lam_lifespan_and_disease.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<a href=\"https://colab.research.google.com/github/genelam26/Lifespan-and-Disease/blob/main/lam_lifespan_and_disease.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>\n",
    "\n",
    "# Gene's Lifespan and Disease Machine Learning Notebook\n",
    "\n",
    "**Requirements:**\n",
    "+ Talk about what datasets you've selected for your notebook\n",
    "+ Include why you choose these datasets\n",
    "+ Talk about the classification algorithm you plan to use\n",
    "+ Mention what you hope to find over the course of this notebook\n",
    "\n",
    "## Abstract\n",
    "For my machine learning notebook, I was vaguely interested in exploring health outcomes related to sleep and Alzheimer's Disease. My interest in the former was because I do not get any sleep and I was wondering how that might impact my potential lifespan (probably negatively). I should note that the dataset I use ended up not answering this question because it related to general mammalian lifespan. I ended up trying to find how sleep (and other factors) may impact the lifespan of a mammal; I still enjoyed working with it. I performed linear regression to predict lifespan based on several feature vectors. I then used k-fold cross validation to improve upon my regression, but spoiler warning, both methods didn't work too well.\n",
    "\n",
    "My interest in the Alzheimer's dataset is out of fear. Out of all the diseases, I fear not remembering the people around me or anything about myself. Computer science is one of the tools I can use to explore this fear without getting too in-depth with the biological mechanisms. I took a biology class at CCNY and found out I was not good at biology... lol. I built a classifier for Alzheimer's to explore how an identification system could potentially work and it did quite well. I used a neural network at first and then an SVM. I hope that in a few years, modern medicine becomes so great that a classifier won't even be necessary for this disease. \n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Section 1: Prepping the Regression datasets\n",
    "\n",
    "**Requirements:**\n",
    "+ At a high level, discuss what columns are included in the data\n",
    "+ Load the dataset you've selected using pandas\n",
    "+ Show the **\"head\"** of the data\n",
    "+ For the columns you will be using in your regression, describe their setup in more detail\n",
    "+ Clean the data if needed using pandas\n",
    "\n",
    "The data comes from the study \"Sleep in Mammals: Ecological and Constitutional Correlates\" by Allison, Truett and Cicchetti, Domenic V. (1976). The following columns are used in the dataset and are characterstics of different mammals:\n",
    "- body weight (kg)\n",
    "- brain weight (g)\n",
    "- maximum life span (years)\n",
    "- gestation time (days)\n",
    "- predation index (1-5) \n",
    "- sleep exposure index (1-5) \n",
    "- overall danger index (1-5) \n",
    "- total sleep\n",
    "\n",
    "Dataset Source: https://www.kaggle.com/mathurinache/sleep-dataset\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Imports for the entire notebook\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt # for data visualization\n",
    "import seaborn as sns # for data visualization\n",
    "\n",
    "from sklearn.model_selection import train_test_split # split the data into our train and test cases\n",
    "from sklearn import metrics # model accuracy\n",
    "from sklearn.metrics import accuracy_score\n",
    "\n",
    "from sklearn.linear_model import LinearRegression # performs linear regression\n",
    "from sklearn.svm import SVR\n",
    "from sklearn.model_selection import KFold\n",
    "from sklearn.model_selection import cross_val_predict\n",
    "\n",
    "from keras.models import Sequential # for da neural network\n",
    "from keras.layers import Dense\n",
    "from sklearn.preprocessing import MinMaxScaler # scales the feature vector from 0 to 1 \n",
    "\n",
    "from sklearn import svm # support vector machine\n",
    "from sklearn.metrics import plot_confusion_matrix # plots the confusion matrix for svm\n",
    "from sklearn.metrics import confusion_matrix\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "ename": "ParserError",
     "evalue": "Error tokenizing data. C error: Expected 1 fields in line 80, saw 2\n",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mParserError\u001b[0m                               Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-4-ec08dba333ee>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[1;32m      1\u001b[0m \u001b[0;31m# Use pandas to load your dataset and output the first 15 rows\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 2\u001b[0;31m \u001b[0mhealth\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mpd\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mread_csv\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m'https://github.com/genelam26/Lifespan-and-Disease/blob/main/dataset_sleep.csv'\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;31m# Loading the dataset using pandas\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m      3\u001b[0m \u001b[0mhealth\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mhead\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;36m15\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;31m# Showing 15 rows of data\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m~/anaconda3/lib/python3.7/site-packages/pandas/io/parsers.py\u001b[0m in \u001b[0;36mparser_f\u001b[0;34m(filepath_or_buffer, sep, delimiter, header, names, index_col, usecols, squeeze, prefix, mangle_dupe_cols, dtype, engine, converters, true_values, false_values, skipinitialspace, skiprows, skipfooter, nrows, na_values, keep_default_na, na_filter, verbose, skip_blank_lines, parse_dates, infer_datetime_format, keep_date_col, date_parser, dayfirst, iterator, chunksize, compression, thousands, decimal, lineterminator, quotechar, quoting, doublequote, escapechar, comment, encoding, dialect, tupleize_cols, error_bad_lines, warn_bad_lines, delim_whitespace, low_memory, memory_map, float_precision)\u001b[0m\n\u001b[1;32m    700\u001b[0m                     skip_blank_lines=skip_blank_lines)\n\u001b[1;32m    701\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 702\u001b[0;31m         \u001b[0;32mreturn\u001b[0m \u001b[0m_read\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mfilepath_or_buffer\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mkwds\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    703\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    704\u001b[0m     \u001b[0mparser_f\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m__name__\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mname\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m~/anaconda3/lib/python3.7/site-packages/pandas/io/parsers.py\u001b[0m in \u001b[0;36m_read\u001b[0;34m(filepath_or_buffer, kwds)\u001b[0m\n\u001b[1;32m    433\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    434\u001b[0m     \u001b[0;32mtry\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 435\u001b[0;31m         \u001b[0mdata\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mparser\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mread\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mnrows\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    436\u001b[0m     \u001b[0;32mfinally\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    437\u001b[0m         \u001b[0mparser\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mclose\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m~/anaconda3/lib/python3.7/site-packages/pandas/io/parsers.py\u001b[0m in \u001b[0;36mread\u001b[0;34m(self, nrows)\u001b[0m\n\u001b[1;32m   1137\u001b[0m     \u001b[0;32mdef\u001b[0m \u001b[0mread\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mnrows\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;32mNone\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1138\u001b[0m         \u001b[0mnrows\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0m_validate_integer\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m'nrows'\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mnrows\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1139\u001b[0;31m         \u001b[0mret\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_engine\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mread\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mnrows\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   1140\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1141\u001b[0m         \u001b[0;31m# May alter columns / col_dict\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m~/anaconda3/lib/python3.7/site-packages/pandas/io/parsers.py\u001b[0m in \u001b[0;36mread\u001b[0;34m(self, nrows)\u001b[0m\n\u001b[1;32m   1993\u001b[0m     \u001b[0;32mdef\u001b[0m \u001b[0mread\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mnrows\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;32mNone\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1994\u001b[0m         \u001b[0;32mtry\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1995\u001b[0;31m             \u001b[0mdata\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_reader\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mread\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mnrows\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   1996\u001b[0m         \u001b[0;32mexcept\u001b[0m \u001b[0mStopIteration\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1997\u001b[0m             \u001b[0;32mif\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_first_chunk\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32mpandas/_libs/parsers.pyx\u001b[0m in \u001b[0;36mpandas._libs.parsers.TextReader.read\u001b[0;34m()\u001b[0m\n",
      "\u001b[0;32mpandas/_libs/parsers.pyx\u001b[0m in \u001b[0;36mpandas._libs.parsers.TextReader._read_low_memory\u001b[0;34m()\u001b[0m\n",
      "\u001b[0;32mpandas/_libs/parsers.pyx\u001b[0m in \u001b[0;36mpandas._libs.parsers.TextReader._read_rows\u001b[0;34m()\u001b[0m\n",
      "\u001b[0;32mpandas/_libs/parsers.pyx\u001b[0m in \u001b[0;36mpandas._libs.parsers.TextReader._tokenize_rows\u001b[0;34m()\u001b[0m\n",
      "\u001b[0;32mpandas/_libs/parsers.pyx\u001b[0m in \u001b[0;36mpandas._libs.parsers.raise_parser_error\u001b[0;34m()\u001b[0m\n",
      "\u001b[0;31mParserError\u001b[0m: Error tokenizing data. C error: Expected 1 fields in line 80, saw 2\n"
     ]
    }
   ],
   "source": [
    "# Use pandas to load your dataset and output the first 15 rows\n",
    "health = pd.read_csv('dataset_sleep.csv') # Loading the dataset using pandas\n",
    "health.head(15) # Showing 15 rows of data"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The original authors of the study seemed to primarily look at sleep, but I want to focus on max lifespan. A possible error that I anticipate is with null values. The dataset's author denoted null with '?' so I will delete all rows with '?' and luckily, we still have 51 values to work with."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 160,
   "metadata": {},
   "outputs": [],
   "source": [
    "health = health[health.max_life_span != '?']\n",
    "health = health[health.gestation_time != '?']\n",
    "health = health[health.total_sleep != '?']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 161,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>body_weight</th>\n",
       "      <th>brain_weight</th>\n",
       "      <th>predation_index</th>\n",
       "      <th>sleep_exposure_index</th>\n",
       "      <th>danger_index</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <td>51.000000</td>\n",
       "      <td>51.000000</td>\n",
       "      <td>51.000000</td>\n",
       "      <td>51.000000</td>\n",
       "      <td>51.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mean</th>\n",
       "      <td>224.363078</td>\n",
       "      <td>317.502353</td>\n",
       "      <td>2.843137</td>\n",
       "      <td>2.450980</td>\n",
       "      <td>2.627451</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>std</th>\n",
       "      <td>988.465573</td>\n",
       "      <td>1018.888421</td>\n",
       "      <td>1.488255</td>\n",
       "      <td>1.578781</td>\n",
       "      <td>1.427736</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>min</th>\n",
       "      <td>0.005000</td>\n",
       "      <td>0.140000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25%</th>\n",
       "      <td>0.615000</td>\n",
       "      <td>4.500000</td>\n",
       "      <td>2.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50%</th>\n",
       "      <td>3.385000</td>\n",
       "      <td>21.000000</td>\n",
       "      <td>3.000000</td>\n",
       "      <td>2.000000</td>\n",
       "      <td>2.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>75%</th>\n",
       "      <td>44.245000</td>\n",
       "      <td>172.000000</td>\n",
       "      <td>4.000000</td>\n",
       "      <td>4.000000</td>\n",
       "      <td>4.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>max</th>\n",
       "      <td>6654.000000</td>\n",
       "      <td>5712.000000</td>\n",
       "      <td>5.000000</td>\n",
       "      <td>5.000000</td>\n",
       "      <td>5.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       body_weight  brain_weight  predation_index  sleep_exposure_index  \\\n",
       "count    51.000000     51.000000        51.000000             51.000000   \n",
       "mean    224.363078    317.502353         2.843137              2.450980   \n",
       "std     988.465573   1018.888421         1.488255              1.578781   \n",
       "min       0.005000      0.140000         1.000000              1.000000   \n",
       "25%       0.615000      4.500000         2.000000              1.000000   \n",
       "50%       3.385000     21.000000         3.000000              2.000000   \n",
       "75%      44.245000    172.000000         4.000000              4.000000   \n",
       "max    6654.000000   5712.000000         5.000000              5.000000   \n",
       "\n",
       "       danger_index  \n",
       "count     51.000000  \n",
       "mean       2.627451  \n",
       "std        1.427736  \n",
       "min        1.000000  \n",
       "25%        1.000000  \n",
       "50%        2.000000  \n",
       "75%        4.000000  \n",
       "max        5.000000  "
      ]
     },
     "execution_count": 161,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "health.describe() # basic statistical overview of dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 162,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "body_weight             0\n",
       "brain_weight            0\n",
       "max_life_span           0\n",
       "gestation_time          0\n",
       "predation_index         0\n",
       "sleep_exposure_index    0\n",
       "danger_index            0\n",
       "total_sleep             0\n",
       "dtype: int64"
      ]
     },
     "execution_count": 162,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Use pandas to clean out data cells that have invalid values for columns of interest\n",
    "health.isnull().sum() # checks for null values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 163,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0"
      ]
     },
     "execution_count": 163,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "health.duplicated().sum() # checks for duplicates"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We now have no duplicates and no null values so that's good. Since I am looking at max lifespan, I will use the rest of the columns as my independent variables for my linear regression. I will move my max_life_span column to the end of the table. I will also expand on my descriptions for all the columns and why I think they would influence a mammal's max lifespan.\n",
    "- body weight (kg) - a higher body weight might indicate higher caloric expenditure, meaning more storage energy for the animal to use over its lifetime.\n",
    "- brain weight (g) - heavier brain => smarter? And the smarter a mammal is, the longer it will survive\n",
    "- gestation time (days) - the longer you take to digest, perhaps the more energy you will have, and then the longer you'll live.\n",
    "- predation index (1-5: with 5 being most likely to be preyed upon) - smaller the PI, the less likely you'll be eaten, the longer you'll live\n",
    "- sleep exposure index (1-5: with 5 being most exposed while sleeping) - smaller the SEI, the less likely you'll be eaten, the longer you'll live\n",
    "- overall danger index (1-5: with 5 being at most danger from other animals) - smaller the ODI, the less likely you'll be eaten, the longer you'll live\n",
    "- total sleep - more sleep, more energy, longer life\n",
    "- maximum life span (years) - our dependent variable\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 164,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>body_weight</th>\n",
       "      <th>brain_weight</th>\n",
       "      <th>gestation_time</th>\n",
       "      <th>predation_index</th>\n",
       "      <th>sleep_exposure_index</th>\n",
       "      <th>danger_index</th>\n",
       "      <th>total_sleep</th>\n",
       "      <th>max_life_span</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>6654.000</td>\n",
       "      <td>5712.0</td>\n",
       "      <td>645</td>\n",
       "      <td>3</td>\n",
       "      <td>5</td>\n",
       "      <td>3</td>\n",
       "      <td>3.3</td>\n",
       "      <td>38.6</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1.000</td>\n",
       "      <td>6.6</td>\n",
       "      <td>42</td>\n",
       "      <td>3</td>\n",
       "      <td>1</td>\n",
       "      <td>3</td>\n",
       "      <td>8.3</td>\n",
       "      <td>4.5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3.385</td>\n",
       "      <td>44.5</td>\n",
       "      <td>60</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>12.5</td>\n",
       "      <td>14</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>2547.000</td>\n",
       "      <td>4603.0</td>\n",
       "      <td>624</td>\n",
       "      <td>3</td>\n",
       "      <td>5</td>\n",
       "      <td>4</td>\n",
       "      <td>3.9</td>\n",
       "      <td>69</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>10.550</td>\n",
       "      <td>179.5</td>\n",
       "      <td>180</td>\n",
       "      <td>4</td>\n",
       "      <td>4</td>\n",
       "      <td>4</td>\n",
       "      <td>9.8</td>\n",
       "      <td>27</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>0.023</td>\n",
       "      <td>0.3</td>\n",
       "      <td>35</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>19.7</td>\n",
       "      <td>19</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>160.000</td>\n",
       "      <td>169.0</td>\n",
       "      <td>392</td>\n",
       "      <td>4</td>\n",
       "      <td>5</td>\n",
       "      <td>4</td>\n",
       "      <td>6.2</td>\n",
       "      <td>30.4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>3.300</td>\n",
       "      <td>25.6</td>\n",
       "      <td>63</td>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>14.5</td>\n",
       "      <td>28</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>52.160</td>\n",
       "      <td>440.0</td>\n",
       "      <td>230</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>9.7</td>\n",
       "      <td>50</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>0.425</td>\n",
       "      <td>6.4</td>\n",
       "      <td>112</td>\n",
       "      <td>5</td>\n",
       "      <td>4</td>\n",
       "      <td>4</td>\n",
       "      <td>12.5</td>\n",
       "      <td>7</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>465.000</td>\n",
       "      <td>423.0</td>\n",
       "      <td>281</td>\n",
       "      <td>5</td>\n",
       "      <td>5</td>\n",
       "      <td>5</td>\n",
       "      <td>3.9</td>\n",
       "      <td>30</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>187.100</td>\n",
       "      <td>419.0</td>\n",
       "      <td>365</td>\n",
       "      <td>5</td>\n",
       "      <td>5</td>\n",
       "      <td>5</td>\n",
       "      <td>3.1</td>\n",
       "      <td>40</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>0.075</td>\n",
       "      <td>1.2</td>\n",
       "      <td>42</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>8.4</td>\n",
       "      <td>3.5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>3.000</td>\n",
       "      <td>25.0</td>\n",
       "      <td>28</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>8.6</td>\n",
       "      <td>50</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>0.785</td>\n",
       "      <td>3.5</td>\n",
       "      <td>42</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>10.7</td>\n",
       "      <td>6</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    body_weight  brain_weight gestation_time  predation_index  \\\n",
       "0      6654.000        5712.0            645                3   \n",
       "1         1.000           6.6             42                3   \n",
       "2         3.385          44.5             60                1   \n",
       "4      2547.000        4603.0            624                3   \n",
       "5        10.550         179.5            180                4   \n",
       "6         0.023           0.3             35                1   \n",
       "7       160.000         169.0            392                4   \n",
       "8         3.300          25.6             63                1   \n",
       "9        52.160         440.0            230                1   \n",
       "10        0.425           6.4            112                5   \n",
       "11      465.000         423.0            281                5   \n",
       "13      187.100         419.0            365                5   \n",
       "14        0.075           1.2             42                1   \n",
       "15        3.000          25.0             28                2   \n",
       "16        0.785           3.5             42                2   \n",
       "\n",
       "    sleep_exposure_index  danger_index total_sleep max_life_span  \n",
       "0                      5             3         3.3          38.6  \n",
       "1                      1             3         8.3           4.5  \n",
       "2                      1             1        12.5            14  \n",
       "4                      5             4         3.9            69  \n",
       "5                      4             4         9.8            27  \n",
       "6                      1             1        19.7            19  \n",
       "7                      5             4         6.2          30.4  \n",
       "8                      2             1        14.5            28  \n",
       "9                      1             1         9.7            50  \n",
       "10                     4             4        12.5             7  \n",
       "11                     5             5         3.9            30  \n",
       "13                     5             5         3.1            40  \n",
       "14                     1             1         8.4           3.5  \n",
       "15                     2             2         8.6            50  \n",
       "16                     2             2        10.7             6  "
      ]
     },
     "execution_count": 164,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "health.insert(7, 'max_life_span', health.pop('max_life_span'))\n",
    "health.head(15)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 165,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'pandas.core.frame.DataFrame'>\n",
      "Int64Index: 51 entries, 0 to 60\n",
      "Data columns (total 8 columns):\n",
      "body_weight             51 non-null float64\n",
      "brain_weight            51 non-null float64\n",
      "gestation_time          51 non-null object\n",
      "predation_index         51 non-null int64\n",
      "sleep_exposure_index    51 non-null int64\n",
      "danger_index            51 non-null int64\n",
      "total_sleep             51 non-null object\n",
      "max_life_span           51 non-null object\n",
      "dtypes: float64(2), int64(3), object(3)\n",
      "memory usage: 3.6+ KB\n"
     ]
    }
   ],
   "source": [
    "health.info()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Looks like there are some object-type columns that we want to convert to ints or floats."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 166,
   "metadata": {},
   "outputs": [],
   "source": [
    "health['gestation_time']=health['gestation_time'].astype('float')\n",
    "health['total_sleep']=health['total_sleep'].astype('float')\n",
    "health['max_life_span']=health['max_life_span'].astype('float')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 167,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'pandas.core.frame.DataFrame'>\n",
      "Int64Index: 51 entries, 0 to 60\n",
      "Data columns (total 8 columns):\n",
      "body_weight             51 non-null float64\n",
      "brain_weight            51 non-null float64\n",
      "gestation_time          51 non-null float64\n",
      "predation_index         51 non-null int64\n",
      "sleep_exposure_index    51 non-null int64\n",
      "danger_index            51 non-null int64\n",
      "total_sleep             51 non-null float64\n",
      "max_life_span           51 non-null float64\n",
      "dtypes: float64(5), int64(3)\n",
      "memory usage: 3.6 KB\n"
     ]
    }
   ],
   "source": [
    "health.info()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# nice."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Section 2: Linear Regression\n",
    "**Requirements:**\n",
    "+ Split your cleaned dataset using sklearn into a training and test set\n",
    "+ Fit a linear regression to your training set\n",
    "+ Visualize the regressor using matplotlib\n",
    "+ Report on the error rate for your test set\n",
    "+ Comment on the result of this regression. Did it work well or not? Provide justification why or why not.\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 168,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Use pandas to extract features and prediction sets from our dataframe\n",
    "X0=health.iloc[:,0:7]\n",
    "y0=health.iloc[:,7]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Nice and then we use sklearn to create a training set and test set. I will reserve 20% of the dataset for testing."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 169,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Use sklearn to split our dataset\n",
    "X_train0, X_test0, y_train0, y_test0 = train_test_split(X0, y0, test_size = 0.2, random_state = 1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's use that training set and fit a Linear Regression to it:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 170,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[-7.40628995e-02  8.96067469e-02 -2.15528341e-03 -4.31519215e+00\n",
      "  4.10787093e+00  2.16030799e+00  1.28551443e-01]\n",
      "8.014156553479896\n"
     ]
    }
   ],
   "source": [
    "# Use sklearn to create and Linear Regressor and fit our data to it\n",
    "model = LinearRegression()\n",
    "model.fit(X_train0, y_train0)\n",
    "print(model.coef_)\n",
    "print(model.intercept_)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The model coefficients are our weights and the intercept is our y-intercept in y=mx+b. To make sure we're on the right track let's try a couple made up sample inputs and see what our model returns:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 171,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "27      7.6\n",
      "2      14.0\n",
      "59     24.0\n",
      "32     24.0\n",
      "42     20.2\n",
      "53     20.0\n",
      "38      5.0\n",
      "60      3.0\n",
      "41     12.0\n",
      "33    100.0\n",
      "4      69.0\n",
      "Name: max_life_span, dtype: float64\n",
      "[  8.72641482  15.18151669  17.47657083  12.43921394  26.4250631\n",
      "  29.51282789   8.55860135   8.20592138  15.15711935 124.10910034\n",
      " 247.2272715 ]\n"
     ]
    }
   ],
   "source": [
    "# experiment with the regressor\n",
    "y_pred0 = model.predict(X_test0)\n",
    "print(y_test0)\n",
    "print(y_pred0)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "So the values for y_test0 seem to match with our y_pred0 values. The y_test0 is our actual values whereas the y_pred0 are the values we get from our linear regression model. We all love data visualizations so let's create a plot of our linear regression and our data. I will have the training data be blue dots, the test data green dots, with a red regression line cutting through the data."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 172,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1440x1800 with 8 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Use matplotlib (or other visualization lib) to create a scatter plot and the regression to create a regression line\n",
    "\n",
    "fig, axs = plt.subplots(4,2, figsize=(20, 25), facecolor='w', edgecolor='k')\n",
    "fig.subplots_adjust(hspace = .3, wspace=.1)\n",
    "axs = axs.ravel()\n",
    "for x in range(7):\n",
    "\n",
    "    # Training Data\n",
    "    axs[x].scatter(X_train0.iloc[:,x], y_train0, color = \"blue\")\n",
    "    # Testing Data\n",
    "    axs[x].scatter(X_test0.iloc[:,x], y_test0, color = \"green\")\n",
    "    # Labels\n",
    "    axs[x].set_ylabel(\"Life Span Outcome in Years\")\n",
    "    axs[x].set_xlabel(health.columns.values[x])\n",
    "\n",
    "    # Plotting the trained regression line\n",
    "    numbers = np.linspace(-1000,1000,1000)\n",
    "    y = numbers*model.coef_[x] + model.intercept_\n",
    "    axs[x].plot(numbers, y, color=\"red\")\n",
    "    axs[x].set_ylim(0,100)\n",
    "\n",
    "    if(x==0):\n",
    "        axs[x].set_xlim(-5,25)\n",
    "    elif(x==1):\n",
    "        axs[x].set_xlim(-5,200)\n",
    "    elif(x==2):\n",
    "        axs[x].set_xlim(0,250)\n",
    "    elif(x==6):\n",
    "        axs[x].set_xlim(0,25)\n",
    "    else:\n",
    "        axs[x].set_xlim(0,6)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 173,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Mean Absolute Error: 22.762550148938207\n",
      "Mean Squared Error: 2973.0962442898167\n",
      "Root Mean Squared Error: 54.526106080388836\n",
      "R-Squared Score: -2.619184726733329\n"
     ]
    }
   ],
   "source": [
    "# Use linear regression model to report accuracy/error rate on the test set\n",
    "print('Mean Absolute Error:', metrics.mean_absolute_error(y_test0, y_pred0))\n",
    "print('Mean Squared Error:', metrics.mean_squared_error(y_test0, y_pred0))\n",
    "print('Root Mean Squared Error:', np.sqrt(metrics.mean_squared_error(y_test0, y_pred0)))\n",
    "print('R-Squared Score:', metrics.r2_score(y_test0, y_pred0))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "I was shocked upon seeing the $r^2$ score. I didn't even think it was possible. I looked online and saw a couple of explanations. One on StackOverflow explains that the equation for r^2:\n",
    "$$1 - \\frac{(\\text{residual sum of squares})}{(\\text{total sum of squares})}$$\n",
    "makes it so that if your denominator is small, you get a high negative number. Another explained that a low, negative value means that I would have gotten a better predictive score if I just used the mean each time. Still, I plugged both y_test_actual and y_predicted into an online $r^2$ calculator and got an $r^2$ value of $0.641$, which I don't think is too bad."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "![title]('picture.png')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "A low $r^2$ value means that there is a weak correlation between our feature vectors and the lifespan of a mammal. One possibility could be that I included a feature vector like brain size that negatively impacted the accuracy of my model. I wanted to do another analysis, though, so in my next step, I will use k-fold cross validation where k=10. \n",
    "\n",
    "We first normalize our dataset from 0 to 1. This will help with outliers."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 182,
   "metadata": {},
   "outputs": [],
   "source": [
    "scaler = MinMaxScaler(feature_range=(0, 1))\n",
    "X0 = scaler.fit_transform(X0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 183,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([-0.3269437 , -0.88773445, -0.04229505, -0.20505994,  0.23121037,\n",
       "       -0.06807492,  0.28881369,  0.24739211, -0.16605203, -0.27652334])"
      ]
     },
     "execution_count": 183,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "best_svr = SVR(kernel='rbf') # I used SVR because I saw that SVM was used for classification problems whereas SVR was used for regression\n",
    "cross_val_score(best_svr, X0, y0, cv=10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 184,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[16.09859564 16.44438022 16.49191585 16.28761802 13.27529151 13.94784831\n",
      " 13.26288597 14.7292746  14.02516833 13.33000172 13.65528875 13.23695594\n",
      " 12.8316654  12.82566611 15.52812628 13.47314841 14.6736127  14.68220532\n",
      " 14.81001957 16.9353125  13.25531636 13.37680138 14.16494776 15.06475472\n",
      " 13.23249005 12.7807203  14.03147091 12.90046102 12.61722218 12.69458705\n",
      " 16.93857627 20.33085268 13.39995161 13.26868896 15.35000417 13.39253371\n",
      " 14.43062085 12.80548929 13.22829977 12.9366862 ]\n"
     ]
    }
   ],
   "source": [
    "new_prediction = cross_val_predict(best_svr, X_train0, y_train0, cv=10)\n",
    "print(new_prediction)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 185,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "R-Squared Score: -0.018051927294353787\n"
     ]
    }
   ],
   "source": [
    "print('R-Squared Score:', metrics.r2_score(y_train0, new_prediction))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "LOL. I guess that is better than our linear regression? Honestly, I am not too sure what happened there. I inputted the training dataset into the online $r^2$ calculator again, but this time, it did not help.\n",
    "\n",
    "![title]('picture2.png')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Conclusions - Lifespan Predictor\n",
    "**Requirements:**\n",
    "+ Summarize the findings of your report\n",
    "+ Repeat your methodology and key findings for Section 2. and Section 4.\n",
    "+ Highlight what you found interesting\n",
    "+ Discuss what you would do to extend the project further\n",
    "\n",
    "I think we can safely say, from trying two different methods, that there is a weak correlation between our feature vectors and the lifespan of our mammal. At first, I used linear regression to plot each feature vector against the lifespan of a mammal. Our accuracy scores turned out really bad - actually - surprisingly bad. I thought that $r^2$ values can only go from 0 to 1, but my model was so poor that I got -2.6. I then tried using cross-validation method, a model often used to prevent overfitting, but our $r^2$ improved from -2.6 to -0.18. Future studies should use a larger dataset over a diverse group of mammals. Also looking at our plots, there seems to be a weak correlation between sleep, body weight, and predation index with lifespan. We don't want to fit our data to match the regression by leaving those out. Instead, we can prioritize the research in those areas to see if there's actually a causation occurring. "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Section 3: Prepping the classification dataset\n",
    "**Requirements:**\n",
    "+ At a high level, discuss what columns are included in the data\n",
    "+ Load the dataset you've selected using pandas\n",
    "+ Show the **head** of the data\n",
    "+ For the columns you will be using in your classification, describe the range of values\n",
    "+ For those columns map the values to a set of integers (if they aren't already)\n",
    "+ Clean the data using pandas if needed\n",
    "\n",
    "For my classificaiton dataset, I am using an Alzheimer's Features dataset. Here are the columns:\n",
    "\n",
    "- Group - Demented or Nondemented\n",
    "- Sex - Male or Female\n",
    "- Age in Years\n",
    "- EDUC - Years of Education\n",
    "- SES - Socioeconomic Status from 1 to 5 with 5 being High\n",
    "- MMSE - Mini Mental State Examination\n",
    "- CDR - Clinical Dementia Rating\n",
    "- eTIV - Estimated total intracranial volume\n",
    "- nWBV - Normalize Whole Brain Volume\n",
    "- ASF - Atlas Scaling Factor\n",
    "\n",
    "Dataset Source: https://www.kaggle.com/brsdincer/alzheimer-features"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 187,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Group</th>\n",
       "      <th>M/F</th>\n",
       "      <th>Age</th>\n",
       "      <th>EDUC</th>\n",
       "      <th>SES</th>\n",
       "      <th>MMSE</th>\n",
       "      <th>CDR</th>\n",
       "      <th>eTIV</th>\n",
       "      <th>nWBV</th>\n",
       "      <th>ASF</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Nondemented</td>\n",
       "      <td>M</td>\n",
       "      <td>87</td>\n",
       "      <td>14</td>\n",
       "      <td>2.0</td>\n",
       "      <td>27.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1987</td>\n",
       "      <td>0.696</td>\n",
       "      <td>0.883</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Nondemented</td>\n",
       "      <td>M</td>\n",
       "      <td>88</td>\n",
       "      <td>14</td>\n",
       "      <td>2.0</td>\n",
       "      <td>30.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>2004</td>\n",
       "      <td>0.681</td>\n",
       "      <td>0.876</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>Demented</td>\n",
       "      <td>M</td>\n",
       "      <td>75</td>\n",
       "      <td>12</td>\n",
       "      <td>NaN</td>\n",
       "      <td>23.0</td>\n",
       "      <td>0.5</td>\n",
       "      <td>1678</td>\n",
       "      <td>0.736</td>\n",
       "      <td>1.046</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Demented</td>\n",
       "      <td>M</td>\n",
       "      <td>76</td>\n",
       "      <td>12</td>\n",
       "      <td>NaN</td>\n",
       "      <td>28.0</td>\n",
       "      <td>0.5</td>\n",
       "      <td>1738</td>\n",
       "      <td>0.713</td>\n",
       "      <td>1.010</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>Demented</td>\n",
       "      <td>M</td>\n",
       "      <td>80</td>\n",
       "      <td>12</td>\n",
       "      <td>NaN</td>\n",
       "      <td>22.0</td>\n",
       "      <td>0.5</td>\n",
       "      <td>1698</td>\n",
       "      <td>0.701</td>\n",
       "      <td>1.034</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>Nondemented</td>\n",
       "      <td>F</td>\n",
       "      <td>88</td>\n",
       "      <td>18</td>\n",
       "      <td>3.0</td>\n",
       "      <td>28.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1215</td>\n",
       "      <td>0.710</td>\n",
       "      <td>1.444</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>Nondemented</td>\n",
       "      <td>F</td>\n",
       "      <td>90</td>\n",
       "      <td>18</td>\n",
       "      <td>3.0</td>\n",
       "      <td>27.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1200</td>\n",
       "      <td>0.718</td>\n",
       "      <td>1.462</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>Nondemented</td>\n",
       "      <td>M</td>\n",
       "      <td>80</td>\n",
       "      <td>12</td>\n",
       "      <td>4.0</td>\n",
       "      <td>28.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1689</td>\n",
       "      <td>0.712</td>\n",
       "      <td>1.039</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>Nondemented</td>\n",
       "      <td>M</td>\n",
       "      <td>83</td>\n",
       "      <td>12</td>\n",
       "      <td>4.0</td>\n",
       "      <td>29.0</td>\n",
       "      <td>0.5</td>\n",
       "      <td>1701</td>\n",
       "      <td>0.711</td>\n",
       "      <td>1.032</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>Nondemented</td>\n",
       "      <td>M</td>\n",
       "      <td>85</td>\n",
       "      <td>12</td>\n",
       "      <td>4.0</td>\n",
       "      <td>30.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1699</td>\n",
       "      <td>0.705</td>\n",
       "      <td>1.033</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>Demented</td>\n",
       "      <td>M</td>\n",
       "      <td>71</td>\n",
       "      <td>16</td>\n",
       "      <td>NaN</td>\n",
       "      <td>28.0</td>\n",
       "      <td>0.5</td>\n",
       "      <td>1357</td>\n",
       "      <td>0.748</td>\n",
       "      <td>1.293</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>Demented</td>\n",
       "      <td>M</td>\n",
       "      <td>73</td>\n",
       "      <td>16</td>\n",
       "      <td>NaN</td>\n",
       "      <td>27.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1365</td>\n",
       "      <td>0.727</td>\n",
       "      <td>1.286</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>Demented</td>\n",
       "      <td>M</td>\n",
       "      <td>75</td>\n",
       "      <td>16</td>\n",
       "      <td>NaN</td>\n",
       "      <td>27.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1372</td>\n",
       "      <td>0.710</td>\n",
       "      <td>1.279</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>Nondemented</td>\n",
       "      <td>F</td>\n",
       "      <td>93</td>\n",
       "      <td>14</td>\n",
       "      <td>2.0</td>\n",
       "      <td>30.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1272</td>\n",
       "      <td>0.698</td>\n",
       "      <td>1.380</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>Nondemented</td>\n",
       "      <td>F</td>\n",
       "      <td>95</td>\n",
       "      <td>14</td>\n",
       "      <td>2.0</td>\n",
       "      <td>29.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1257</td>\n",
       "      <td>0.703</td>\n",
       "      <td>1.396</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "          Group M/F  Age  EDUC  SES  MMSE  CDR  eTIV   nWBV    ASF\n",
       "0   Nondemented   M   87    14  2.0  27.0  0.0  1987  0.696  0.883\n",
       "1   Nondemented   M   88    14  2.0  30.0  0.0  2004  0.681  0.876\n",
       "2      Demented   M   75    12  NaN  23.0  0.5  1678  0.736  1.046\n",
       "3      Demented   M   76    12  NaN  28.0  0.5  1738  0.713  1.010\n",
       "4      Demented   M   80    12  NaN  22.0  0.5  1698  0.701  1.034\n",
       "5   Nondemented   F   88    18  3.0  28.0  0.0  1215  0.710  1.444\n",
       "6   Nondemented   F   90    18  3.0  27.0  0.0  1200  0.718  1.462\n",
       "7   Nondemented   M   80    12  4.0  28.0  0.0  1689  0.712  1.039\n",
       "8   Nondemented   M   83    12  4.0  29.0  0.5  1701  0.711  1.032\n",
       "9   Nondemented   M   85    12  4.0  30.0  0.0  1699  0.705  1.033\n",
       "10     Demented   M   71    16  NaN  28.0  0.5  1357  0.748  1.293\n",
       "11     Demented   M   73    16  NaN  27.0  1.0  1365  0.727  1.286\n",
       "12     Demented   M   75    16  NaN  27.0  1.0  1372  0.710  1.279\n",
       "13  Nondemented   F   93    14  2.0  30.0  0.0  1272  0.698  1.380\n",
       "14  Nondemented   F   95    14  2.0  29.0  0.0  1257  0.703  1.396"
      ]
     },
     "execution_count": 187,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Use pandas to map categories to ints, summarize your mapping below\n",
    "disease = pd.read_csv('alzheimer.csv') # loads\n",
    "disease.head(15) # gives head"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 188,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Group     0\n",
       "M/F       0\n",
       "Age       0\n",
       "EDUC      0\n",
       "SES      19\n",
       "MMSE      2\n",
       "CDR       0\n",
       "eTIV      0\n",
       "nWBV      0\n",
       "ASF       0\n",
       "dtype: int64"
      ]
     },
     "execution_count": 188,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "disease.isnull().sum()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "It seems like social-economic status, while being a key indicator in detecting Alzheimer's, would give us some trouble, along with MMSE, the cognitive assessment score. Luckily there is another assessment, CDR score, so I will use that. I plan on using Age, Education, and CDR to classify Alzheimers.\n",
    "\n",
    "Age is written in years and the data goes from 60 to 98 years old. We can expect that the older you are, the more likely you would be susceptible to Alzheimer's.\n",
    "Education is reflected in years as well and goes from 6 to 23. We can expect that the higher your education levels, the less likely you would be to have Alzheimer's.\n",
    "CDR is the other cognitive assessment for characterizing and tracking a patient's level of impairment/dementia:\n",
    "- 0 = Normal\n",
    "- 0.5 = Very Mild Dementia\n",
    "- 1 = Mild Dementia\n",
    "- 2 = Moderate Dementia\n",
    "- 3 = Severe Dementia\n",
    "\n",
    "The CDR score probably gives the most indication of Alzheimer's as the patient is given a mental test."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 189,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Group</th>\n",
       "      <th>Age</th>\n",
       "      <th>EDUC</th>\n",
       "      <th>CDR</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Nondemented</td>\n",
       "      <td>87</td>\n",
       "      <td>14</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Nondemented</td>\n",
       "      <td>88</td>\n",
       "      <td>14</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>Demented</td>\n",
       "      <td>75</td>\n",
       "      <td>12</td>\n",
       "      <td>0.5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Demented</td>\n",
       "      <td>76</td>\n",
       "      <td>12</td>\n",
       "      <td>0.5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>Demented</td>\n",
       "      <td>80</td>\n",
       "      <td>12</td>\n",
       "      <td>0.5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>Nondemented</td>\n",
       "      <td>88</td>\n",
       "      <td>18</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>Nondemented</td>\n",
       "      <td>90</td>\n",
       "      <td>18</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>Nondemented</td>\n",
       "      <td>80</td>\n",
       "      <td>12</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>Nondemented</td>\n",
       "      <td>83</td>\n",
       "      <td>12</td>\n",
       "      <td>0.5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>Nondemented</td>\n",
       "      <td>85</td>\n",
       "      <td>12</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>Demented</td>\n",
       "      <td>71</td>\n",
       "      <td>16</td>\n",
       "      <td>0.5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>Demented</td>\n",
       "      <td>73</td>\n",
       "      <td>16</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>Demented</td>\n",
       "      <td>75</td>\n",
       "      <td>16</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>Nondemented</td>\n",
       "      <td>93</td>\n",
       "      <td>14</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>Nondemented</td>\n",
       "      <td>95</td>\n",
       "      <td>14</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "          Group  Age  EDUC  CDR\n",
       "0   Nondemented   87    14  0.0\n",
       "1   Nondemented   88    14  0.0\n",
       "2      Demented   75    12  0.5\n",
       "3      Demented   76    12  0.5\n",
       "4      Demented   80    12  0.5\n",
       "5   Nondemented   88    18  0.0\n",
       "6   Nondemented   90    18  0.0\n",
       "7   Nondemented   80    12  0.0\n",
       "8   Nondemented   83    12  0.5\n",
       "9   Nondemented   85    12  0.0\n",
       "10     Demented   71    16  0.5\n",
       "11     Demented   73    16  1.0\n",
       "12     Demented   75    16  1.0\n",
       "13  Nondemented   93    14  0.0\n",
       "14  Nondemented   95    14  0.0"
      ]
     },
     "execution_count": 189,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "del disease[\"M/F\"]\n",
    "del disease[\"SES\"]\n",
    "del disease[\"MMSE\"]\n",
    "del disease[\"eTIV\"]\n",
    "del disease[\"nWBV\"]\n",
    "del disease[\"ASF\"]\n",
    "disease.head(15)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "This gives us our one qualititative feature, the group, and quantitative features, age, years of education, and CDR score. Now we we need to map our group to a number. One of the Group values is \"Converted\", which is not documented in the original dataset so I just delete all rows with that value for simplicity sake."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 190,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Group</th>\n",
       "      <th>Age</th>\n",
       "      <th>EDUC</th>\n",
       "      <th>CDR</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>87</td>\n",
       "      <td>14</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0</td>\n",
       "      <td>88</td>\n",
       "      <td>14</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>1</td>\n",
       "      <td>75</td>\n",
       "      <td>12</td>\n",
       "      <td>0.5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>1</td>\n",
       "      <td>76</td>\n",
       "      <td>12</td>\n",
       "      <td>0.5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>1</td>\n",
       "      <td>80</td>\n",
       "      <td>12</td>\n",
       "      <td>0.5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>0</td>\n",
       "      <td>88</td>\n",
       "      <td>18</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>0</td>\n",
       "      <td>90</td>\n",
       "      <td>18</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>0</td>\n",
       "      <td>80</td>\n",
       "      <td>12</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>0</td>\n",
       "      <td>83</td>\n",
       "      <td>12</td>\n",
       "      <td>0.5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>0</td>\n",
       "      <td>85</td>\n",
       "      <td>12</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>1</td>\n",
       "      <td>71</td>\n",
       "      <td>16</td>\n",
       "      <td>0.5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>1</td>\n",
       "      <td>73</td>\n",
       "      <td>16</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>1</td>\n",
       "      <td>75</td>\n",
       "      <td>16</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>0</td>\n",
       "      <td>93</td>\n",
       "      <td>14</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>0</td>\n",
       "      <td>95</td>\n",
       "      <td>14</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   Group  Age  EDUC  CDR\n",
       "0      0   87    14  0.0\n",
       "1      0   88    14  0.0\n",
       "2      1   75    12  0.5\n",
       "3      1   76    12  0.5\n",
       "4      1   80    12  0.5\n",
       "5      0   88    18  0.0\n",
       "6      0   90    18  0.0\n",
       "7      0   80    12  0.0\n",
       "8      0   83    12  0.5\n",
       "9      0   85    12  0.0\n",
       "10     1   71    16  0.5\n",
       "11     1   73    16  1.0\n",
       "12     1   75    16  1.0\n",
       "13     0   93    14  0.0\n",
       "14     0   95    14  0.0"
      ]
     },
     "execution_count": 190,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "disease[\"Group\"]=disease[\"Group\"].replace({\"Demented\":1, \"Nondemented\":0})\n",
    "disease = disease[disease.Group != \"Converted\"]\n",
    "disease.head(15)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Great! Now we have our data correctly defined. Here's a summary of the data mapping:\n",
    "\n",
    "- 0 = Nondemented\n",
    "- 1 = Demented"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 191,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'pandas.core.frame.DataFrame'>\n",
      "Int64Index: 336 entries, 0 to 372\n",
      "Data columns (total 4 columns):\n",
      "Group    336 non-null object\n",
      "Age      336 non-null int64\n",
      "EDUC     336 non-null int64\n",
      "CDR      336 non-null float64\n",
      "dtypes: float64(1), int64(2), object(1)\n",
      "memory usage: 13.1+ KB\n"
     ]
    }
   ],
   "source": [
    "disease.info()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Uh oh! Our group is of the type - object. Lemme change that real quick and then check for other problems."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 192,
   "metadata": {},
   "outputs": [],
   "source": [
    "disease['Group']=disease['Group'].astype('int')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 193,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'pandas.core.frame.DataFrame'>\n",
      "Int64Index: 336 entries, 0 to 372\n",
      "Data columns (total 4 columns):\n",
      "Group    336 non-null int64\n",
      "Age      336 non-null int64\n",
      "EDUC     336 non-null int64\n",
      "CDR      336 non-null float64\n",
      "dtypes: float64(1), int64(3)\n",
      "memory usage: 13.1 KB\n"
     ]
    }
   ],
   "source": [
    "disease.info()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 194,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Group    0\n",
       "Age      0\n",
       "EDUC     0\n",
       "CDR      0\n",
       "dtype: int64"
      ]
     },
     "execution_count": 194,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "disease.isnull().sum()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 195,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Group</th>\n",
       "      <th>Age</th>\n",
       "      <th>EDUC</th>\n",
       "      <th>CDR</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <td>336.000000</td>\n",
       "      <td>336.000000</td>\n",
       "      <td>336.000000</td>\n",
       "      <td>336.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mean</th>\n",
       "      <td>0.434524</td>\n",
       "      <td>76.711310</td>\n",
       "      <td>14.502976</td>\n",
       "      <td>0.294643</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>std</th>\n",
       "      <td>0.496434</td>\n",
       "      <td>7.614793</td>\n",
       "      <td>2.900333</td>\n",
       "      <td>0.385678</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>min</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>60.000000</td>\n",
       "      <td>6.000000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25%</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>71.000000</td>\n",
       "      <td>12.000000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50%</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>76.000000</td>\n",
       "      <td>14.000000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>75%</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>82.000000</td>\n",
       "      <td>16.000000</td>\n",
       "      <td>0.500000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>max</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>98.000000</td>\n",
       "      <td>23.000000</td>\n",
       "      <td>2.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "            Group         Age        EDUC         CDR\n",
       "count  336.000000  336.000000  336.000000  336.000000\n",
       "mean     0.434524   76.711310   14.502976    0.294643\n",
       "std      0.496434    7.614793    2.900333    0.385678\n",
       "min      0.000000   60.000000    6.000000    0.000000\n",
       "25%      0.000000   71.000000   12.000000    0.000000\n",
       "50%      0.000000   76.000000   14.000000    0.000000\n",
       "75%      1.000000   82.000000   16.000000    0.500000\n",
       "max      1.000000   98.000000   23.000000    2.000000"
      ]
     },
     "execution_count": 195,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "disease.describe()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Sexy"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Section 4: Classification with a Neural Network (or classifier of choice)\n",
    "**Requirements:**\n",
    "+ Split your cleaned dataset using sklearn into a training and test set\n",
    "+ **(Situational)** Scale your data to prevent overfitting\n",
    "+ Fit a classifier to your training set\n",
    "+ **(Optional)** visualize the classifier on a data plot\n",
    "+ Report on the error rate for your test set\n",
    "+ Comment on the result of this regression. Did it work well or not? Provide justification why or why not.\n",
    "\n",
    "I will be using a neural network for the classifier model. First, let's split out our data into a training and test set. I will hold 20% of the data for testing."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 217,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Use sklearn to split our dataset\n",
    "features = ['Age','EDUC','CDR']\n",
    "x = disease[features]\n",
    "y = disease['Group']\n",
    "X_train, X_test, y_train, y_test = train_test_split(x, y, test_size = 0.2, random_state = 1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Before we train we also need to scale our data for our neural network. This is because neural networks are suspetible to overfitting on features that have a higher magnitude. To scale this data I will use a [MinMaxScaler](https://scikit-learn.org/stable/modules/generated/sklearn.preprocessing.MinMaxScaler.html)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 218,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[0.45714286, 0.66666667, 0.        ],\n",
       "       [0.22857143, 0.33333333, 0.        ],\n",
       "       [0.71428571, 0.46666667, 0.5       ],\n",
       "       [1.        , 0.6       , 1.        ],\n",
       "       [0.62857143, 0.26666667, 0.5       ],\n",
       "       [0.22857143, 0.53333333, 0.        ],\n",
       "       [0.34285714, 0.53333333, 1.        ],\n",
       "       [0.54285714, 0.4       , 0.5       ],\n",
       "       [0.34285714, 0.26666667, 0.5       ],\n",
       "       [0.42857143, 0.46666667, 0.        ],\n",
       "       [0.71428571, 0.6       , 0.        ],\n",
       "       [0.6       , 0.33333333, 0.        ],\n",
       "       [0.54285714, 0.26666667, 0.5       ],\n",
       "       [0.45714286, 0.4       , 0.        ],\n",
       "       [0.25714286, 0.2       , 0.5       ],\n",
       "       [0.05714286, 0.46666667, 0.        ],\n",
       "       [0.77142857, 0.26666667, 0.5       ],\n",
       "       [0.57142857, 0.        , 0.5       ],\n",
       "       [0.88571429, 1.        , 0.        ],\n",
       "       [0.42857143, 0.4       , 1.        ],\n",
       "       [0.82857143, 0.66666667, 0.        ],\n",
       "       [0.22857143, 0.53333333, 1.        ],\n",
       "       [0.22857143, 0.46666667, 0.5       ],\n",
       "       [0.25714286, 0.53333333, 0.5       ],\n",
       "       [0.34285714, 0.6       , 0.        ],\n",
       "       [0.        , 0.53333333, 0.        ],\n",
       "       [0.25714286, 0.2       , 1.        ],\n",
       "       [0.14285714, 0.53333333, 1.        ],\n",
       "       [0.48571429, 0.66666667, 0.        ],\n",
       "       [0.42857143, 0.26666667, 0.5       ],\n",
       "       [0.34285714, 0.4       , 0.        ],\n",
       "       [0.2       , 0.26666667, 0.5       ],\n",
       "       [0.34285714, 0.53333333, 0.5       ],\n",
       "       [0.28571429, 0.53333333, 0.        ],\n",
       "       [0.11428571, 0.33333333, 0.        ],\n",
       "       [0.62857143, 0.46666667, 0.        ],\n",
       "       [0.74285714, 0.26666667, 0.        ],\n",
       "       [0.45714286, 0.53333333, 0.        ],\n",
       "       [0.28571429, 0.26666667, 1.        ],\n",
       "       [0.31428571, 0.53333333, 0.5       ],\n",
       "       [0.34285714, 0.66666667, 0.        ],\n",
       "       [0.34285714, 0.4       , 0.5       ],\n",
       "       [0.28571429, 0.26666667, 0.5       ],\n",
       "       [0.4       , 0.33333333, 0.5       ],\n",
       "       [0.37142857, 0.53333333, 0.        ],\n",
       "       [0.57142857, 0.26666667, 0.        ],\n",
       "       [0.77142857, 0.53333333, 0.        ],\n",
       "       [0.45714286, 0.26666667, 1.        ],\n",
       "       [0.4       , 0.26666667, 0.5       ],\n",
       "       [0.48571429, 0.26666667, 1.        ],\n",
       "       [0.17142857, 0.26666667, 0.        ],\n",
       "       [0.4       , 0.53333333, 1.        ],\n",
       "       [0.77142857, 0.        , 0.5       ],\n",
       "       [0.54285714, 0.26666667, 0.5       ],\n",
       "       [0.11428571, 0.33333333, 0.        ],\n",
       "       [0.8       , 0.33333333, 0.        ],\n",
       "       [0.54285714, 0.26666667, 0.        ],\n",
       "       [0.85714286, 0.66666667, 0.        ],\n",
       "       [0.82857143, 0.26666667, 0.        ],\n",
       "       [0.        , 0.33333333, 0.        ],\n",
       "       [0.8       , 0.26666667, 0.        ],\n",
       "       [0.37142857, 0.66666667, 0.5       ],\n",
       "       [0.05714286, 0.33333333, 0.        ],\n",
       "       [0.77142857, 0.26666667, 0.        ],\n",
       "       [0.62857143, 0.26666667, 0.        ],\n",
       "       [0.85714286, 0.33333333, 0.        ],\n",
       "       [0.25714286, 0.53333333, 0.        ],\n",
       "       [0.48571429, 0.        , 1.        ]])"
      ]
     },
     "execution_count": 218,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Use sklearn to scale your feature vector (if you use a neural network)\n",
    "\n",
    "scaler = MinMaxScaler()\n",
    "scaler.fit_transform(X_train)\n",
    "scaler.fit_transform(X_test)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Alright now that our data is ready, let's create a neural network. For this project I will use a neural network with 3 hidden layers. My neural network is a connected graph using the relu activation functions in the middle and sigmoid function at the last layer. The sigmoid function makes it easy for us to map to 0 or 1, whether or not the patient has Alzheimer's."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 219,
   "metadata": {},
   "outputs": [],
   "source": [
    "# source code from https://machinelearningmastery.com/tutorial-first-neural-network-python-keras/\n",
    "# define the keras model - connected graph using relu and sigmoid activation functions\n",
    "model = Sequential()\n",
    "model.add(Dense(12, input_dim=3, activation='relu')) # input is 3\n",
    "model.add(Dense(8, activation='relu'))\n",
    "model.add(Dense(1, activation='sigmoid')) # last layer being sigmoid allows us to easily map to either 0 or 1\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 220,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/100\n",
      "27/27 [==============================] - 0s 498us/step - loss: 0.8397 - accuracy: 0.4813\n",
      "Epoch 2/100\n",
      "27/27 [==============================] - 0s 467us/step - loss: 0.6965 - accuracy: 0.5410\n",
      "Epoch 3/100\n",
      "27/27 [==============================] - 0s 483us/step - loss: 0.6888 - accuracy: 0.5746\n",
      "Epoch 4/100\n",
      "27/27 [==============================] - 0s 480us/step - loss: 0.6874 - accuracy: 0.5821\n",
      "Epoch 5/100\n",
      "27/27 [==============================] - 0s 468us/step - loss: 0.6837 - accuracy: 0.5746\n",
      "Epoch 6/100\n",
      "27/27 [==============================] - 0s 508us/step - loss: 0.6793 - accuracy: 0.5784\n",
      "Epoch 7/100\n",
      "27/27 [==============================] - 0s 496us/step - loss: 0.6796 - accuracy: 0.5896\n",
      "Epoch 8/100\n",
      "27/27 [==============================] - 0s 467us/step - loss: 0.6780 - accuracy: 0.5821\n",
      "Epoch 9/100\n",
      "27/27 [==============================] - 0s 501us/step - loss: 0.6757 - accuracy: 0.5784\n",
      "Epoch 10/100\n",
      "27/27 [==============================] - 0s 515us/step - loss: 0.6733 - accuracy: 0.5896\n",
      "Epoch 11/100\n",
      "27/27 [==============================] - 0s 430us/step - loss: 0.6740 - accuracy: 0.5709\n",
      "Epoch 12/100\n",
      "27/27 [==============================] - 0s 460us/step - loss: 0.6751 - accuracy: 0.5821\n",
      "Epoch 13/100\n",
      "27/27 [==============================] - 0s 407us/step - loss: 0.6701 - accuracy: 0.5784\n",
      "Epoch 14/100\n",
      "27/27 [==============================] - 0s 401us/step - loss: 0.6696 - accuracy: 0.5933\n",
      "Epoch 15/100\n",
      "27/27 [==============================] - 0s 478us/step - loss: 0.6662 - accuracy: 0.5672\n",
      "Epoch 16/100\n",
      "27/27 [==============================] - 0s 421us/step - loss: 0.6666 - accuracy: 0.5933\n",
      "Epoch 17/100\n",
      "27/27 [==============================] - 0s 511us/step - loss: 0.6622 - accuracy: 0.6007\n",
      "Epoch 18/100\n",
      "27/27 [==============================] - 0s 398us/step - loss: 0.6636 - accuracy: 0.6455\n",
      "Epoch 19/100\n",
      "27/27 [==============================] - 0s 475us/step - loss: 0.6636 - accuracy: 0.5933\n",
      "Epoch 20/100\n",
      "27/27 [==============================] - 0s 441us/step - loss: 0.6570 - accuracy: 0.5896\n",
      "Epoch 21/100\n",
      "27/27 [==============================] - 0s 373us/step - loss: 0.6540 - accuracy: 0.6007\n",
      "Epoch 22/100\n",
      "27/27 [==============================] - 0s 406us/step - loss: 0.6526 - accuracy: 0.6231\n",
      "Epoch 23/100\n",
      "27/27 [==============================] - 0s 487us/step - loss: 0.6515 - accuracy: 0.6157\n",
      "Epoch 24/100\n",
      "27/27 [==============================] - 0s 498us/step - loss: 0.6475 - accuracy: 0.6007\n",
      "Epoch 25/100\n",
      "27/27 [==============================] - 0s 407us/step - loss: 0.6454 - accuracy: 0.6343\n",
      "Epoch 26/100\n",
      "27/27 [==============================] - 0s 494us/step - loss: 0.6410 - accuracy: 0.6231\n",
      "Epoch 27/100\n",
      "27/27 [==============================] - 0s 472us/step - loss: 0.6379 - accuracy: 0.6493\n",
      "Epoch 28/100\n",
      "27/27 [==============================] - 0s 472us/step - loss: 0.6335 - accuracy: 0.6604\n",
      "Epoch 29/100\n",
      "27/27 [==============================] - 0s 357us/step - loss: 0.6303 - accuracy: 0.6530\n",
      "Epoch 30/100\n",
      "27/27 [==============================] - 0s 431us/step - loss: 0.6307 - accuracy: 0.6530\n",
      "Epoch 31/100\n",
      "27/27 [==============================] - 0s 385us/step - loss: 0.6236 - accuracy: 0.6866\n",
      "Epoch 32/100\n",
      "27/27 [==============================] - 0s 457us/step - loss: 0.6240 - accuracy: 0.6343\n",
      "Epoch 33/100\n",
      "27/27 [==============================] - 0s 440us/step - loss: 0.6173 - accuracy: 0.6903\n",
      "Epoch 34/100\n",
      "27/27 [==============================] - 0s 475us/step - loss: 0.6178 - accuracy: 0.6716\n",
      "Epoch 35/100\n",
      "27/27 [==============================] - 0s 492us/step - loss: 0.6077 - accuracy: 0.6828\n",
      "Epoch 36/100\n",
      "27/27 [==============================] - 0s 497us/step - loss: 0.6041 - accuracy: 0.7090\n",
      "Epoch 37/100\n",
      "27/27 [==============================] - 0s 431us/step - loss: 0.5975 - accuracy: 0.6903\n",
      "Epoch 38/100\n",
      "27/27 [==============================] - 0s 451us/step - loss: 0.5896 - accuracy: 0.7127\n",
      "Epoch 39/100\n",
      "27/27 [==============================] - 0s 430us/step - loss: 0.5931 - accuracy: 0.6978\n",
      "Epoch 40/100\n",
      "27/27 [==============================] - 0s 532us/step - loss: 0.5839 - accuracy: 0.7313\n",
      "Epoch 41/100\n",
      "27/27 [==============================] - 0s 492us/step - loss: 0.5790 - accuracy: 0.7201\n",
      "Epoch 42/100\n",
      "27/27 [==============================] - 0s 438us/step - loss: 0.5796 - accuracy: 0.7052\n",
      "Epoch 43/100\n",
      "27/27 [==============================] - 0s 437us/step - loss: 0.5691 - accuracy: 0.7500\n",
      "Epoch 44/100\n",
      "27/27 [==============================] - 0s 483us/step - loss: 0.5608 - accuracy: 0.7276\n",
      "Epoch 45/100\n",
      "27/27 [==============================] - 0s 391us/step - loss: 0.5528 - accuracy: 0.7500\n",
      "Epoch 46/100\n",
      "27/27 [==============================] - 0s 360us/step - loss: 0.5508 - accuracy: 0.7164\n",
      "Epoch 47/100\n",
      "27/27 [==============================] - 0s 376us/step - loss: 0.5478 - accuracy: 0.7388\n",
      "Epoch 48/100\n",
      "27/27 [==============================] - 0s 416us/step - loss: 0.5379 - accuracy: 0.7388\n",
      "Epoch 49/100\n",
      "27/27 [==============================] - 0s 554us/step - loss: 0.5236 - accuracy: 0.7873\n",
      "Epoch 50/100\n",
      "27/27 [==============================] - 0s 464us/step - loss: 0.5253 - accuracy: 0.7873\n",
      "Epoch 51/100\n",
      "27/27 [==============================] - 0s 504us/step - loss: 0.5135 - accuracy: 0.7910\n",
      "Epoch 52/100\n",
      "27/27 [==============================] - 0s 429us/step - loss: 0.5035 - accuracy: 0.8134\n",
      "Epoch 53/100\n",
      "27/27 [==============================] - 0s 413us/step - loss: 0.4961 - accuracy: 0.8022\n",
      "Epoch 54/100\n",
      "27/27 [==============================] - 0s 440us/step - loss: 0.4927 - accuracy: 0.7873\n",
      "Epoch 55/100\n",
      "27/27 [==============================] - 0s 439us/step - loss: 0.4766 - accuracy: 0.8470\n",
      "Epoch 56/100\n",
      "27/27 [==============================] - 0s 466us/step - loss: 0.4752 - accuracy: 0.8172\n",
      "Epoch 57/100\n",
      "27/27 [==============================] - 0s 420us/step - loss: 0.4629 - accuracy: 0.8284\n",
      "Epoch 58/100\n",
      "27/27 [==============================] - 0s 434us/step - loss: 0.4685 - accuracy: 0.7948\n",
      "Epoch 59/100\n",
      "27/27 [==============================] - 0s 418us/step - loss: 0.4634 - accuracy: 0.7910\n",
      "Epoch 60/100\n",
      "27/27 [==============================] - 0s 377us/step - loss: 0.4579 - accuracy: 0.8097\n",
      "Epoch 61/100\n",
      "27/27 [==============================] - 0s 369us/step - loss: 0.4627 - accuracy: 0.7985\n",
      "Epoch 62/100\n",
      "27/27 [==============================] - 0s 444us/step - loss: 0.4224 - accuracy: 0.8470\n",
      "Epoch 63/100\n",
      "27/27 [==============================] - 0s 448us/step - loss: 0.4207 - accuracy: 0.8619\n",
      "Epoch 64/100\n",
      "27/27 [==============================] - 0s 440us/step - loss: 0.4033 - accuracy: 0.8955\n",
      "Epoch 65/100\n",
      "27/27 [==============================] - 0s 410us/step - loss: 0.3953 - accuracy: 0.8806\n",
      "Epoch 66/100\n",
      "27/27 [==============================] - 0s 441us/step - loss: 0.3906 - accuracy: 0.8731\n",
      "Epoch 67/100\n",
      "27/27 [==============================] - 0s 475us/step - loss: 0.3732 - accuracy: 0.9216\n",
      "Epoch 68/100\n",
      "27/27 [==============================] - 0s 480us/step - loss: 0.3740 - accuracy: 0.8843\n",
      "Epoch 69/100\n",
      "27/27 [==============================] - 0s 352us/step - loss: 0.3590 - accuracy: 0.9030\n",
      "Epoch 70/100\n",
      "27/27 [==============================] - 0s 394us/step - loss: 0.3584 - accuracy: 0.9067\n",
      "Epoch 71/100\n",
      "27/27 [==============================] - 0s 477us/step - loss: 0.3476 - accuracy: 0.9291\n",
      "Epoch 72/100\n",
      "27/27 [==============================] - 0s 425us/step - loss: 0.3396 - accuracy: 0.9254\n",
      "Epoch 73/100\n",
      "27/27 [==============================] - 0s 431us/step - loss: 0.3276 - accuracy: 0.9216\n",
      "Epoch 74/100\n",
      "27/27 [==============================] - 0s 417us/step - loss: 0.3151 - accuracy: 0.9403\n",
      "Epoch 75/100\n",
      "27/27 [==============================] - 0s 448us/step - loss: 0.3032 - accuracy: 0.9366\n",
      "Epoch 76/100\n",
      "27/27 [==============================] - 0s 487us/step - loss: 0.3003 - accuracy: 0.9366\n",
      "Epoch 77/100\n",
      "27/27 [==============================] - 0s 420us/step - loss: 0.2916 - accuracy: 0.9366\n",
      "Epoch 78/100\n",
      "27/27 [==============================] - 0s 457us/step - loss: 0.2793 - accuracy: 0.9552\n",
      "Epoch 79/100\n",
      "27/27 [==============================] - 0s 332us/step - loss: 0.2752 - accuracy: 0.9627\n",
      "Epoch 80/100\n",
      "27/27 [==============================] - 0s 465us/step - loss: 0.2675 - accuracy: 0.9440\n",
      "Epoch 81/100\n",
      "27/27 [==============================] - 0s 436us/step - loss: 0.2633 - accuracy: 0.9552\n",
      "Epoch 82/100\n",
      "27/27 [==============================] - 0s 407us/step - loss: 0.2750 - accuracy: 0.9254\n",
      "Epoch 83/100\n",
      "27/27 [==============================] - 0s 473us/step - loss: 0.2486 - accuracy: 0.9627\n",
      "Epoch 84/100\n",
      "27/27 [==============================] - 0s 423us/step - loss: 0.2339 - accuracy: 0.9739\n",
      "Epoch 85/100\n",
      "27/27 [==============================] - 0s 385us/step - loss: 0.2265 - accuracy: 0.9664\n",
      "Epoch 86/100\n",
      "27/27 [==============================] - 0s 418us/step - loss: 0.2203 - accuracy: 0.9813\n",
      "Epoch 87/100\n",
      "27/27 [==============================] - 0s 483us/step - loss: 0.2274 - accuracy: 0.9627\n",
      "Epoch 88/100\n",
      "27/27 [==============================] - 0s 570us/step - loss: 0.2091 - accuracy: 0.9813\n",
      "Epoch 89/100\n",
      "27/27 [==============================] - 0s 528us/step - loss: 0.2088 - accuracy: 0.9813\n",
      "Epoch 90/100\n",
      "27/27 [==============================] - 0s 487us/step - loss: 0.1929 - accuracy: 0.9851\n",
      "Epoch 91/100\n",
      "27/27 [==============================] - 0s 450us/step - loss: 0.1935 - accuracy: 0.9888\n",
      "Epoch 92/100\n",
      "27/27 [==============================] - 0s 469us/step - loss: 0.1882 - accuracy: 0.9776\n",
      "Epoch 93/100\n",
      "27/27 [==============================] - 0s 424us/step - loss: 0.1757 - accuracy: 0.9888\n",
      "Epoch 94/100\n",
      "27/27 [==============================] - 0s 450us/step - loss: 0.1695 - accuracy: 0.9925\n",
      "Epoch 95/100\n",
      "27/27 [==============================] - 0s 372us/step - loss: 0.1649 - accuracy: 0.9925\n",
      "Epoch 96/100\n",
      "27/27 [==============================] - 0s 412us/step - loss: 0.1604 - accuracy: 0.9925\n",
      "Epoch 97/100\n",
      "27/27 [==============================] - 0s 479us/step - loss: 0.1547 - accuracy: 0.9925\n",
      "Epoch 98/100\n",
      "27/27 [==============================] - 0s 496us/step - loss: 0.1511 - accuracy: 0.9925\n",
      "Epoch 99/100\n",
      "27/27 [==============================] - 0s 428us/step - loss: 0.1484 - accuracy: 0.9851\n",
      "Epoch 100/100\n",
      "27/27 [==============================] - 0s 475us/step - loss: 0.1419 - accuracy: 0.9925\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<tensorflow.python.keras.callbacks.History at 0x7f425aef7b00>"
      ]
     },
     "execution_count": 220,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# compile the keras model\n",
    "model.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])\n",
    "# fit the keras model on the dataset\n",
    "model.fit(X_train, y_train, epochs=100, batch_size=10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 221,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "9/9 [==============================] - 0s 598us/step - loss: 0.1445 - accuracy: 0.9925\n",
      "Accuracy: 99.25\n"
     ]
    }
   ],
   "source": [
    "_, accuracy = model.evaluate(X_train, y_train)\n",
    "print('Accuracy: %.2f' % (accuracy*100))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "NICE! let's print the weights."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 222,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[array([[-0.26616108, -0.3924657 ,  0.2494779 ,  0.13159908, -0.36632594,\n",
       "         -0.5015903 ,  0.48492858, -0.2979524 ,  0.42693114, -0.568185  ,\n",
       "          0.22650643,  0.12289394],\n",
       "        [-0.37975538,  0.01612872, -0.31725222,  0.1862542 ,  0.58237976,\n",
       "          0.34082073, -0.46144047,  0.2969423 ,  0.12773553,  0.51122147,\n",
       "          0.23595025, -0.4430783 ],\n",
       "        [-0.29332304,  0.4803794 ,  2.1093862 ,  2.0021405 ,  0.2692179 ,\n",
       "         -0.59819955, -2.5583823 , -0.07597119,  2.6149063 ,  0.41492015,\n",
       "         -2.1352055 ,  2.6183856 ]], dtype=float32),\n",
       " array([ 0.        ,  0.        ,  0.0622514 ,  0.04212363,  0.        ,\n",
       "         0.        , -0.10512767,  0.        ,  0.11132807,  0.        ,\n",
       "        -0.08387619,  0.07043348], dtype=float32),\n",
       " array([[-0.15996048, -0.08219397,  0.11079621,  0.24847001, -0.12174439,\n",
       "         -0.03816634,  0.08047372, -0.2981299 ],\n",
       "        [-0.07785887,  0.15405285,  0.49240708,  0.09907353,  0.3338827 ,\n",
       "          0.18477142, -0.12123328,  0.18585962],\n",
       "        [ 0.42160982, -0.27218255,  0.37258542,  0.5647708 ,  0.4131711 ,\n",
       "          0.5353298 , -0.3669768 ,  0.08465278],\n",
       "        [-0.15418844,  0.5086698 ,  0.04501871,  0.12002797, -0.42533818,\n",
       "         -0.48454186, -0.4774015 ,  0.05979878],\n",
       "        [-0.02547026, -0.0651812 , -0.4497922 ,  0.5037805 , -0.44294983,\n",
       "          0.28425342, -0.16303241,  0.06719559],\n",
       "        [ 0.28259915,  0.04947013, -0.533764  ,  0.059232  ,  0.3648426 ,\n",
       "         -0.38082945,  0.212569  ,  0.30371457],\n",
       "        [-0.0030241 ,  0.17818041,  0.05119044, -0.52986056,  0.04233633,\n",
       "         -0.17360464, -0.13649553, -0.15635341],\n",
       "        [-0.2001077 , -0.11439872,  0.0577504 , -0.26885614,  0.4320634 ,\n",
       "          0.32788223, -0.27622336,  0.42084193],\n",
       "        [ 0.37024534, -0.45643696, -0.5049935 ,  0.34954348, -0.35314068,\n",
       "         -0.49092758, -0.15236148, -0.3088007 ],\n",
       "        [ 0.22153717, -0.09044054,  0.14428586,  0.35576403, -0.4729137 ,\n",
       "          0.39529073,  0.3367347 , -0.41015416],\n",
       "        [ 0.39230868,  0.10438598,  0.2434137 , -0.32161307,  0.3334215 ,\n",
       "         -0.0996848 , -0.05971479, -0.54538035],\n",
       "        [-0.20148358, -0.3128886 ,  0.2561587 ,  0.36195305,  0.29317898,\n",
       "          0.31020397,  0.21207094, -0.15786248]], dtype=float32),\n",
       " array([ 0.02539652, -0.01677674, -0.00952166,  0.08789153, -0.03383492,\n",
       "         0.        ,  0.        ,  0.        ], dtype=float32),\n",
       " array([[-0.12528057],\n",
       "        [ 0.6576957 ],\n",
       "        [-0.29948986],\n",
       "        [ 1.352713  ],\n",
       "        [-0.40327623],\n",
       "        [ 0.00166208],\n",
       "        [-0.79249895],\n",
       "        [-0.25973785]], dtype=float32),\n",
       " array([-0.02876866], dtype=float32)]"
      ]
     },
     "execution_count": 222,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.get_weights()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "And try our model on the test set."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 224,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "100.0\n"
     ]
    }
   ],
   "source": [
    "# Error in testing data\n",
    "predictions = model.predict(X_test)\n",
    "# round predictions \n",
    "rounded = [round(x[0]) for x in predictions]\n",
    "correct=0\n",
    "\n",
    "for x in range(len(rounded)):\n",
    "    if(rounded[x]==y_test.iloc[x]):\n",
    "        correct+=1\n",
    "error = 100*correct/len(rounded)\n",
    "print(error)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Great, this means that we have an accuracy of 99.25% for our Alzheimer's neural network model on our training set and 100% on our test set. I am a bit surprised at how well it worked given our lack of feature variables. I would expect that we would need more to verify whether the patient has Alzheimer's, but I could be wrong and those three worked just as well."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Let's try another technique tho, SVM Classifier."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 226,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1.0\n"
     ]
    }
   ],
   "source": [
    "# Initialize a SVM Classifier with your choice of parameters\n",
    "\n",
    "clf = svm.SVC(kernel=\"linear\")\n",
    "clf.fit(X_train, y_train)\n",
    "y_pred = clf.predict(X_test)\n",
    "clf_accuracy = accuracy_score(y_test, y_pred)\n",
    "print(clf_accuracy)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 227,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(33.0, 0.5, 'Actual')"
      ]
     },
     "execution_count": 227,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAW4AAAGDCAYAAAALePrTAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDMuMC4zLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvnQurowAAH99JREFUeJzt3XmYXFWd//H3t9OBBBCiQIKQCMq+jAYFnUGUEBEDAqI4IjrKplEcZ1wYcGF+Dii4j8u4oFE2hUFRxAVJABkjBJE9LAFkUZaIIYiDSiCQ5fv7496ORU8vle6uqj7t+5XnPk/VvbfOPV3p/vSp77n3dmQmkqRydHW6A5KktWNwS1JhDG5JKozBLUmFMbglqTAGtyQVxuAeJSJiYkT8JCL+FBHfG0Y7b46IS0ayb50QEXMj4vAhvvbkiPhDRCwZ6X71c7z5EfG2QfbZKiIyIrqH0P5zIuKxiBg39F5qLDG411JEvCkirqt/kH5fB8yeI9D064EpwMaZ+Y9DbSQzz8nMfUegP08TETPq4PlBr/UvqNfPb7KdEyPi7MH2y8z9MvOsIfRzGnAssFNmbra2rx+g3YiI30TEbSPVZrMy8/7M3CAzV7XjeBFxZkQc0Y5jaWgM7rUQEe8HvgB8nCpknwN8FXjNCDS/JXBnZq4cgbZa5WFgj4jYuGHd4cCdI3WAOiCH8325JfBIZi4dwrEHGg2/HJgMPC8idh9q50Y7R/VlMLibFBEbAR8F/jkzf5CZyzJzRWb+JDOPq/dZNyK+EBEP1ssXImLdetuMiFgcEcdGxNJ6tH5kve0k4CPAofVI/ujeI9PeH7Uj4oh6BPiXiPhtRLy5Yf2ChtftERHX1iWYayNij4Zt8yPiYxFxZd3OJRGxyQBvw1PAD4E31q8fB7wBOKfXe/XFiHggIv4cEddHxMvq9bOADzd8nTc19OOUiLgSeJwqHNeUHyLi1Ij4fkP7n4qIyyIieh13H+BSYPO6/TPr9QdFxKKIeLRud8eG19wbER+IiJuBZQOE9+HAj4CL6sd9ioib6mP3LBkRMxp2eXNE3F+Xck5oeF1XRHwwIu6JiEci4ryIeFa9rff//fyoykG/rI/xk4jYOCLOqd/zayNiq4a2d4iISyPijxHx64h4Q8O2M+v396KIWAbs3evr2SYiflF///whIr7b39euNspMlyYWYBawEugeYJ+PAr+iGpltCvwS+Fi9bUb9+o8C44H9qULqmfX2E4GzG9rq/XwrIIFuYH3gz8D29bZnAzvXj48AFtSPnwX8L/CW+nWH1c83rrfPB+4BtgMm1s8/2c/XNgNYDOwBXF2v2x+4GHgbML9h338CNq6PeSywBJjQ19fV0I/7gZ3r14yv172t3r4e1aj+COBlwB+AqQP1s+H5dsAy4JV1u8cDdwPr1NvvBRYC04CJ/bS5Xv1+7w8cUh9/nV79f1sfr5sN3AFs2PD/9436vX4B8CSwY73ve6m+d6YC6wJfB87t/X/fcLy7ga2BjYDb6vdnn/r9+xZwRr3v+sADwJH1thfW/e/5fjkT+BPwUqqB3IReX8O5wAk924A9O/2z6JKOuNfCxsAfcuBSxpuBj2bm0sx8GDiJKjR7rKi3r8jMi4DHgO2H2J/VwC4RMTEzf5+Zi/rY59XAXZn57cxcmZnnUgXJgQ37nJGZd2bmE8B5wPSBDpqZvwSeFRHbA2+lCone+5ydmY/Ux/xPqiAa7Os8MzMX1a9Z0au9x6l+GXwOOBv4l8xcPEh7PQ4FfpqZl9btfpYqOPdo2Oe/MvOB+j3oy+uoQvYS4EKqAHz1QAeNat7jZOCgzPxzw6aTMvOJzLwJuIkqwAHeAZyQmYsz80mqX3CvH+ATwBmZeU9m/gmYC9yTmT+rvz+/B+xa73cAcG9mnlG/tzcA51PNqfT4UWZemZmrM3N5r+OsoCo/bZ6ZyzNzAeo4g7t5jwCbDFIH3Ry4r+H5ffW6NW30Cv7HgQ3WtiOZuYwqkN4J/D4ifhoROzTRn54+bdHwvPHMi2b7823g3VQfqy/ovbEuB91ef7x+lGpUOFAJBqpRYb8y8xrgN0BQ/YJp1tPeg8xcXR+r8T0Y8NhUpZHz6uB7EvgBA5dLptV9PDwze9f/+3u/twQuqMs5jwK3A6uo5lL68lDD4yf6eN7Y7kt62q3bfjPQOHE70Nd/PNV7fk1dbjpqgH3VJgZ3864ClgMHD7DPg1Q/KD2eU68bimVUH9F7PO0Micy8ODNfSVUmuYPqI/hg/enp0++G2Kce3wbeBVxUj4bXqOvZH6CqfT8zMydRfRTvqUf3dzvKAW9TGRH/TDVyf5AqTJr1tPegrotP4+nvQb/HjoipwEzgnyJiSVSnGL4e2L+v+YCImEg1D/CFzJy7Fv18ANgvMyc1LBMyc7j/Vw8Av+jV7gaZeUzDPv1+/Zm5JDPfnpmbU30q+GpEbDPMPmmYDO4m1R9JPwJ8JSIOjoj1ImJ8ROwXEZ+udzsX+PeI2LT+of4I1Uf7oVgIvDyqc3g3Aj7UsyEiptQTbutTfYR/jGp01ttFwHZRncLYHRGHAjtRfdwfssz8LbAXVe2zt2dQ1fIfBroj4iNUNd4eDwFbxVqcORIR21GVHf6JqvR0fEQMWNJpcB7w6oh4RUSMp6q5P0k1/9CMt1DVj7enKiNNp6qbL6aaM+jtdOCOzPx0H9sG8jXglIjYEqD+HhqJs5UupPoeeEv9/To+InZvnKAdSET8Y/3LC6r5kaTv7zW1kcG9FjLzc8D7gX+nCqYHqEoGP6x3ORm4DrgZuAW4oV43lGNdCny3but6nh62XVQB9CDwR6oQfVcfbTxCVeM8lqrUczxwQGb+YSh96tX2gszs69PExVQ11zupShTLefpH8Z6Lix6JiBsGO05dmjob+FRm3pSZd1GdmfLtqM/YGaSfv6YK/C9RTcodCByYmU8N9tra4cBX65HnmoUqaPsql7wReG2vM0te1sRxvgj8GLgkIv5CNVH5kib72K/M/Auwb92vB6lKNZ+i+vTSjN2BqyPisbp/76l/cauDItM/pCBJJXHELUmFMbglqQ0iYkJEXFNfpLUoqgvveq4WPiUi7qzPxvrXwdpa6xveSJKG5ElgZmY+Vk+UL4iIucCOVGc67ZCZqyNi8mANGdyS1AZZTSg+Vj8dXy8JHAO8qb7GgGziPjuWSiSpTSJiXEQsBJYCl2bm1VS3Ljg0qruOzo2IbQdrZ9SOuOOVUz3dRf/HE/NG7EaEGkMmjFsvBt9rYMPOnJ/97h1U96fpMScz5zTuktWteadHxCSqK2V3oTo1c3lm7hYRr6O6FmDAU0hHbXBLUlvF8LK/Duk5g+5Y7ftoVPewn0V1Mdf59aYLgDMGe72lEklqg/pq2En144lUd3O8g+oCvpn1bnvRxP3tHXFLErRjGPts4Kyo7mPfRXXjsgujun/+ORHxPqrJywH/DB4Y3JJUGWapZDCZeTN/vd1u4/pHGeQ2wb0Z3JIEf71/ZQGscUtSYRxxSxK0vFQykgxuSYKi6g8GtySBI25JKk45uV3ShwNJEjjilqRKVzlDboNbkqCoUonBLUng5KQkFaec3HZyUpJK44hbksDJSUkqTjm5bXBLElDU5KQ1bkkqjCNuSQJr3JJUnHJy2+CWJKCoGrfBLUlQ1IjbyUlJKowjbkkCJyclqTjl5LbBLUmAk5OSVJyCZvwK6qokCRxxS1LFUokkFaac3Da4JQkoasRtjVuSCuOIW5KgqGGswS1JUFSpxOCWJHByUpKKU9C9Sgqq6kiSwBG3JFWscUtSYcrJbYNbkgDCEbcklaWk4HZyUpIK44hbkihqbtLgliSAroKS2+CWJKxxS5JayBG3JFHWiNvgliTKCm5LJZJEdVbJcJbB248JEXFNRNwUEYsi4qR6/XMj4uqIuCsivhsR6wzWlsEtSVQj7uEsTXgSmJmZLwCmA7Mi4u+BTwGfz8xtgf8Fjh6sIYNbktogK4/VT8fXSwIzge/X688CDh6sLYNbkhj+iDsiZkfEdQ3L7D6OMS4iFgJLgUuBe4BHM3NlvctiYIvB+urkpCQBMczbA2bmHGDOIPusAqZHxCTgAmDHvnYb7FgGtyTR3rNKMvPRiJgP/D0wKSK661H3VODBwV5vqUSSaMtZJZvWI20iYiKwD3A78HPg9fVuhwM/GqwtR9yS1B7PBs6KiHFUg+bzMvPCiLgN+E5EnAzcCJw2WEMGtyTR+ptMZebNwK59rP8N8OK1acvgliTKunLS4JYkygpuJyclqTCOuCUJ/wKOJBWnpFKJwS1JGNySVJySgtvJSUkqjCNuSaKsEbfBLUl4VokkFccRtyQVpqTgdnJSkgrjiFuSaP3dAUeSwS1JODkpScWxxi1JahmDe5RZd/y6XP2lC1n4tUu49RuXceJbj12z7eQjj+fXZ1zObaf9nH85+KgO9lKdduUVV3LQ/gdzwKsO4rRvnN7p7owJMcx/7WSpZJR5csWTzDzuDSxb/jjd47pZ8PkLmHvtz9nxOdswbdPN2eGovchMNp20cae7qg5ZtWoVHz/5k3z9m6cyZcoU3nTom5mx915svc3Wne5a0UoqlbQsuCNiB+A1wBZAUv3J+R9n5u2tOuZYsWz54wCM7+5mfHc3mckxB7yVN33i3WQmAA8/+kgnu6gOuvWWW5n2nGlMnTYVgFn7vYr5/zPf4B6mkoK7JaWSiPgA8B0ggGuAa+vH50bEB1txzLGkq6uLG792MUu/dxOX3nAF19xxI1tvviWHzjiQa7/yUy465dtss8VzO91NdcjSh5ay2WZT1jyfvNkUHlr6cAd7NDZEDG9pp1aNuI8Gds7MFY0rI+JzwCLgk329KCJmA7MB2GESTF2/Rd0b3VavXs2u73wVG62/IRec+E123mp71h2/DsufepLd//nVvHbP/Tj92M/y8vcf0umuqgPqD11PU85YUSOhVZOTq4HN+1j/7HpbnzJzTmbulpm7/a2GdqM/Lfsz82+6ilm7zWDxw7/n/CsuAuCCBXN5/vN27HDv1ClTNpvMkiUPrXm+dMlDTJ68aQd7NDZExLCWdmpVcL8XuCwi5kbEnHqZB1wGvKdFxxwTNtnoWWy0/oYATFhnAvu8cE/ueOBufvjLi5k5/aUA7PX8f+DOxb/pZDfVQTvvsjP333c/ixf/jhVPrWDe3IvZa+8Zne5W8UoK7paUSjJzXkRsB7yYanIygMXAtZm5qhXHHCue/awpnHX85xnXNY6uCM67/EJ+evVlLLj1Ws750Jd43yFv57EnlvG2zx3X6a6qQ7q7u/nQCR/gmLe/i9WrV3Pwa1/DNts6MTlcJU1ORvZVMBsF4pVTR2fH1FFPzLuz013QKDRh3HrDTt3tPjdrWJlz5/vntS35PY9bkvBeJZJUnJJKJQa3JGFwS1JxSgpubzIlSYVxxC1JODkpScUpqVRicEsSBrckFaek4HZyUpIK44hbknByUpKKU1KpxOCWJChqyG2NW5IK44hbkrBUIknFKSi3DW5JAkfcklSckoLbyUlJKozBLUm0/q+8R8S0iPh5RNweEYsi4j29tv9bRGREbDJYW5ZKJIm2TE6uBI7NzBsi4hnA9RFxaWbeFhHTgFcC9zfTkCNuSaL1I+7M/H1m3lA//gtwO7BFvfnzwPFAU39p3hG3JDH8ycmImA3Mblg1JzPn9LPvVsCuwNURcRDwu8y8qdk+GNySNALqkO4zqBtFxAbA+cB7qconJwD7rs2xDG5Joj2nA0bEeKrQPiczfxARfwc8F+gZbU8FboiIF2fmkv7aMbglidYHd1QHOA24PTM/B5CZtwCTG/a5F9gtM/8wUFtOTkoS1Vklw1ma8FLgLcDMiFhYL/sPpa+OuCWpDTJzATBgxGfmVs20ZXBLEmVd8m5wSxIGtyQVx+CWpMIUlNueVSJJpXHELUlYKpGk8hjcklQWR9ySVJiucnLbyUlJKo0jbknCUokkFafL4JakspQ04rbGLUmFccQtSZQ1ijW4JQlr3JJUnJJq3Aa3JFHWiLukso4kCUfckgRYKpGk4pRUfjC4JYmyatwGtyRRVqmkpE8HkiQccUsSYKlEkopTTmwb3JIElDXitsYtSYVxxC1JlDXiNrglibJOBzS4JQlH3JJUnHJi28lJSSqOI25JwlKJJBXH4JakwoyJs0oi4idA9rc9Mw9qSY8kqQPGyoj7s23rhSSpaf0Gd2b+op0dkaROKme83USNOyK2BT4B7ARM6Fmfmc9rYb8kqa3GSqmkxxnAfwCfB/YGjqSsX06SNKiSgruZC3AmZuZlQGTmfZl5IjCztd2SJPWnmRH38ojoAu6KiHcDvwMmt7ZbktReJZ0O2MyI+73AesC/Ai8C3gIc3spOSVK7dQ1zaadBR9yZeW398DGq+rYkjTkljbibOavk5/RxIU5mWueWNGa0enIyIk4HDgCWZuYu9brpwNeozthbCbwrM68ZrK1matz/1vB4AnBIfQBJUvPOBL4MfKth3aeBkzJzbkTsXz+fMVhDzZRKru+16sqI8OIcSWNKq0fcmXl5RGzVezWwYf14I+DBZtpqplTyrIanXVQTlJs10/hwPDHvzlYfQgWa+O7dO90FjUJ56qJht9GhGvd7gYsj4rNU+bpHMy9qplRyPdVvhaAqkfwWOHqInZSkUalrmNcVRsRsYHbDqjmZOWeQlx0DvC8zz4+INwCnAfsMdqxmgnvHzFzeq4PrNvE6SSrGcEfcdUgPFtS9HQ68p378PeCbzbyomdMPf9nHuqua7JQkqX8PAnvVj2cCdzXzooHux70ZsAUwMSJ25a/3J9mQ6oIcSRoz2nA64LlUZ4xsEhGLqe4B9XbgixHRDSzn6aWWfg1UKnkVcAQwFfhP/hrcfwY+PJSOS9JoFS2+d15mHtbPphetbVsD3Y/7LOCsiDgkM89f24YlqSQlXTnZTI37RRExqedJRDwzIk5uYZ8kSQNoJrj3y8xHe55k5v8C+7euS5LUfl0Rw1raqZnTAcdFxLqZ+SRAREwEPB1Q0pgSbb/H39A1E9xnA5dFxBn18yOBs1rXJUlqv5L+Ak4z9yr5dETcTHU1TwDzgC1b3TFJaqexNjkJsARYTXVnwFcAt7esR5KkAQ10Ac52wBuBw4BHgO9S/d3JvdvUN0lqm1afxz2SBiqV3AFcARyYmXcDRMT72tIrSWqzkmrcA5VKDqEqkfw8Ir4REa+Agn4lSdJaiIhhLe3Ub3Bn5gWZeSiwAzAfeB8wJSJOjYh929Q/SVIvg05OZuayzDwnMw+gum/JQuCDLe+ZJLVR1zD/tVMz53GvkZl/BL5eL5I0ZpR0OuBaBbckjVUGtyQVZrh/uqydyrk4X5IEOOKWJMBSiSQVp6QLcAxuSWLsXPIuSX8zuqKcKb9yeipJAhxxSxLg5KQkFccatyQVpqSzSqxxS1JhHHFLEpZKJKk4JZVKDG5JAqKg87gNbkmirFJJOb9iJEmAI25JAqxxS1JxvHJSkgpT0l/AMbglibJG3E5OSlJhHHFLEp7HLUnFscYtSYWxxi1JahlH3JJEWZe8G9ySRFmlEoNbknByUpKKU9LpgOX0VJIEOOKWJMDJSUkqTkmTk5ZKJIlqxD2cf4O2H3F6RCyNiFsb1n0mIu6IiJsj4oKImNRMXw1uSaIacQ9nacKZwKxe6y4FdsnM5wN3Ah9qpiGDW5LaIDMvB/7Ya90lmbmyfvorYGozbRnckkR1HvdwloiYHRHXNSyz17ILRwFzm9nRyUlJYviTk5k5B5gzxGOfAKwEzmlmf4NbkoDoUAEiIg4HDgBekZnZzGsMbknqkIiYBXwA2CszH2/2dQa3JNH687gj4lxgBrBJRCwG/oPqLJJ1gUvr4/8qM985WFsGtyTR+isnM/OwPlafNpS2DG5JAroKunLS4JYkyrpXiedxS1JhHHFLEmXdZMrgliQ6dx73UBjckoQjbkkqTkl/c7KczwaSJMARtyQBlkokqTglncdtcEsSZY24rXFLUmEccUsSnsctScXxJlOSVBgnJyWpME5OSpJaxhH3KHflFVfyqU98htWrVvPa1x/M0W8/qtNdUges270Olx/7LdbtXofurnF8/8ZLOPHCr3D2kZ9ity13ZsWqlVxz7y2845yTWLl6Zae7W6SSSiWOuEexVatW8fGTP8lXv/5lLvjJ+cy7aB733H1Pp7ulDnhy5VPM/MJRTD/ldUw/5RBm7bQnL3nu8znnmgvZ4cQD+LuPHczE8RN4256HdLqrxYqIYS3tZHCPYrfecivTnjONqdOmMn6d8cza71XM/5/5ne6WOmTZk9UfAR8/rpvx47rJTOYuumLN9mvuvYWpk6Z0qnvF6xrmv/b2tc0i4sh2H7NUSx9aymab/fUHcfJmU3ho6cMd7JE6qSu6uPHD57P001dw6e1Xcc29t6zZ1t3VzVteciDzblvQwR6WzRH3wE7qb0NEzI6I6yLiutO+cXo7+zQqZf7fdeVU4TTSVudqdv34IUz98ExevNXfsfPm26zZ9tXD/h+X3309C+6+oYM9VLu0ZHIyIm7ubxPQ72e5zJwDzAFYvurxPmLrb8uUzSazZMlDa54vXfIQkydv2sEeaTT40xN/Yf5d1zBrpz1Z9ODdfOTVx7DpBs/kHXNO7HTXiubkZBXObwUO7GN5pEXHHHN23mVn7r/vfhYv/h0rnlrBvLkXs9feMzrdLXXAJhs8k40mPgOACePXZZ8d/oE7lvyWo196CK/a8aUcdvpxZF8f0dS0kkolrTod8EJgg8xc2HtDRMxv0THHnO7ubj50wgc45u3vYvXq1Rz82tewzbZbd7pb6oBnb7QpZx3+ccZFF11dXZx3/cX89NZfsOLLN3HfHx/kquP+G4AfLPwZH7vo1A73tkwljbhjtP6WtlSivkx89+6d7oJGoTx10bBT99qHFwwrc3bfdM+2Jb8X4EgSZY24DW5JAijoXiUGtyThiFuSiuPdASVJLeOIW5KwVCJJxTG4JakwJdW4DW5JoqwRt5OTklQYR9ySRFkjboNbkrDGLUnFKWnEbY1bkgrjiFuSsFQiScUpqVRicEsSBrckFaekUomTk5JUGINbkqhKJcP519QxIiZFxPcj4o6IuD0i/mEofbVUIkm0rcb9RWBeZr4+ItYB1htKIwa3JNH6GndEbAi8HDgCIDOfAp4aSluWSiRpBETE7Ii4rmGZ3WuX5wEPA2dExI0R8c2IWH8oxzK4JQmAGNaSmXMyc7eGZU6vA3QDLwROzcxdgWXAB4fSU4NbkqhKJcNZmrAYWJyZV9fPv08V5GvN4JYkWn9WSWYuAR6IiO3rVa8AbhtKX52clCTadlbJvwDn1GeU/AY4ciiNGNyS1CaZuRDYbbjtGNySRFmXvBvckoQ3mZKk4hjcklSYkkolng4oSYVxxC1JWCqRpOKUVCoxuCWJskbc1rglqTCOuCUJoKARt8EtSZQU2wa3JAFOTkpSgcoJbicnJakwjrgliZLG2wa3JNXKiW6DW5Ioa3LSGrckFcbglqTCWCqRJMq6V4nBLUmUFdyWSiSpMAa3JBXGUokk4emAkqQWcsQtSZQ1OWlwSxJQ0iXvlkokqTCOuCWJksbbBrckAWWdVWJwSxJQ0pjb4JYkSoptJyclqTiOuCUJKGnMbXBLEmVNTloqkaTCGNySVBhLJZKE9yqRpAIZ3JJUlHJi2+CWJMCzSiRJLeSIW5KAkoolBrckUVJsG9ySVCsnuq1xSxLV5ORwliaPMSsifh0Rd0fEB4faV4NbktogIsYBXwH2A3YCDouInYbSlsEtSe3xYuDuzPxNZj4FfAd4zVAassYtSbTlkvctgAcani8GXjKUhkZtcE8Yt145MwUtFhGzM3NOp/sxGuSpizrdhVHD74uRNdzMiYjZwOyGVXN6/f/01X4O5ViWSsowe/Bd9DfI74tRJDPnZOZuDUvvX6qLgWkNz6cCDw7lWAa3JLXHtcC2EfHciFgHeCPw46E0NGpLJZI0lmTmyoh4N3AxMA44PTOHVPszuMtgHVN98fuiMJl5EXDRcNuJzCHVxiVJHWKNW5IKY3CPciN1iazGjog4PSKWRsStne6LOsPgHsVG8hJZjSlnArM63Ql1jsE9uo3YJbIaOzLzcuCPne6HOsfgHt36ukR2iw71RdIoYXCPbiN2iaykscPgHt1G7BJZSWOHwT26jdglspLGDoN7FMvMlUDPJbK3A+cN9RJZjR0RcS5wFbB9RCyOiKM73Se1l1dOSlJhHHFLUmEMbkkqjMEtSYUxuCWpMAa3JBXG4NaIi4hVEbEwIm6NiO9FxHrDaGtGRFxYPz5ooDskRsSkiHjXEI5xYkT821D7KLWbwa1WeCIzp2fmLsBTwDsbN0Zlrb/3MvPHmfnJAXaZBKx1cEulMbjValcA20TEVhFxe0R8FbgBmBYR+0bEVRFxQz0y3wDW3IP8johYALyup6GIOCIivlw/nhIRF0TETfWyB/BJYOt6tP+Zer/jIuLaiLg5Ik5qaOuE+j7nPwO2b9u7IY0Ag1stExHdVPcSv6VetT3wrczcFVgG/DuwT2a+ELgOeH9ETAC+ARwIvAzYrJ/m/wv4RWa+AHghsAj4IHBPPdo/LiL2Bbaluj3udOBFEfHyiHgR1e0DdqX6xbD7CH/pUkv5x4LVChMjYmH9+ArgNGBz4L7M/FW9/u+p/jjElREBsA7VZdw7AL/NzLsAIuJsYHYfx5gJvBUgM1cBf4qIZ/baZ996ubF+vgFVkD8DuCAzH6+P4f1fVBSDW63wRGZOb1xRh/OyxlXApZl5WK/9pjNyt64N4BOZ+fVex3jvCB5DajtLJeqUXwEvjYhtACJivYjYDrgDeG5EbF3vd1g/r78MOKZ+7biI2BD4C9VousfFwFENtfMtImIycDnw2oiYGBHPoCrLSMUwuNURmfkwcARwbkTcTBXkO2TmcqrSyE/rycn7+mniPcDeEXELcD2wc2Y+QlV6uTUiPpOZlwD/DVxV7/d94BmZeQPwXWAhcD5VOUcqhncHlKTCOOKWpMIY3JJUGINbkgpjcEtSYQxuSSqMwS1JhTG4JakwBrckFeb/A+Pkk/0WJJP6AAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 432x432 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "matrix = confusion_matrix(y_test, y_pred)\n",
    "\n",
    "tn, fp, fn, tp = confusion_matrix(y_test, y_pred).ravel()\n",
    "fig = plt.figure(figsize=(6,6))\n",
    "ax = sns.heatmap(matrix, annot=True,cmap='Greens', fmt='g')\n",
    "plt.title(\"Confusion Matrix for Alzheimer's\")\n",
    "plt.xlabel('Predicted')\n",
    "plt.ylabel('Actual')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "This is suspiciously high. I would expect ~ some ~ error, but perhaps the SVM is more subject to overfitting and the data just worked really well. It could be that I am using data specifically used to fit this type of classifier. This is worth looking into in the future."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Conclusions - Alzheimer's Classifier\n",
    "**Requirements:**\n",
    "+ Summarize the findings of your report\n",
    "+ Repeat your methodology and key findings for Section 2. and Section 4.\n",
    "+ Highlight what you found interesting\n",
    "+ Discuss what you would do to extend the project further\n",
    "\n",
    "It seems that by using only age, Clinical Dementia Rating, and years of education, I could accurately predict if the patient had Alzheimers. However, I do not trust my machine learning models one bit. After splitting my data set, I scaled my feature vectors, defined a model using Keras, compiled, and trained it. The total accuracy score ended up being 99%, which in my mind is disturbingly too high. I wrote some code to test the model on my testing data and I got a similar score. I definitely feel like I did something wrong. When I ran my code again, the accuracy jumped down to 68% and I was able to re-run it multiple times to get a higher score. I don't know what happened there. I tried using the SVM model the next time, and the results were even worse (better). This time, the accuracy went up to 100%, which means I probably used a dataset where my models were overfitting. For the SVM model, if I changed the kernel to RBF, my accuracy score goes down to 58%. I would love to hear your thoughts on this phenomenon. To extend this project further, I would implement the rest of my variables that I left out for simplicity's sake. I would also collect more data. Another discussion remains for the sample demographics of my data: Is it inclusive? Am I training a model for all race, ethnicity, ages, genders?"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# General thoughts: I really enjoyed this project and wished we had more like it. I am scared about my results though. For my regression models, the accuracy was really low and for my classifcation algorithms, my accuracy was too high. Python is fun that we have all these useful packages to abstract our methods, but I also have a hard time debugging since I don't know what's going on underneath the abstraction. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}