2019-10-19-2.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import pandas as pd \n",
    "import statsmodels.api as sm\n",
    "import statsmodels.formula.api as smf\n",
    "import seaborn as sns\n",
    "from sklearn.preprocessing import scale \n",
    "from sklearn.model_selection import train_test_split, GridSearchCV, cross_val_score\n",
    "from sklearn.metrics import confusion_matrix, accuracy_score, classification_report\n",
    "from sklearn.metrics import roc_auc_score,roc_curve\n",
    "import statsmodels.formula.api as smf\n",
    "import matplotlib.pyplot as plt\n",
    "from sklearn.neighbors import KNeighborsClassifier\n",
    "from sklearn.discriminant_analysis import LinearDiscriminantAnalysis\n",
    "from sklearn.linear_model import LogisticRegression\n",
    "from sklearn.svm import SVC\n",
    "from sklearn.naive_bayes import GaussianNB\n",
    "from sklearn import tree\n",
    "from sklearn.ensemble import RandomForestClassifier\n",
    "from sklearn.ensemble import GradientBoostingClassifier\n",
    "from xgboost import XGBClassifier\n",
    "from lightgbm import LGBMClassifier\n",
    "from catboost import CatBoostClassifier \n",
    "from sklearn.decomposition import PCA\n",
    "from sklearn.preprocessing import scale\n",
    "from sklearn.tree import DecisionTreeClassifier\n",
    "import warnings            \n",
    "warnings.filterwarnings(\"ignore\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "def prepare_data():\n",
    "    # verinin okunması\n",
    "    data = pd.read_csv('M30_EURUSD.csv',low_memory=False)\n",
    "    # soru işaretleri olan satırlar veriden silindi.\n",
    "    data = data.replace(\"?\", np.nan)\n",
    "    data = data.dropna()\n",
    "    # verinin feature larının ayrılması\n",
    "    df = data.iloc[:,5:430]\n",
    "    df_first = data.iloc[:,1:5]\n",
    "    df_result = data.iloc[:,430:-1]\n",
    "    return df, df_first, df_result"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "# parametrik fonk. tanımlaması\n",
    "# 1 : \n",
    "# 2 :\n",
    "\n",
    "def kategorikleri_dummy_yap(df):\n",
    "    cat_column_names = ['ind_7','ind_11','ind_24','ind_38','ind_54','ind_57','ind_60','ind_63','ind_66','ind_69','ind_72','ind_75',\n",
    "                    'ind_78','ind_81','ind_84','ind_87','ind_89','ind_91','ind_93','ind_95','ind_97','ind_99','ind_101',\n",
    "                    'ind_103','ind_105','ind_107','ind_109', 'ind_111', 'ind_113', 'ind_115','ind_138','ind_141','ind_144',\n",
    "                    'ind_157','ind_159','ind_161','ind_163','ind_165','ind_167','ind_169','ind_171','ind_173','ind_175',\n",
    "                    'ind_177','ind_182','ind_184','ind_187','ind_190','ind_193','ind_196','ind_199','ind_202','ind_205',\n",
    "                    'ind_208','ind_211','ind_213','ind_384','ind_386','ind_388','ind_390']\n",
    "    # categorical kolonların dummy var. oalrak değiştirdik\n",
    "    dms = pd.get_dummies(df[cat_column_names])\n",
    "    dms_none_cols = dms.filter(regex = '_NONE').columns\n",
    "    for i in dms_none_cols:\n",
    "        dms.drop(i,axis=1,inplace=True)\n",
    "    dms_red_cols = dms.filter(regex = '_RED').columns\n",
    "    for i in dms_red_cols:\n",
    "        dms.drop(i,axis=1,inplace=True)\n",
    "    #datadan categorical olan kolonları çıkarıyoruz ve type nı değiştiriyoruz\n",
    "    df_noncategoric = df.drop(cat_column_names,axis=1).astype(\"float64\")\n",
    "    df_all = pd.concat([df_noncategoric, dms], axis=1)\n",
    "    # y değerlerinin alınması\n",
    "    return df_all, df_noncategoric, dms"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 1.1 - dropping correlaritions\n",
    "def corr_df(df, corr_val):\n",
    "    corr_matrix = df_noncategoric.corr().abs()\n",
    "    # Select upper triangle of correlation matrix\n",
    "    upper = corr_matrix.where(np.triu(np.ones(corr_matrix.shape), k=1).astype(np.bool))\n",
    "\n",
    "    # Find index of feature columns with correlation greater than 0.95\n",
    "    to_high = [column for column in upper.columns if any(upper[column] > corr_val)]\n",
    "    df.drop(to_high, axis = 1, inplace = True)\n",
    "    return df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 1.2 RandomForest\n",
    "# bütün değişkenlerle yapılan random forest sonucu importance değeri verilen parametreden büyük olan değişkenleri döner\n",
    "def rand_forest(X, y, imp_value):\n",
    "    rf_model = RandomForestClassifier().fit(X, y)\n",
    "    Importance = pd.DataFrame({'Importance':rf_model.feature_importances_*100}, index = X.columns)\n",
    "    imp_values = Importance.sort_values(by = 'Importance', axis = 0, ascending = True)\n",
    "    imp_values = imp_values[imp_values['Importance']>imp_value]\n",
    "    col_names = imp_values.index   \n",
    "    return X[col_names]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 1.3 - pca\n",
    "def pca_fon(X, threshold):\n",
    "    pca = PCA()\n",
    "    X_pca = pca.fit_transform(scale(X))\n",
    "    arr = np.cumsum(np.round(pca.explained_variance_ratio_, decimals = 4)*100)\n",
    "    num_var = sum((arr < threshold*100)) + 1 \n",
    "    print('pca sonrası değişken sayısı: ',num_var)\n",
    "    X_pcad = pd.DataFrame(X_pca[:,0:num_var], index = X.index)\n",
    "    return X_pcad"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "def splitting(X, y, test_size):\n",
    "    X_train, X_test, y_train, y_test = train_test_split(X, y, test_size = test_size, shuffle = False)\n",
    "    return X_train, X_test, y_train, y_test"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 2.1.1 - multi lojistik\n",
    "def multi_logit(X_train, X_test, y_train, y_test):\n",
    "    logreg = LogisticRegression(C=1e5, solver='lbfgs', multi_class='multinomial')\n",
    "    log = logreg.fit(X_train, y_train)\n",
    "    y_pred = log.predict(X_test)\n",
    "    confusion_mat = confusion_matrix(y_test, y_pred)\n",
    "    accuracy = accuracy_score(y_test, y_pred)\n",
    "    print(\"Accuracy: \",accuracy)\n",
    "    print('-------------------------------')\n",
    "    print(\"Counfusion matrix: \\n\",confusion_mat)\n",
    "    print('-------------------------------')\n",
    "    print('Classification report')\n",
    "    print(classification_report(y_test, y_pred))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 2.1.2 - decision tree\n",
    "def dec_tree(X_train, X_test, y_train, y_test):\n",
    "    cart = DecisionTreeClassifier()\n",
    "    cart_model = cart.fit(X_train, y_train)\n",
    "    y_pred = cart_model.predict(X_test)\n",
    "    confusion_mat = confusion_matrix(y_test, y_pred)\n",
    "    accuracy = accuracy_score(y_test, y_pred)\n",
    "    print(\"Accuracy: \",accuracy)\n",
    "    print('-------------------------------')\n",
    "    print(\"Counfusion matrix: \\n\",confusion_mat)\n",
    "    print('-------------------------------')\n",
    "    print('Classification report')\n",
    "    print(classification_report(y_test, y_pred))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "def grad_boost(X_train, X_test, y_train, y_test):\n",
    "    from sklearn.ensemble import GradientBoostingClassifier\n",
    "    print('grad_boost----------------')\n",
    "    gbm_model = GradientBoostingClassifier().fit(X_train, y_train)\n",
    "    y_pred = gbm_model.predict(X_test)\n",
    "    confusion_mat = confusion_matrix(y_test, y_pred)\n",
    "    accuracy = accuracy_score(y_test, y_pred)\n",
    "    print(\"Accuracy: \",accuracy)\n",
    "    print('-------------------------------')\n",
    "    print(\"Counfusion matrix: \\n\",confusion_mat)\n",
    "    print('-------------------------------')\n",
    "    print('Classification report')\n",
    "    print(classification_report(y_test, y_pred))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "def xgb_boost(X_train, X_test, y_train, y_test):\n",
    "    from xgboost import XGBClassifier\n",
    "    print('xgb_boost----------------')\n",
    "    xgb_model = XGBClassifier().fit(X_train, y_train)\n",
    "    y_pred = xgb_model.predict(X_test)\n",
    "    confusion_mat = confusion_matrix(y_test, y_pred)\n",
    "    accuracy = accuracy_score(y_test, y_pred)\n",
    "    print(\"Accuracy: \",accuracy)\n",
    "    print('-------------------------------')\n",
    "    print(\"Counfusion matrix: \\n\",confusion_mat)\n",
    "    print('-------------------------------')\n",
    "    print('Classification report')\n",
    "    print(classification_report(y_test, y_pred))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "def lightGBM(X_train, X_test, y_train, y_test):\n",
    "    from lightgbm import LGBMClassifier\n",
    "    print('lightGBM----------------')\n",
    "    lgbm_model = LGBMClassifier(verbose=-1).fit(X_train,y_train)    \n",
    "    y_pred = lgbm_model.predict(X_test)\n",
    "    confusion_mat = confusion_matrix(y_test, y_pred)\n",
    "    accuracy = accuracy_score(y_test, y_pred)\n",
    "    print(\"Accuracy: \",accuracy)\n",
    "    print('-------------------------------')\n",
    "    print(\"Counfusion matrix: \\n\",confusion_mat)\n",
    "    print('-------------------------------')\n",
    "    print('Classification report')\n",
    "    print(classification_report(y_test, y_pred))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "def catBoost(X_train, X_test, y_train, y_test):\n",
    "    from catboost import CatBoostClassifier\n",
    "    print('CatBoost----------------')\n",
    "    cat_model = CatBoostClassifier().fit(X_train, y_train)\n",
    "    y_pred = cat_model.predict(X_test)\n",
    "    confusion_mat = confusion_matrix(y_test, y_pred)\n",
    "    accuracy = accuracy_score(y_test, y_pred)\n",
    "    print(\"Accuracy: \",accuracy)\n",
    "    print('-------------------------------')\n",
    "    print(\"Counfusion matrix: \\n\",confusion_mat)\n",
    "    print('-------------------------------')\n",
    "    print('Classification report')\n",
    "    print(classification_report(y_test, y_pred))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 2.1.3 - Boosting\n",
    "def boostings(X_train, X_test, y_train, y_test):\n",
    "    grad_boost(X_train, X_test, y_train, y_test)\n",
    "    xgb_boost(X_train, X_test, y_train, y_test)\n",
    "    # lightGBM(X_train, X_test, y_train, y_test)\n",
    "    # catBoost(X_train, X_test, y_train, y_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 2.1.2 - decision tree\n",
    "def dec_tree_withcv(X_train, X_test, y_train, y_test):\n",
    "    cart = DecisionTreeClassifier()\n",
    "    cart_model = cart.fit(X_train, y_train)\n",
    "    y_pred = cart_model.predict(X_test)\n",
    "    confusion_mat = confusion_matrix(y_test, y_pred)\n",
    "    accuracy = accuracy_score(y_test, y_pred)\n",
    "    print(\"Accuracy: \",accuracy)\n",
    "    print('-------------------------------')\n",
    "    print(\"Counfusion matrix: \\n\",confusion_mat)\n",
    "    print('-------------------------------')\n",
    "    print('Classification report')\n",
    "    print(classification_report(y_test, y_pred))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "# verinin okunması - df: ilk 5 kolon ve result'lar hariç kolonlar, df_first: ilk 5 kolon, df_result: sonuç kolonları\n",
    "df, df_first, df_result = prepare_data()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "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>0</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>2001</th>\n",
       "      <td>BUY</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2002</th>\n",
       "      <td>BUY</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2003</th>\n",
       "      <td>BUY</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2004</th>\n",
       "      <td>BUY</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2005</th>\n",
       "      <td>BUY</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        0\n",
       "2001  BUY\n",
       "2002  BUY\n",
       "2003  BUY\n",
       "2004  BUY\n",
       "2005  BUY"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# y sütunlarında geçen verilerden en çok tekrar edenler y sütunu olarak alınmıştır.\n",
    "# desc = y.T.describe(include='all')\n",
    "y_max1 = df_result.mode(axis=1)\n",
    "y = pd.DataFrame(y_max1[0])\n",
    "y.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 1.1 den gelen veriler (non correlatedlardan gelenler)\n",
    "df_all, df_noncategoric, dms = kategorikleri_dummy_yap(df)\n",
    "df_noncorr = corr_df(df_noncategoric, 0.50)\n",
    "X1_1 = pd.concat([df_first, df_noncorr, dms], axis=1)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 1.2 den gelen veriler. \n",
    "# Notlar: \n",
    "# 1 - y için iterasyon denenebilir. y kolonu '220_signal' seçilmiştir.\n",
    "# 2- importance treshold'u 0.05 seçilmiştir, cv yapılabilir.\n",
    "\n",
    "X_raw = pd.concat([df_first,df_all], axis=1) \n",
    "X1_2 = rand_forest(X_raw, y, 0.05)    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "pca sonrası değişken sayısı:  191\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "((19901, 486), (19901, 191))"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# 1.3 den gelen veriler.\n",
    "X_raw2 = pd.concat([df_first,df_all], axis=1) \n",
    "X1_3 = pca_fon(X_raw2, 0.99)\n",
    "X_raw2.shape, X1_3.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "X1_1 için multi log\n",
      "Accuracy:  0.4153691612255148\n",
      "-------------------------------\n",
      "Counfusion matrix: \n",
      " [[293   0 201]\n",
      " [197   0 119]\n",
      " [647   0 534]]\n",
      "-------------------------------\n",
      "Classification report\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "         BUY       0.26      0.59      0.36       494\n",
      "        NONE       0.00      0.00      0.00       316\n",
      "        SELL       0.63      0.45      0.52      1181\n",
      "\n",
      "    accuracy                           0.42      1991\n",
      "   macro avg       0.29      0.35      0.29      1991\n",
      "weighted avg       0.43      0.42      0.40      1991\n",
      "\n",
      "*************************************************************************************\n"
     ]
    }
   ],
   "source": [
    "X_train1, X_test1, y_train, y_test = splitting(X1_1, y, 0.10)\n",
    "print('X1_1 için multi log')\n",
    "multi_logit(X_train1, X_test1, y_train, y_test)\n",
    "print('*************************************************************************************')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "X1_2 için multi log\n",
      "Accuracy:  0.38623807132094423\n",
      "-------------------------------\n",
      "Counfusion matrix: \n",
      " [[331   3 160]\n",
      " [238  12  66]\n",
      " [735  20 426]]\n",
      "-------------------------------\n",
      "Classification report\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "         BUY       0.25      0.67      0.37       494\n",
      "        NONE       0.34      0.04      0.07       316\n",
      "        SELL       0.65      0.36      0.46      1181\n",
      "\n",
      "    accuracy                           0.39      1991\n",
      "   macro avg       0.42      0.36      0.30      1991\n",
      "weighted avg       0.50      0.39      0.38      1991\n",
      "\n",
      "*************************************************************************************\n"
     ]
    }
   ],
   "source": [
    "X_train2, X_test2, y_train, y_test = splitting(X1_2, y, 0.10)\n",
    "print('X1_2 için multi log')\n",
    "multi_logit(X_train2, X_test2, y_train, y_test)\n",
    "print('*************************************************************************************')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "X1_3 için multi log\n",
      "Accuracy:  0.3219487694625816\n",
      "-------------------------------\n",
      "Counfusion matrix: \n",
      " [[362   4 128]\n",
      " [248  19  49]\n",
      " [848  73 260]]\n",
      "-------------------------------\n",
      "Classification report\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "         BUY       0.25      0.73      0.37       494\n",
      "        NONE       0.20      0.06      0.09       316\n",
      "        SELL       0.59      0.22      0.32      1181\n",
      "\n",
      "    accuracy                           0.32      1991\n",
      "   macro avg       0.35      0.34      0.26      1991\n",
      "weighted avg       0.45      0.32      0.30      1991\n",
      "\n",
      "*************************************************************************************\n"
     ]
    }
   ],
   "source": [
    "X_train3, X_test3, y_train, y_test = splitting(X1_3, y, 0.10)\n",
    "print('X1_3 için multi log')\n",
    "multi_logit(X_train3, X_test3, y_train, y_test)\n",
    "print('*************************************************************************************')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "X1_1 için dec tree\n",
      "Accuracy:  0.29884480160723254\n",
      "-------------------------------\n",
      "Counfusion matrix: \n",
      " [[164  90 240]\n",
      " [138  78 100]\n",
      " [425 403 353]]\n",
      "-------------------------------\n",
      "Classification report\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "         BUY       0.23      0.33      0.27       494\n",
      "        NONE       0.14      0.25      0.18       316\n",
      "        SELL       0.51      0.30      0.38      1181\n",
      "\n",
      "    accuracy                           0.30      1991\n",
      "   macro avg       0.29      0.29      0.27      1991\n",
      "weighted avg       0.38      0.30      0.32      1991\n",
      "\n",
      "*************************************************************************************\n"
     ]
    }
   ],
   "source": [
    "X_train1, X_test1, y_train, y_test = splitting(X1_1, y, 0.10)\n",
    "print('X1_1 için dec tree')\n",
    "dec_tree(X_train1, X_test1, y_train, y_test)\n",
    "print('*************************************************************************************')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "X1_2 için dec tree\n",
      "Accuracy:  0.27122049221496736\n",
      "-------------------------------\n",
      "Counfusion matrix: \n",
      " [[345  85  64]\n",
      " [275  36   5]\n",
      " [663 359 159]]\n",
      "-------------------------------\n",
      "Classification report\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "         BUY       0.27      0.70      0.39       494\n",
      "        NONE       0.07      0.11      0.09       316\n",
      "        SELL       0.70      0.13      0.23      1181\n",
      "\n",
      "    accuracy                           0.27      1991\n",
      "   macro avg       0.35      0.32      0.23      1991\n",
      "weighted avg       0.49      0.27      0.24      1991\n",
      "\n",
      "*************************************************************************************\n"
     ]
    }
   ],
   "source": [
    "X_train2, X_test2, y_train, y_test = splitting(X1_2, y, 0.10)\n",
    "print('X1_2 için dec tree')\n",
    "dec_tree(X_train2, X_test2, y_train, y_test)\n",
    "print('*************************************************************************************')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0  kolonu için sonuçlar:\n",
      "Accuracy:  0.3751883475640382\n",
      "-------------------------------\n",
      "Counfusion matrix: \n",
      " [[237  46 211]\n",
      " [141  35 140]\n",
      " [564 142 475]]\n",
      "-------------------------------\n",
      "Classification report\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "         BUY       0.25      0.48      0.33       494\n",
      "        NONE       0.16      0.11      0.13       316\n",
      "        SELL       0.58      0.40      0.47      1181\n",
      "\n",
      "    accuracy                           0.38      1991\n",
      "   macro avg       0.33      0.33      0.31      1991\n",
      "weighted avg       0.43      0.38      0.38      1991\n",
      "\n",
      "*************************************************************************************\n"
     ]
    }
   ],
   "source": [
    "X_train3, X_test3, y_train, y_test = splitting(X1_3, y, 0.10)\n",
    "for i in y.columns:\n",
    "    print(i, ' kolonu için sonuçlar:')\n",
    "    dec_tree(X_train3, X_test3, y_train[i], y_test[i])\n",
    "    print('*************************************************************************************')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0  kolonu için sonuçlar:\n",
      "grad_boost----------------\n",
      "Accuracy:  0.34433093284207\n",
      "-------------------------------\n",
      "Counfusion matrix: \n",
      " [[1929   54   40]\n",
      " [1151    2   82]\n",
      " [2583    5  125]]\n",
      "-------------------------------\n",
      "Classification report\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "         BUY       0.34      0.95      0.50      2023\n",
      "        NONE       0.03      0.00      0.00      1235\n",
      "        SELL       0.51      0.05      0.08      2713\n",
      "\n",
      "    accuracy                           0.34      5971\n",
      "   macro avg       0.29      0.33      0.20      5971\n",
      "weighted avg       0.35      0.34      0.21      5971\n",
      "\n",
      "xgb_boost----------------\n",
      "Accuracy:  0.3481828839390387\n",
      "-------------------------------\n",
      "Counfusion matrix: \n",
      " [[1996    0   27]\n",
      " [1154    0   81]\n",
      " [2630    0   83]]\n",
      "-------------------------------\n",
      "Classification report\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "         BUY       0.35      0.99      0.51      2023\n",
      "        NONE       0.00      0.00      0.00      1235\n",
      "        SELL       0.43      0.03      0.06      2713\n",
      "\n",
      "    accuracy                           0.35      5971\n",
      "   macro avg       0.26      0.34      0.19      5971\n",
      "weighted avg       0.31      0.35      0.20      5971\n",
      "\n",
      "*************************************************************************************\n"
     ]
    }
   ],
   "source": [
    "X_train1, X_test1, y_train, y_test = splitting(X1_1, y, 0.30)\n",
    "for i in y.columns:\n",
    "    print(i, ' kolonu için sonuçlar:')\n",
    "    boostings(X_train1, X_test1, y_train[i], y_test[i])\n",
    "    print('*************************************************************************************')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0  kolonu için sonuçlar:\n",
      "grad_boost----------------\n",
      "Accuracy:  0.36911740077039024\n",
      "-------------------------------\n",
      "Counfusion matrix: \n",
      " [[1843    9  171]\n",
      " [1118    0  117]\n",
      " [2348    4  361]]\n",
      "-------------------------------\n",
      "Classification report\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "         BUY       0.35      0.91      0.50      2023\n",
      "        NONE       0.00      0.00      0.00      1235\n",
      "        SELL       0.56      0.13      0.21      2713\n",
      "\n",
      "    accuracy                           0.37      5971\n",
      "   macro avg       0.30      0.35      0.24      5971\n",
      "weighted avg       0.37      0.37      0.27      5971\n",
      "\n",
      "xgb_boost----------------\n",
      "Accuracy:  0.4086417685479819\n",
      "-------------------------------\n",
      "Counfusion matrix: \n",
      " [[1730    0  293]\n",
      " [ 970    0  265]\n",
      " [2003    0  710]]\n",
      "-------------------------------\n",
      "Classification report\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "         BUY       0.37      0.86      0.51      2023\n",
      "        NONE       0.00      0.00      0.00      1235\n",
      "        SELL       0.56      0.26      0.36      2713\n",
      "\n",
      "    accuracy                           0.41      5971\n",
      "   macro avg       0.31      0.37      0.29      5971\n",
      "weighted avg       0.38      0.41      0.34      5971\n",
      "\n",
      "*************************************************************************************\n"
     ]
    }
   ],
   "source": [
    "X_train2, X_test2, y_train, y_test = splitting(X1_2, y, 0.30)\n",
    "for i in y.columns:\n",
    "    print(i, ' kolonu için sonuçlar:')\n",
    "    boostings(X_train2, X_test2, y_train[i], y_test[i])\n",
    "    print('*************************************************************************************')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0  kolonu için sonuçlar:\n",
      "grad_boost----------------\n",
      "Accuracy:  0.38737229944732876\n",
      "-------------------------------\n",
      "Counfusion matrix: \n",
      " [[1010   21  992]\n",
      " [ 676   20  539]\n",
      " [1419   11 1283]]\n",
      "-------------------------------\n",
      "Classification report\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "         BUY       0.33      0.50      0.39      2023\n",
      "        NONE       0.38      0.02      0.03      1235\n",
      "        SELL       0.46      0.47      0.46      2713\n",
      "\n",
      "    accuracy                           0.39      5971\n",
      "   macro avg       0.39      0.33      0.30      5971\n",
      "weighted avg       0.40      0.39      0.35      5971\n",
      "\n",
      "xgb_boost----------------\n",
      "Accuracy:  0.3937363925640596\n",
      "-------------------------------\n",
      "Counfusion matrix: \n",
      " [[1042    3  978]\n",
      " [ 687    1  547]\n",
      " [1405    0 1308]]\n",
      "-------------------------------\n",
      "Classification report\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "         BUY       0.33      0.52      0.40      2023\n",
      "        NONE       0.25      0.00      0.00      1235\n",
      "        SELL       0.46      0.48      0.47      2713\n",
      "\n",
      "    accuracy                           0.39      5971\n",
      "   macro avg       0.35      0.33      0.29      5971\n",
      "weighted avg       0.37      0.39      0.35      5971\n",
      "\n",
      "*************************************************************************************\n"
     ]
    }
   ],
   "source": [
    "X_train3, X_test3, y_train, y_test = splitting(X1_3, y, 0.30)\n",
    "for i in y.columns:\n",
    "    print(i, ' kolonu için sonuçlar:')\n",
    "    boostings(X_train3, X_test3, y_train[i], y_test[i])\n",
    "    print('*************************************************************************************')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# yol haritası\n",
    "\n",
    "1. verilerin sadeleştirilmesi \n",
    "\n",
    "    1.1 correlatedları atarak non correlated ları bul \n",
    " \n",
    "    1.2 RandomForest'dan important değişkenleri bul \n",
    " \n",
    "    1.3 pca  \n",
    " \n",
    " \n",
    "2. algoritmalar \n",
    "\n",
    "    2.1 algoritmaları fonk. olarak yaz\n",
    " \n",
    "        2.1.1 loj reg\n",
    "  \n",
    "        2.1.2 decision tree\n",
    "  \n",
    "        2.1.3 boosting\n",
    "      \n",
    "    2.2 cross validations\n",
    " \n",
    "    2.3 1'de bulduğun verilerle bütün algoritmaları çalıştır, sonuçları kıyasla \n",
    " \n",
    " Notlar:\n",
    " - ilk 5 sütun correlation a koyulmadı. bunların da koyulması gerekir mi?"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "X1_1 için dec tree\n",
      "Accuracy:  0.2812656956303365\n",
      "-------------------------------\n",
      "Counfusion matrix: \n",
      " [[208  72 214]\n",
      " [222  44  50]\n",
      " [525 348 308]]\n",
      "-------------------------------\n",
      "Classification report\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "         BUY       0.22      0.42      0.29       494\n",
      "        NONE       0.09      0.14      0.11       316\n",
      "        SELL       0.54      0.26      0.35      1181\n",
      "\n",
      "    accuracy                           0.28      1991\n",
      "   macro avg       0.28      0.27      0.25      1991\n",
      "weighted avg       0.39      0.28      0.30      1991\n",
      "\n"
     ]
    }
   ],
   "source": [
    "X_train1, X_test1, y_train, y_test = splitting(X1_1, y, 0.10)\n",
    "print('X1_1 için dec tree')\n",
    "dec_tree(X_train1, X_test1, y_train, y_test)\n",
    "print\n",
    "cart = DecisionTreeClassifier()\n",
    "cart_model = cart.fit(X_train1, y_train)\n",
    "y_pred = cart_model.predict(X_test1)\n",
    "confusion_mat = confusion_matrix(y_test, y_pred)\n",
    "accuracy = accuracy_score(y_test, y_pred)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 10 folds for each of 196 candidates, totalling 1960 fits\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Using backend LokyBackend with 8 concurrent workers.\n",
      "[Parallel(n_jobs=-1)]: Done  25 tasks      | elapsed:    3.7s\n",
      "[Parallel(n_jobs=-1)]: Done 146 tasks      | elapsed:    8.3s\n",
      "[Parallel(n_jobs=-1)]: Done 349 tasks      | elapsed:   23.4s\n",
      "[Parallel(n_jobs=-1)]: Done 632 tasks      | elapsed:   53.8s\n",
      "[Parallel(n_jobs=-1)]: Done 997 tasks      | elapsed:  1.7min\n",
      "[Parallel(n_jobs=-1)]: Done 1442 tasks      | elapsed:  2.8min\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "En iyi parametreler : {'max_depth': 2, 'min_samples_split': 2}\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Done 1960 out of 1960 | elapsed:  4.0min finished\n"
     ]
    }
   ],
   "source": [
    "cart_grid = {\"max_depth\":[2,5,8,10,15,20,25,30,40,50,70,90,120,150], \"min_samples_split\":[2,5,8,10,15,20,25,30,40,50,70,90,120,150]}\n",
    "cart_cv = GridSearchCV(cart, cart_grid, cv=10, n_jobs =-1, verbose = 2)\n",
    "cart_cv_model = cart_cv.fit(X_train1, y_train)\n",
    "print('En iyi parametreler : ' + str(cart_cv_model.best_params_))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "accuracy :  0.24962330487192366\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "         BUY       0.31      0.54      0.40       494\n",
      "        NONE       0.20      0.73      0.32       316\n",
      "        SELL       0.00      0.00      0.00      1181\n",
      "\n",
      "    accuracy                           0.25      1991\n",
      "   macro avg       0.17      0.42      0.24      1991\n",
      "weighted avg       0.11      0.25      0.15      1991\n",
      "\n"
     ]
    }
   ],
   "source": [
    "cart = DecisionTreeClassifier(max_depth = 2, min_samples_split=2 )\n",
    "cart_tuned = cart.fit(X_train1, y_train)\n",
    "y_pred = cart_tuned.predict(X_test1)\n",
    "accuracy = accuracy_score(y_test, y_pred)\n",
    "print('accuracy : ', accuracy)\n",
    "print(classification_report(y_test, y_pred))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 10 folds for each of 162 candidates, totalling 1620 fits\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Using backend LokyBackend with 8 concurrent workers.\n",
      "[Parallel(n_jobs=-1)]: Done  25 tasks      | elapsed:    3.4s\n",
      "[Parallel(n_jobs=-1)]: Done 146 tasks      | elapsed:    6.7s\n",
      "[Parallel(n_jobs=-1)]: Done 349 tasks      | elapsed:   13.7s\n",
      "[Parallel(n_jobs=-1)]: Done 632 tasks      | elapsed:   27.0s\n",
      "[Parallel(n_jobs=-1)]: Done 997 tasks      | elapsed:   52.2s\n",
      "[Parallel(n_jobs=-1)]: Done 1442 tasks      | elapsed:  1.6min\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "En iyi parametreler : {'max_depth': 1, 'min_samples_split': 2}\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Done 1620 out of 1620 | elapsed:  1.9min finished\n"
     ]
    }
   ],
   "source": [
    "cart_grid = {\"max_depth\":range(1,10), \"min_samples_split\":range(2,20)}\n",
    "cart_cv = GridSearchCV(cart, cart_grid, cv=10, n_jobs =-1, verbose = 2)\n",
    "cart_cv_model = cart_cv.fit(X_train1, y_train)\n",
    "print('En iyi parametreler : ' + str(cart_cv_model.best_params_))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "accuracy :  0.24811652435961828\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "         BUY       0.25      1.00      0.40       494\n",
      "        NONE       0.00      0.00      0.00       316\n",
      "        SELL       0.00      0.00      0.00      1181\n",
      "\n",
      "    accuracy                           0.25      1991\n",
      "   macro avg       0.08      0.33      0.13      1991\n",
      "weighted avg       0.06      0.25      0.10      1991\n",
      "\n"
     ]
    }
   ],
   "source": [
    "cart = DecisionTreeClassifier(max_depth = 1, min_samples_split=2)\n",
    "cart_tuned = cart.fit(X_train1, y_train)\n",
    "y_pred = cart_tuned.predict(X_test1)\n",
    "accuracy_score = accuracy_score(y_test, y_pred)\n",
    "print('accuracy : ', accuracy_score)\n",
    "print(classification_report(y_test, y_pred))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train1, X_test1, y_train, y_test = splitting(X1_2, y, 0.10)\n",
    "print('X1_1 için dec tree')\n",
    "dec_tree(X_train1, X_test1, y_train, y_test)\n",
    "print\n",
    "cart = DecisionTreeClassifier()\n",
    "cart_model = cart.fit(X_train1, y_train)\n",
    "y_pred = cart_model.predict(X_test1)\n",
    "confusion_mat = confusion_matrix(y_test, y_pred)\n",
    "accuracy = accuracy_score(y_test, y_pred)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# kmeans - 3 küme\n",
    "from sklearn.cluster import KMeans\n",
    "kmeans = KMeans(n_clusters=3, random_state=0)\n",
    "kmeans.fit(X_train1)\n",
    "labels = kmeans.predict(X_train1)\n",
    "\n",
    "plt.scatter(X_train1['ind_14'],X_train1['ind_176'],c = labels)\n",
    "plt.xlabel('1')\n",
    "plt.xlabel('2')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 79,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[   0    1    2]\n",
      " [2906 8719 6285]]\n"
     ]
    }
   ],
   "source": [
    "unique_elements, counts_elements = np.unique(labels, return_counts=True)\n",
    "print(np.asarray((unique_elements, counts_elements)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 115,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<bound method NDFrame.head of        0\n",
       "2001   1\n",
       "2002   1\n",
       "2003   1\n",
       "2004   1\n",
       "2005   1\n",
       "...   ..\n",
       "25623  2\n",
       "25624  2\n",
       "25625  2\n",
       "25626  2\n",
       "25627  2\n",
       "\n",
       "[17910 rows x 1 columns]>"
      ]
     },
     "execution_count": 115,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cleanup_nums = {'BUY':1,'SELL':2,'NONE':0}\n",
    "y_cluster = y_train.replace(cleanup_nums)\n",
    "y_cluster.head"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 116,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "False    10742\n",
       "True      7168\n",
       "Name: 0, dtype: int64"
      ]
     },
     "execution_count": 116,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "(y_cluster[0] == labels).value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 118,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x240c2f724e0>"
      ]
     },
     "execution_count": 118,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "sns.countplot((y_cluster[0] == labels), palette=\"bwr\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 128,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[   0    1    2]\n",
      " [7404 5006 5500]]\n"
     ]
    }
   ],
   "source": [
    "# kmeans random forest verisi ile\n",
    "X_train2, X_test2, y_train, y_test = splitting(X1_2, y, 0.10)\n",
    "kmeans = KMeans(n_clusters=3, random_state=0)\n",
    "kmeans.fit(X_train2)\n",
    "labels1 = kmeans.predict(X_train2)\n",
    "unique_elements, counts_elements = np.unique(labels1, return_counts=True)\n",
    "print(np.asarray((unique_elements, counts_elements)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 129,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "False    12322\n",
       "True      5588\n",
       "Name: 0, dtype: int64"
      ]
     },
     "execution_count": 129,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cleanup_nums = {'BUY':2,'SELL':1,'NONE':0}\n",
    "y_cluster = y_train.replace(cleanup_nums)\n",
    "(y_cluster[0] == labels1).value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 130,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[   0    1]\n",
      " [9252 8658]]\n"
     ]
    }
   ],
   "source": [
    "kmeans = KMeans(n_clusters=2, random_state=0)\n",
    "kmeans.fit(X_train2)\n",
    "labels2 = kmeans.predict(X_train2)\n",
    "unique_elements, counts_elements = np.unique(labels2, return_counts=True)\n",
    "print(np.asarray((unique_elements, counts_elements)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 131,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[   0    1]\n",
      " [9152 8758]]\n"
     ]
    }
   ],
   "source": [
    "a = ( labels1 == labels2)\n",
    "unique_elements, counts_elements = np.unique(a, return_counts=True)\n",
    "print(np.asarray((unique_elements, counts_elements)))"
   ]
  },
  {
   "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": 2
}