examples/train_and_evaluate_custom_model.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Document Classification Demo\n",
    "\n",
    "In this notebook we will learn how to use the **Document Classification** service using a python based client library. \n",
    "In this demo, we train a model for classification and evaluate its performance on a small example dataset.\n",
    "\n",
    "For the demo, we prepared this Jupyter Notebook which demonstrates the use of this client library to invoke the most important functions of the Document Classification REST API. \n",
    "\n",
    "### Demo content\n",
    "\n",
    "1. Create a dataset\n",
    "1. Train and deploy a custom model\n",
    "1. Run classification requests using the custom model\n",
    "\n",
    "### Initial steps\n",
    "\n",
    "**Note:** The following steps are required for the execution of this notebook(Dataset creation and training/deployment features are not available on trial accounts):\n",
    "\n",
    "1. See this [tutorial](https://developers.sap.com/tutorials/hcp-create-trial-account.html) to learn how to create a **SAP Cloud Platform Trial account**.\n",
    "2. See this [tutorial](https://developers.sap.com/tutorials/cp-aibus-dc-service-instance.html) to learn how to create a **Document Classification service instance**.\n",
    "\n",
    "Please keep the **service key** for the Document Classification service created in the 2nd step above at hand as we will use it below!"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 1. Installation and invocation of the client library\n",
    "\n",
    "In this section we will:\n",
    "- Pass the service key we obtained in the **initial steps above** to the client library which will use it to communicate with the [Document Classification REST API](https://aiservices-trial-dc.cfapps.eu10.hana.ondemand.com/document-classification/v1/)\n",
    "- Import the client library package and create an instance of the client\n",
    "- Try out and learn about the [API of the client library](https://github.com/SAP/document-classification-client/blob/master/API.md)\n",
    "\n",
    "An example dataset is provided in the repo, you can exlpore the structure of the dataset required [here](https://github.com/SAP/document-classification-client/tree/master/examples/data/training_data)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-11-09T04:41:04.535541Z",
     "start_time": "2019-11-09T04:41:04.531191Z"
    }
   },
   "outputs": [],
   "source": [
    "# Pass the credentials to the client library\n",
    "\n",
    "# Please copy the content of the service key you created in the Prerequisites above here after \"service_key = \"\n",
    "# An EXAMPLE is given to show how the service key needs to be pasted here\n",
    "service_key = {\n",
    "  \"url\": \"https://environment.eu10.hana.ondemand.com\",\n",
    "  \"uaa\": {\n",
    "    \"tenantmode\": \"dedicated\",\n",
    "    \"sburl\": \"*******\",\n",
    "    \"subaccountid\": \"*******\",\n",
    "    \"clientid\": \"*******\",\n",
    "    \"xsappname\": \"*******\",\n",
    "    \"clientsecret\": \"*******\",\n",
    "    \"url\": \"*******\",\n",
    "    \"uaadomain\": \"*******\",\n",
    "    \"verificationkey\": \"*******\",\n",
    "    \"apiurl\": \"*******\",\n",
    "    \"identityzone\": \"*******\",\n",
    "    \"identityzoneid\": \"*******\",\n",
    "    \"tenantid\": \"*******\",\n",
    "    \"zoneid\": \"*******\"\n",
    "  },\n",
    "  \"swagger\": \"/document-classification/v1\"\n",
    "}\n",
    "\n",
    "# These 4 values are needed for communicating with the Document Classification REST API\n",
    "api_url = service_key[\"url\"]\n",
    "uaa_server = service_key[\"uaa\"][\"url\"]\n",
    "client_id = service_key[\"uaa\"][\"clientid\"]\n",
    "client_secret = service_key[\"uaa\"][\"clientsecret\"]\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<div class=\"alert alert-block alert-warning\">\n",
    "<b>Warning:</b> Please provide a model name for training.\n",
    "</div>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Model specific configuration\n",
    "model_name = \"\" # choose an arbitrary model name for the model trained here, will be assigned to the trained model for identification purposes\n",
    "dataset_folder = \"data/training_data/\" # should point to (relative or absolute) path containing dataset"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Initialize Demo"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-11-09T04:35:31.032608Z",
     "start_time": "2019-11-09T04:35:31.018851Z"
    }
   },
   "outputs": [],
   "source": [
    "# Import client class from document classification client python package\n",
    "from sap_document_classification_client.dc_api_client import DCApiClient\n",
    "# Create an instance of this class with the credentials defined in the last cell\n",
    "my_dc_client = DCApiClient(api_url, client_id, client_secret, uaa_server)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Create Dataset for training of a new model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-11-09T03:53:04.220110Z",
     "start_time": "2019-11-09T03:53:03.806452Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Dataset created with datasetId: 1b6656f3-1774-41ab-bc27-07bd87876052\n"
     ]
    }
   ],
   "source": [
    "# Create Training dataset\n",
    "response = my_dc_client.create_dataset()\n",
    "training_dataset_id = response[\"datasetId\"]\n",
    "print(\"Dataset created with datasetId: {}\".format(training_dataset_id))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-09-04T09:26:01.068519Z",
     "start_time": "2019-09-04T09:23:45.230188Z"
    },
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Uploading training documents to the dataset\n",
      "Finished uploading training documents to the dataset\n"
     ]
    }
   ],
   "source": [
    "# Upload training documents to the dataset from training directory\n",
    "print(\"Uploading training documents to the dataset\")\n",
    "my_dc_client.upload_documents_directory_to_dataset(training_dataset_id, dataset_folder)\n",
    "print(\"Finished uploading training documents to the dataset\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-09-04T09:36:49.651509Z",
     "start_time": "2019-09-04T09:36:49.370143Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Dataset statistics\n",
      "{'createdAt': '2020-03-17T15:35:13.120000',\n",
      " 'datasetId': '1b6656f3-1774-41ab-bc27-07bd87876052',\n",
      " 'documents': {'FAILED': 0, 'RUNNING': 0, 'SUCCEEDED': 202, 'total': 202},\n",
      " 'groundTruths': [{'characteristic': 'Language',\n",
      "                   'classes': [{'total': 50, 'value': 'English'},\n",
      "                               {'total': 50, 'value': 'German'},\n",
      "                               {'total': 52, 'value': 'Other'},\n",
      "                               {'total': 50, 'value': 'Both'}],\n",
      "                   'total': 202}],\n",
      " 'updatedAt': '2020-03-17T15:35:13.120000'}\n"
     ]
    }
   ],
   "source": [
    "# Pretty print the dataset statistics\n",
    "from pprint import pprint\n",
    "print(\"Dataset statistics\")\n",
    "dataset_stats = my_dc_client.get_dataset_info(training_dataset_id)\n",
    "pprint(dataset_stats)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 800x1200 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Visualization of label distribution\n",
    "%matplotlib inline\n",
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "\n",
    "nrCharacteristics = len(dataset_stats[\"groundTruths\"])\n",
    "fig, (ax) = plt.subplots(nrCharacteristics,1, figsize=(10, 15), dpi=80, facecolor='w', edgecolor='k')\n",
    "if nrCharacteristics==1:\n",
    "    ax = np.array((ax,)) \n",
    "for i in range(nrCharacteristics):\n",
    "    keys = [element[\"value\"] for element in  dataset_stats[\"groundTruths\"][i][\"classes\"]]\n",
    "    total = [element[\"total\"] for element in  dataset_stats[\"groundTruths\"][i][\"classes\"]]\n",
    "    ax[i].set_ylabel(\"Absolute\")\n",
    "    ax[i].bar(keys, total)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Training"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-09-04T09:49:52.977862Z",
     "start_time": "2019-09-04T09:49:52.633337Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Start training job from model with modelName dc_example_dataset\n",
      "{'accuracy': 0.8846153846153846,\n",
      " 'datasetId': '1b6656f3-1774-41ab-bc27-07bd87876052',\n",
      " 'details': 'Model training logs:\\n'\n",
      "            '------\\n'\n",
      "            '[2020-03-17 15:37:29 +0000] [8] [INFO] [data] Loading data from '\n",
      "            'bdp-c70d39c1-d258-4752-bca8-18051c5ace9e ...\\n'\n",
      "            '[2020-03-17 15:37:29 +0000] [8] [INFO] [data] Preprocessing the '\n",
      "            \"'training' dataset ...\\n\"\n",
      "            '[2020-03-17 15:37:29 +0000] [8] [INFO] [data] Preprocessing the '\n",
      "            \"'validation' dataset ...\\n\"\n",
      "            '[2020-03-17 15:37:29 +0000] [8] [INFO] [data] Preprocessing the '\n",
      "            \"'test' dataset ...\\n\"\n",
      "            '[2020-03-17 15:37:29 +0000] [8] [INFO] [tfidf] Vectorizing '\n",
      "            'data...\\n'\n",
      "            '[2020-03-17 15:37:29 +0000] [8] [INFO] [tfidf] Transformed '\n",
      "            'datasets via tfidf. Found 238 (or hit max number of features '\n",
      "            '100000) distinct ngrams (1 <= n <= 2) occurring at least 10 '\n",
      "            'times.\\n'\n",
      "            '[2020-03-17 15:37:29 +0000] [8] [INFO] [classifier] Training with '\n",
      "            \"the 'rf' classifier\\n\"\n",
      "            '[2020-03-17 15:37:29 +0000] [8] [INFO] [classifier] Performing '\n",
      "            'grid search using validation set...\\n'\n",
      "            '[2020-03-17 15:37:32 +0000] [8] [INFO] [classifier] Sample-based '\n",
      "            'precision, recall, fscore is not meaningful outside multilabel '\n",
      "            'classification. Switching instead to weighted averages.\\n'\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.895 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 1, \"\n",
      "            \"'min_samples_split': 2, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 32, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.947 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 1, \"\n",
      "            \"'min_samples_split': 2, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 64, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.947 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 1, \"\n",
      "            \"'min_samples_split': 2, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 128, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.947 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 1, \"\n",
      "            \"'min_samples_split': 2, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 256, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.947 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 1, \"\n",
      "            \"'min_samples_split': 2, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 512, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.895 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 1, \"\n",
      "            \"'min_samples_split': 5, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 32, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.895 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 1, \"\n",
      "            \"'min_samples_split': 5, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 64, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.947 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 1, \"\n",
      "            \"'min_samples_split': 5, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 128, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.947 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 1, \"\n",
      "            \"'min_samples_split': 5, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 256, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.947 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 1, \"\n",
      "            \"'min_samples_split': 5, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 512, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.895 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 1, \"\n",
      "            \"'min_samples_split': 10, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 32, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.947 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 1, \"\n",
      "            \"'min_samples_split': 10, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 64, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.947 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 1, \"\n",
      "            \"'min_samples_split': 10, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 128, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.947 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 1, \"\n",
      "            \"'min_samples_split': 10, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 256, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.947 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 1, \"\n",
      "            \"'min_samples_split': 10, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 512, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 1.000 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 1, \"\n",
      "            \"'min_samples_split': 20, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 32, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 1.000 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 1, \"\n",
      "            \"'min_samples_split': 20, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 64, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 1.000 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 1, \"\n",
      "            \"'min_samples_split': 20, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 128, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 1.000 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 1, \"\n",
      "            \"'min_samples_split': 20, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 256, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 1.000 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 1, \"\n",
      "            \"'min_samples_split': 20, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 512, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.895 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 5, \"\n",
      "            \"'min_samples_split': 2, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 32, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.895 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 5, \"\n",
      "            \"'min_samples_split': 2, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 64, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.947 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 5, \"\n",
      "            \"'min_samples_split': 2, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 128, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.895 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 5, \"\n",
      "            \"'min_samples_split': 2, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 256, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.895 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 5, \"\n",
      "            \"'min_samples_split': 2, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 512, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.895 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 5, \"\n",
      "            \"'min_samples_split': 5, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 32, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.895 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 5, \"\n",
      "            \"'min_samples_split': 5, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 64, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.947 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 5, \"\n",
      "            \"'min_samples_split': 5, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 128, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.895 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 5, \"\n",
      "            \"'min_samples_split': 5, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 256, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.895 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 5, \"\n",
      "            \"'min_samples_split': 5, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 512, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.895 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 5, \"\n",
      "            \"'min_samples_split': 10, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 32, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.895 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 5, \"\n",
      "            \"'min_samples_split': 10, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 64, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.947 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 5, \"\n",
      "            \"'min_samples_split': 10, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 128, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.895 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 5, \"\n",
      "            \"'min_samples_split': 10, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 256, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.895 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 5, \"\n",
      "            \"'min_samples_split': 10, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 512, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.947 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 5, \"\n",
      "            \"'min_samples_split': 20, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 32, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.947 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 5, \"\n",
      "            \"'min_samples_split': 20, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 64, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.947 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 5, \"\n",
      "            \"'min_samples_split': 20, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 128, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.947 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 5, \"\n",
      "            \"'min_samples_split': 20, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 256, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.947 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 5, \"\n",
      "            \"'min_samples_split': 20, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 512, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.842 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 10, \"\n",
      "            \"'min_samples_split': 2, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 32, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.842 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 10, \"\n",
      "            \"'min_samples_split': 2, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 64, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.842 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 10, \"\n",
      "            \"'min_samples_split': 2, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 128, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.947 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 10, \"\n",
      "            \"'min_samples_split': 2, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 256, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.947 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 10, \"\n",
      "            \"'min_samples_split': 2, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 512, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.842 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 10, \"\n",
      "            \"'min_samples_split': 5, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 32, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.842 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 10, \"\n",
      "            \"'min_samples_split': 5, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 64, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.842 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 10, \"\n",
      "            \"'min_samples_split': 5, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 128, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.947 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 10, \"\n",
      "            \"'min_samples_split': 5, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 256, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.947 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 10, \"\n",
      "            \"'min_samples_split': 5, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 512, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.842 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 10, \"\n",
      "            \"'min_samples_split': 10, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 32, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.842 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 10, \"\n",
      "            \"'min_samples_split': 10, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 64, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.842 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 10, \"\n",
      "            \"'min_samples_split': 10, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 128, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.947 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 10, \"\n",
      "            \"'min_samples_split': 10, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 256, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.947 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 10, \"\n",
      "            \"'min_samples_split': 10, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 512, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.842 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 10, \"\n",
      "            \"'min_samples_split': 20, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 32, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.842 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 10, \"\n",
      "            \"'min_samples_split': 20, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 64, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.842 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 10, \"\n",
      "            \"'min_samples_split': 20, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 128, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.947 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 10, \"\n",
      "            \"'min_samples_split': 20, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 256, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.947 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 10, \"\n",
      "            \"'min_samples_split': 20, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 512, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.895 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 20, \"\n",
      "            \"'min_samples_split': 2, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 32, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.947 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 20, \"\n",
      "            \"'min_samples_split': 2, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 64, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.947 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 20, \"\n",
      "            \"'min_samples_split': 2, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 128, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.947 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 20, \"\n",
      "            \"'min_samples_split': 2, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 256, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.947 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 20, \"\n",
      "            \"'min_samples_split': 2, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 512, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.895 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 20, \"\n",
      "            \"'min_samples_split': 5, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 32, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.947 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 20, \"\n",
      "            \"'min_samples_split': 5, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 64, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.947 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 20, \"\n",
      "            \"'min_samples_split': 5, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 128, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.947 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 20, \"\n",
      "            \"'min_samples_split': 5, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 256, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.947 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 20, \"\n",
      "            \"'min_samples_split': 5, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 512, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.895 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 20, \"\n",
      "            \"'min_samples_split': 10, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 32, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.947 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 20, \"\n",
      "            \"'min_samples_split': 10, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 64, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.947 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 20, \"\n",
      "            \"'min_samples_split': 10, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 128, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.947 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 20, \"\n",
      "            \"'min_samples_split': 10, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 256, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.947 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 20, \"\n",
      "            \"'min_samples_split': 10, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 512, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.895 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 20, \"\n",
      "            \"'min_samples_split': 20, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 32, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.947 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 20, \"\n",
      "            \"'min_samples_split': 20, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 64, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.947 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 20, \"\n",
      "            \"'min_samples_split': 20, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 128, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.947 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 20, \"\n",
      "            \"'min_samples_split': 20, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 256, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] 0.947 for '\n",
      "            \"{'bootstrap': True, 'class_weight': None, 'criterion': 'gini', \"\n",
      "            \"'max_depth': None, 'max_features': 'auto', 'max_leaf_nodes': \"\n",
      "            \"None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 20, \"\n",
      "            \"'min_samples_split': 20, 'min_weight_fraction_leaf': 0.0, \"\n",
      "            \"'n_estimators': 512, 'n_jobs': 8, 'oob_score': False, 'verbose': \"\n",
      "            \"0, 'warm_start': False}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] accuracy on '\n",
      "            'train data: 0.9772727272727273\\n'\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] Hamming score '\n",
      "            'on test data: 0.8846153846153846\\n'\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] accuracy on '\n",
      "            'test data: 0.8846153846153846\\n'\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [classifier] The '\n",
      "            \"classification_report on dataset: {'Language_Both': {'precision': \"\n",
      "            \"1.0, 'recall': 0.5, 'f1-score': 0.6666666666666666}, \"\n",
      "            \"'Language_English': {'precision': 0.8, 'recall': 1.0, 'f1-score': \"\n",
      "            \"0.888888888888889}, 'Language_German': {'precision': \"\n",
      "            \"0.8181818181818182, 'recall': 1.0, 'f1-score': 0.9}, \"\n",
      "            \"'Language_Other': {'precision': 1.0, 'recall': 1.0, 'f1-score': \"\n",
      "            \"1.0}, 'hamming_score': 0.8846153846153846, 'micro': {'precision': \"\n",
      "            \"0.8846153846153846, 'recall': 0.8846153846153846, 'f1-score': \"\n",
      "            \"0.8846153846153846}, 'macro': {'precision': 0.9045454545454545, \"\n",
      "            \"'recall': 0.875, 'f1-score': 0.8638888888888889}, 'weighted': \"\n",
      "            \"{'precision': 0.9062937062937062, 'recall': 0.8846153846153846, \"\n",
      "            \"'f1-score': 0.8713675213675214}}\\n\"\n",
      "            '[2020-03-17 15:41:36 +0000] [8] [INFO] [util] Generate meta-data: '\n",
      "            \"{'jobId': \"\n",
      "            \"'bdp-c70d39c1-d258-4752-bca8-18051c5ace9e-2020-03-17t1537z100508', \"\n",
      "            \"'zoneId': '12e18df8-1823-423e-8f44-0464c96f2acb', 'language': \"\n",
      "            \"'none', 'encoding': 'utf-8', 'modelType': 'tfidf+lc', \"\n",
      "            \"'serializionProtocol': '1', 'mlf_custom': 'tfidf+lc', \"\n",
      "            \"'min_occurrences_of_terms': '10', 'max_num_of_terms_to_consider': \"\n",
      "            \"'100000', 'ngram_max': '2', 'accuracy': '0.8846153846153846', \"\n",
      "            \"'precision': '0.9062937062937062', 'recall': \"\n",
      "            \"'0.8846153846153846', 'f1': '0.8713675213675214', 'bootstrap': \"\n",
      "            \"'True', 'class_weight': 'None', 'criterion': 'gini', 'max_depth': \"\n",
      "            \"'None', 'max_features': 'auto', 'max_leaf_nodes': 'None', \"\n",
      "            \"'min_impurity_decrease': '0.0', 'min_samples_leaf': '1', \"\n",
      "            \"'min_samples_split': '20', 'min_weight_fraction_leaf': '0.0', \"\n",
      "            \"'n_estimators': '32', 'n_jobs': '8', 'oob_score': 'False', \"\n",
      "            \"'verbose': '0', 'warm_start': 'False'}\\n\"\n",
      "            '[2020-03-17 15:44:43 +0000] [8] [INFO] '\n",
      "            '[sap.mlf.serving.python.exporter.mlf_exporter] Zipped saved model '\n",
      "            '/data/jobs/bdp-c70d39c1-d258-4752-bca8-18051c5ace9e-2020-03-17t1537z100508-30fbb460-6865-11ea-b22d-eeee0a92ff28/dist/bdp-c70d39c1-d258-4752-bca8-18051c5ace9e/model_1.zip\\n'\n",
      "            '[2020-03-17 15:45:36 +0000] [8] [INFO] [util] Uploaded model '\n",
      "            'bdp-c70d39c1-d258-4752-bca8-18051c5ace9e ver. 1 to the '\n",
      "            'repository.\\n'\n",
      "            '\\n'\n",
      "            '------\\n',\n",
      " 'finishedAt': '2020-03-17 15:45:37',\n",
      " 'modelName': 'dc_example_dataset',\n",
      " 'modelVersion': '1',\n",
      " 'precision': 0.9062937062937062,\n",
      " 'recall': 0.8846153846153846,\n",
      " 'startedAt': '2020-03-17 15:37:23',\n",
      " 'status': 'SUCCEEDED'}\n",
      "Model training finished with status: SUCCEEDED\n",
      "Trained model: dc_example_dataset\n",
      "Trained model version: 1\n"
     ]
    }
   ],
   "source": [
    "# Train the model\n",
    "print(\"Start training job from model with modelName {}\".format(model_name))\n",
    "response = my_dc_client.train_model(model_name, training_dataset_id)\n",
    "pprint(response)\n",
    "print(\"Model training finished with status: {}\".format(response.get(\"status\")))\n",
    "if response.get(\"status\") == \"SUCCEEDED\":\n",
    "    model_version = response.get(\"modelVersion\")\n",
    "    print(\"Trained model: {}\".format(model_name))\n",
    "    print(\"Trained model version: {}\".format(model_version))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-09-04T10:01:14.801991Z",
     "start_time": "2019-09-04T10:01:14.229766Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'accuracy': 0.8846153846153846,\n",
      " 'datasetId': '1b6656f3-1774-41ab-bc27-07bd87876052',\n",
      " 'finishedAt': '2020-03-17 15:45:37',\n",
      " 'modelName': 'dc_example_dataset',\n",
      " 'modelVersion': '1',\n",
      " 'precision': 0.9062937062937062,\n",
      " 'recall': 0.8846153846153846,\n",
      " 'startedAt': '2020-03-17 15:37:23',\n",
      " 'status': 'SUCCEEDED'}\n"
     ]
    }
   ],
   "source": [
    "# Check training statistics\n",
    "reponse = my_dc_client.get_trained_model_info(model_name, model_version)\n",
    "training_details = response.pop(\"details\")\n",
    "pprint(response)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Deployment"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'deploymentId': '6585dac8-c969-4259-8d95-6342aeb10c8c',\n",
      " 'finishedAt': '2020-03-17 15:50:02.063000',\n",
      " 'modelName': 'dc_example_dataset',\n",
      " 'modelVersion': '1',\n",
      " 'startedAt': '2020-03-17 15:46:55.045000',\n",
      " 'status': 'SUCCEEDED'}\n"
     ]
    }
   ],
   "source": [
    "# Deploy model\n",
    "response = my_dc_client.deploy_model(model_name, model_version)\n",
    "pprint(response)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Classification\n",
    "\n",
    "This runs inference on all documents in the test set (stratification is done inside DC service and reproduced here).  \n",
    "We are working on exposing the stratification results so that this cell can be shortend."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'documentId': '6f74fb45-c243-482f-ba01-014234db18a3',\n",
      " 'predictions': [{'characteristic': 'Language',\n",
      "                  'results': [{'label': 'Both', 'score': 0.11972336063156244},\n",
      "                              {'label': 'English', 'score': 0.0078125},\n",
      "                              {'label': 'German', 'score': 0.7748078893684377},\n",
      "                              {'label': 'Other', 'score': 0.09765625}]}],\n",
      " 'processedTime': '2020-03-17T15:50:26.461000',\n",
      " 'status': 'SUCCEEDED'}\n",
      "{'documentId': '6e06e197-3496-49bd-b8c5-1845fdc90cae',\n",
      " 'predictions': [{'characteristic': 'Language',\n",
      "                  'results': [{'label': 'Both', 'score': 0.28932334741180255},\n",
      "                              {'label': 'English', 'score': 0.505862210850733},\n",
      "                              {'label': 'German', 'score': 0.03466208735909823},\n",
      "                              {'label': 'Other',\n",
      "                               'score': 0.17015235437836637}]}],\n",
      " 'processedTime': '2020-03-17T15:50:26.319000',\n",
      " 'status': 'SUCCEEDED'}\n",
      "{'documentId': '29941b55-7a77-4b51-ba8b-3550361103bb',\n",
      " 'predictions': [{'characteristic': 'Language',\n",
      "                  'results': [{'label': 'Both', 'score': 0.5971805085518136},\n",
      "                              {'label': 'English',\n",
      "                               'score': 0.14738568215892053},\n",
      "                              {'label': 'German', 'score': 0.16272547595593262},\n",
      "                              {'label': 'Other',\n",
      "                               'score': 0.09270833333333332}]}],\n",
      " 'processedTime': '2020-03-17T15:50:26.912000',\n",
      " 'status': 'SUCCEEDED'}\n",
      "{'documentId': '4f3933b0-b892-464e-bc50-5ee5362030f6',\n",
      " 'predictions': [{'characteristic': 'Language',\n",
      "                  'results': [{'label': 'Both', 'score': 0.01310185185185185},\n",
      "                              {'label': 'English',\n",
      "                               'score': 0.00837962962962963},\n",
      "                              {'label': 'German', 'score': 0.04416666666666667},\n",
      "                              {'label': 'Other',\n",
      "                               'score': 0.9343518518518519}]}],\n",
      " 'processedTime': '2020-03-17T15:50:26.604000',\n",
      " 'status': 'SUCCEEDED'}\n",
      "{'documentId': 'b49b1dc9-3e42-4a54-ad17-0378c8ae35db',\n",
      " 'predictions': [{'characteristic': 'Language',\n",
      "                  'results': [{'label': 'Both', 'score': 0.005952380952380952},\n",
      "                              {'label': 'English',\n",
      "                               'score': 0.008928571428571428},\n",
      "                              {'label': 'German', 'score': 0.0},\n",
      "                              {'label': 'Other',\n",
      "                               'score': 0.9851190476190477}]}],\n",
      " 'processedTime': '2020-03-17T15:50:26.420000',\n",
      " 'status': 'SUCCEEDED'}\n",
      "{'documentId': '61307b89-00a0-4e05-a2ef-452a6a5ee16c',\n",
      " 'predictions': [{'characteristic': 'Language',\n",
      "                  'results': [{'label': 'Both', 'score': 0.19463158125759156},\n",
      "                              {'label': 'English', 'score': 0.7664739294950967},\n",
      "                              {'label': 'German',\n",
      "                               'score': 0.0026041666666666665},\n",
      "                              {'label': 'Other',\n",
      "                               'score': 0.036290322580645164}]}],\n",
      " 'processedTime': '2020-03-17T15:50:33.577000',\n",
      " 'status': 'SUCCEEDED'}\n",
      "{'documentId': 'fd300097-518a-4f44-8aec-a9eb42289cef',\n",
      " 'predictions': [{'characteristic': 'Language',\n",
      "                  'results': [{'label': 'Both', 'score': 0.4704353532067348},\n",
      "                              {'label': 'English',\n",
      "                               'score': 0.03211805555555555},\n",
      "                              {'label': 'German', 'score': 0.4974465912377097},\n",
      "                              {'label': 'Other', 'score': 0.0}]}],\n",
      " 'processedTime': '2020-03-17T15:50:33.884000',\n",
      " 'status': 'SUCCEEDED'}\n",
      "{'documentId': '2c857f0d-8cc5-447d-b3b6-105516db2ea3',\n",
      " 'predictions': [{'characteristic': 'Language',\n",
      "                  'results': [{'label': 'Both', 'score': 0.17104341860750663},\n",
      "                              {'label': 'English',\n",
      "                               'score': 0.0011574074074074073},\n",
      "                              {'label': 'German', 'score': 0.7244271120192741},\n",
      "                              {'label': 'Other',\n",
      "                               'score': 0.10337206196581196}]}],\n",
      " 'processedTime': '2020-03-17T15:50:34.342000',\n",
      " 'status': 'SUCCEEDED'}\n",
      "{'documentId': 'e96cf3ff-eeb6-4581-a464-72f5e8ae2f6d',\n",
      " 'predictions': [{'characteristic': 'Language',\n",
      "                  'results': [{'label': 'Both', 'score': 0.11213224291736434},\n",
      "                              {'label': 'English', 'score': 0.883310465415969},\n",
      "                              {'label': 'German', 'score': 0.001953125},\n",
      "                              {'label': 'Other',\n",
      "                               'score': 0.0026041666666666665}]}],\n",
      " 'processedTime': '2020-03-17T15:50:34.704000',\n",
      " 'status': 'SUCCEEDED'}\n",
      "{'documentId': 'a7eccfde-3882-4f6c-ba6f-9160a7c7547c',\n",
      " 'predictions': [{'characteristic': 'Language',\n",
      "                  'results': [{'label': 'Both', 'score': 0.319119649685475},\n",
      "                              {'label': 'English',\n",
      "                               'score': 0.15923447924437098},\n",
      "                              {'label': 'German', 'score': 0.4079751848518832},\n",
      "                              {'label': 'Other',\n",
      "                               'score': 0.1136706862182708}]}],\n",
      " 'processedTime': '2020-03-17T15:50:35.079000',\n",
      " 'status': 'SUCCEEDED'}\n",
      "{'documentId': 'b2c37ced-045c-4ec5-85b5-dafe0b9fa677',\n",
      " 'predictions': [{'characteristic': 'Language',\n",
      "                  'results': [{'label': 'Both', 'score': 0.3320472692898601},\n",
      "                              {'label': 'English', 'score': 0.6168677221391566},\n",
      "                              {'label': 'German',\n",
      "                               'score': 0.023047504025764894},\n",
      "                              {'label': 'Other',\n",
      "                               'score': 0.02803750454521843}]}],\n",
      " 'processedTime': '2020-03-17T15:50:35.695000',\n",
      " 'status': 'SUCCEEDED'}\n",
      "{'documentId': '941f16ee-c233-47d7-af8d-df888dbd4c8c',\n",
      " 'predictions': [{'characteristic': 'Language',\n",
      "                  'results': [{'label': 'Both', 'score': 0.005952380952380952},\n",
      "                              {'label': 'English',\n",
      "                               'score': 0.008928571428571428},\n",
      "                              {'label': 'German',\n",
      "                               'score': 0.004166666666666667},\n",
      "                              {'label': 'Other', 'score': 0.980952380952381}]}],\n",
      " 'processedTime': '2020-03-17T15:50:35.967000',\n",
      " 'status': 'SUCCEEDED'}\n",
      "{'documentId': '73136685-ef9d-4c1d-99e0-bfac4bff75fa',\n",
      " 'predictions': [{'characteristic': 'Language',\n",
      "                  'results': [{'label': 'Both', 'score': 0.6492727198566964},\n",
      "                              {'label': 'English',\n",
      "                               'score': 0.12195298985122824},\n",
      "                              {'label': 'German', 'score': 0.19752429029207533},\n",
      "                              {'label': 'Other', 'score': 0.03125}]}],\n",
      " 'processedTime': '2020-03-17T15:50:36.618000',\n",
      " 'status': 'SUCCEEDED'}\n",
      "{'documentId': '0c7853bc-e2fb-44c0-a5fe-eb6f4f4dbb39',\n",
      " 'predictions': [{'characteristic': 'Language',\n",
      "                  'results': [{'label': 'Both', 'score': 0.031018518518518518},\n",
      "                              {'label': 'English',\n",
      "                               'score': 0.03230820105820106},\n",
      "                              {'label': 'German', 'score': 0.03720238095238095},\n",
      "                              {'label': 'Other',\n",
      "                               'score': 0.8994708994708995}]}],\n",
      " 'processedTime': '2020-03-17T15:50:36.898000',\n",
      " 'status': 'SUCCEEDED'}\n",
      "{'documentId': '29eea36e-d23a-47c8-99e5-c2ec0e7bb5c7',\n",
      " 'predictions': [{'characteristic': 'Language',\n",
      "                  'results': [{'label': 'Both', 'score': 0.03631114130434782},\n",
      "                              {'label': 'English',\n",
      "                               'score': 0.09488224637681159},\n",
      "                              {'label': 'German',\n",
      "                               'score': 0.019791666666666666},\n",
      "                              {'label': 'Other', 'score': 0.849014945652174}]}],\n",
      " 'processedTime': '2020-03-17T15:50:37.202000',\n",
      " 'status': 'SUCCEEDED'}\n",
      "{'documentId': '9bb3fb9d-77f1-4696-af9a-e24980554856',\n",
      " 'predictions': [{'characteristic': 'Language',\n",
      "                  'results': [{'label': 'Both', 'score': 0.1716306595690146},\n",
      "                              {'label': 'English', 'score': 0.0},\n",
      "                              {'label': 'German', 'score': 0.7998236673540625},\n",
      "                              {'label': 'Other',\n",
      "                               'score': 0.028545673076923076}]}],\n",
      " 'processedTime': '2020-03-17T15:50:37.478000',\n",
      " 'status': 'SUCCEEDED'}\n",
      "{'documentId': '186cb39d-39cf-4485-b90e-1dc733a2b1fc',\n",
      " 'predictions': [{'characteristic': 'Language',\n",
      "                  'results': [{'label': 'Both', 'score': 0.1327699458002126},\n",
      "                              {'label': 'English', 'score': 0.0},\n",
      "                              {'label': 'German', 'score': 0.8672300541997876},\n",
      "                              {'label': 'Other', 'score': 0.0}]}],\n",
      " 'processedTime': '2020-03-17T15:50:37.817000',\n",
      " 'status': 'SUCCEEDED'}\n",
      "{'documentId': '755c955c-31b2-4090-85d3-396fc7b6eec8',\n",
      " 'predictions': [{'characteristic': 'Language',\n",
      "                  'results': [{'label': 'Both', 'score': 0.22850969569790466},\n",
      "                              {'label': 'English', 'score': 0.5251361376354287},\n",
      "                              {'label': 'German', 'score': 0.0},\n",
      "                              {'label': 'Other',\n",
      "                               'score': 0.24635416666666665}]}],\n",
      " 'processedTime': '2020-03-17T15:50:37.850000',\n",
      " 'status': 'SUCCEEDED'}\n",
      "{'documentId': 'c4746eea-86dc-4ca9-a36e-86de0fca8bc6',\n",
      " 'predictions': [{'characteristic': 'Language',\n",
      "                  'results': [{'label': 'Both', 'score': 0.5629550019096201},\n",
      "                              {'label': 'English', 'score': 0.0880396483305966},\n",
      "                              {'label': 'German', 'score': 0.2368608391777728},\n",
      "                              {'label': 'Other',\n",
      "                               'score': 0.11214451058201058}]}],\n",
      " 'processedTime': '2020-03-17T15:50:38.243000',\n",
      " 'status': 'SUCCEEDED'}\n",
      "{'documentId': 'ef63064b-5ca1-4cd3-ad14-da5c5dfeeb30',\n",
      " 'predictions': [{'characteristic': 'Language',\n",
      "                  'results': [{'label': 'Both', 'score': 0.12882674190275428},\n",
      "                              {'label': 'English', 'score': 0.0},\n",
      "                              {'label': 'German', 'score': 0.7969545080972458},\n",
      "                              {'label': 'Other', 'score': 0.07421875}]}],\n",
      " 'processedTime': '2020-03-17T15:50:38.252000',\n",
      " 'status': 'SUCCEEDED'}\n",
      "{'documentId': 'f6fb4f62-f598-4931-8527-3e4901ad4044',\n",
      " 'predictions': [{'characteristic': 'Language',\n",
      "                  'results': [{'label': 'Both', 'score': 0.025275297619047617},\n",
      "                              {'label': 'English',\n",
      "                               'score': 0.020491071428571428},\n",
      "                              {'label': 'German', 'score': 0.01578125},\n",
      "                              {'label': 'Other', 'score': 0.938452380952381}]}],\n",
      " 'processedTime': '2020-03-17T15:50:38.583000',\n",
      " 'status': 'SUCCEEDED'}\n",
      "{'documentId': '30632266-ea5e-4120-a90c-5fa6101d36de',\n",
      " 'predictions': [{'characteristic': 'Language',\n",
      "                  'results': [{'label': 'Both', 'score': 0.04293218908708038},\n",
      "                              {'label': 'English',\n",
      "                               'score': 0.07926831723027376},\n",
      "                              {'label': 'German', 'score': 0.09291529248730335},\n",
      "                              {'label': 'Other',\n",
      "                               'score': 0.7848842011953425}]}],\n",
      " 'processedTime': '2020-03-17T15:50:38.911000',\n",
      " 'status': 'SUCCEEDED'}\n",
      "{'documentId': '29bacb16-186e-42b5-84e2-74167a1517c7',\n",
      " 'predictions': [{'characteristic': 'Language',\n",
      "                  'results': [{'label': 'Both', 'score': 0.1634994133038506},\n",
      "                              {'label': 'English', 'score': 0.0},\n",
      "                              {'label': 'German', 'score': 0.8331523724104352},\n",
      "                              {'label': 'Other',\n",
      "                               'score': 0.0033482142857142855}]}],\n",
      " 'processedTime': '2020-03-17T15:50:39.591000',\n",
      " 'status': 'SUCCEEDED'}\n",
      "{'documentId': '763946a9-1645-4f2a-8cd2-3246f529019c',\n",
      " 'predictions': [{'characteristic': 'Language',\n",
      "                  'results': [{'label': 'Both', 'score': 0.16511075007168755},\n",
      "                              {'label': 'English', 'score': 0.0},\n",
      "                              {'label': 'German', 'score': 0.8036392499283125},\n",
      "                              {'label': 'Other', 'score': 0.03125}]}],\n",
      " 'processedTime': '2020-03-17T15:50:39.841000',\n",
      " 'status': 'SUCCEEDED'}\n",
      "{'documentId': 'afb222e2-673e-4f5f-9c46-c3d5840dc847',\n",
      " 'predictions': [{'characteristic': 'Language',\n",
      "                  'results': [{'label': 'Both', 'score': 0.13889542605253108},\n",
      "                              {'label': 'English', 'score': 0.0},\n",
      "                              {'label': 'German', 'score': 0.8454795739474691},\n",
      "                              {'label': 'Other', 'score': 0.015625}]}],\n",
      " 'processedTime': '2020-03-17T15:50:39.923000',\n",
      " 'status': 'SUCCEEDED'}\n",
      "{'documentId': '25edd49e-3bc5-4243-8cc4-e95bac196d0f',\n",
      " 'predictions': [{'characteristic': 'Language',\n",
      "                  'results': [{'label': 'Both', 'score': 0.1458659360751247},\n",
      "                              {'label': 'English', 'score': 0.0},\n",
      "                              {'label': 'German', 'score': 0.846879599639161},\n",
      "                              {'label': 'Other',\n",
      "                               'score': 0.007254464285714286}]}],\n",
      " 'processedTime': '2020-03-17T15:50:40.201000',\n",
      " 'status': 'SUCCEEDED'}\n"
     ]
    }
   ],
   "source": [
    "# Test usage of the model by classifying a few documents and collecting results and ground truth\n",
    "import binascii\n",
    "import time\n",
    "import json\n",
    "import numpy as np\n",
    "from collections import defaultdict\n",
    "\n",
    "filenames = my_dc_client._find_files(dataset_folder, \"*.PDF\")\n",
    "test_filenames = []\n",
    "for filename in filenames:\n",
    "    # Check whether it is a test document\n",
    "    with open(filename, 'rb') as pdf_file:\n",
    "        is_test_document = (int(str(binascii.crc32(pdf_file.read()))) % 100) in range(90,100)\n",
    "    if is_test_document:\n",
    "        test_filenames.append(filename)\n",
    "\n",
    "# Classify all test documents\n",
    "responses = my_dc_client.classify_documents(test_filenames, model_name, model_version)\n",
    "\n",
    "# Iterate over responses and store results in convenient format\n",
    "test_prediction = defaultdict(lambda : [])\n",
    "test_probability = defaultdict(lambda : defaultdict(lambda : []))\n",
    "test_ground_truth = defaultdict(lambda : [])\n",
    "for response, filename in zip(responses, test_filenames):\n",
    "    pprint(response)\n",
    "    try:\n",
    "        # Parse response from DC service\n",
    "        prediction = response[\"predictions\"]\n",
    "        for element in prediction:\n",
    "            labels = []\n",
    "            scores = []\n",
    "            for subelement in element[\"results\"]:\n",
    "                labels.append(subelement[\"label\"])\n",
    "                scores.append(subelement[\"score\"])\n",
    "                test_probability[element[\"characteristic\"]][subelement[\"label\"]].append(subelement[\"score\"])\n",
    "            test_prediction[element[\"characteristic\"]].append(labels[np.argmax(np.asarray(scores))])\n",
    "        # Collect ground truth of all test documents\n",
    "        with open(filename.replace(\".pdf\", \".json\")) as gt_file:\n",
    "            gt = json.load(gt_file)\n",
    "        for element in gt[\"classification\"]:\n",
    "            test_ground_truth[element[\"characteristic\"]].append(element[\"value\"])\n",
    "    except KeyError:\n",
    "        print(\"Document not used\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Ground truth for characteristic 'Language': 'German'\n",
      "Model predictions:\n",
      "{'documentId': '6f74fb45-c243-482f-ba01-014234db18a3',\n",
      " 'predictions': [{'characteristic': 'Language',\n",
      "                  'results': [{'label': 'Both', 'score': 0.11972336063156244},\n",
      "                              {'label': 'English', 'score': 0.0078125},\n",
      "                              {'label': 'German', 'score': 0.7748078893684377},\n",
      "                              {'label': 'Other', 'score': 0.09765625}]}],\n",
      " 'processedTime': '2020-03-17T15:50:26.461000',\n",
      " 'status': 'SUCCEEDED'}\n"
     ]
    }
   ],
   "source": [
    "# display the ground truth and classification result for a certain document with index idx\n",
    "idx = 0\n",
    "\n",
    "for i in range(nrCharacteristics):\n",
    "    characteristic =dataset_stats[\"groundTruths\"][i][\"characteristic\"]\n",
    "    print(\"Ground truth for characteristic '{}'\".format(str(characteristic)) + \": '{}'\".format(test_ground_truth[str(characteristic)][idx]))\n",
    "\n",
    "print(\"Model predictions:\")\n",
    "pprint(responses[idx])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Confusion Matrix\n",
    "\n",
    "The confusion matrix is a way to visualize the results of a classiifcation task."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1008x2016 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "from sklearn.metrics import confusion_matrix\n",
    "\n",
    "font = {'size'   : 22}\n",
    "plt.rc('font', **font)\n",
    "\n",
    "def plot_confusion_matrix(ax, char, y_true, y_pred, classes,\n",
    "                          normalize=True,\n",
    "                          title=None,\n",
    "                          cmap=plt.cm.Blues):\n",
    "    \"\"\"\n",
    "    This function prints and plots the confusion matrix.\n",
    "    Normalization can be applied by setting `normalize=True`.\n",
    "    \"\"\"\n",
    "    if not title:\n",
    "        if normalize:\n",
    "            title = \"{}: Normalized confusion matrix\".format(char)\n",
    "        else:\n",
    "            title = \"{}: Confusion matrix without normalization\".format(char)\n",
    "\n",
    "    # Compute confusion matrix\n",
    "    cm = confusion_matrix(y_true, y_pred)\n",
    "    if normalize:\n",
    "        cm = cm.astype('float') / cm.sum(axis=1)[:, np.newaxis]\n",
    "    im = ax.imshow(cm, interpolation='nearest', cmap=cmap)\n",
    "\n",
    "        # We want to show all ticks...\n",
    "    ax.set(xticks=np.arange(cm.shape[1]),\n",
    "           yticks=np.arange(cm.shape[0]),\n",
    "           # ... and label them with the respective list entries\n",
    "           xticklabels=classes, yticklabels=classes,\n",
    "           title=title,\n",
    "           ylabel=\"True label\",\n",
    "           xlabel=\"Predicted label\",\n",
    "           xlim=(-0.5,len(classes)-0.5),\n",
    "           ylim=(-0.5,len(classes)-0.5))\n",
    "\n",
    "\n",
    "    # Rotate the tick labels and set their alignment.\n",
    "    plt.setp(ax.get_xticklabels(), rotation=45, ha='right',\n",
    "             rotation_mode='anchor')\n",
    "\n",
    "    # Loop over data dimensions and create text annotations.\n",
    "    fmt = '.2f' if normalize else 'd'\n",
    "    thresh = cm.max() / 2.\n",
    "    for i in range(cm.shape[0]):\n",
    "        for j in range(cm.shape[1]):\n",
    "            ax.text(j, i, format(cm[i, j], fmt),\n",
    "                    ha='center', va='center', color='white' if cm[i, j] > thresh else 'black')\n",
    "    fig.show()\n",
    "\n",
    "fig, ax = plt.subplots(len(test_ground_truth.keys()), 1, figsize=(14,28))\n",
    "if len(test_ground_truth.keys())==1:\n",
    "    ax = np.array((ax,)) \n",
    "for idx, characteristic in enumerate(test_ground_truth.keys()):\n",
    "    plot_confusion_matrix(ax[idx], characteristic,\n",
    "                          test_ground_truth[characteristic], \n",
    "                          test_prediction[characteristic], \n",
    "                          np.unique(np.asarray(test_ground_truth[characteristic])), \n",
    "                          normalize=False)\n",
    "fig.subplots_adjust(hspace=0.5)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Precision Recall curves\n",
    "\n",
    "Note that a Precision recall curve is only well defined for binary classification tasks.  \n",
    "For the multi-class case each class is treated independently based on the confidence scroes provided by the service."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 960x1920 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "## Visualize PR curve for each characteristic (NOTE this as a bit boring in this example, create a more challenging dataset for algorithm?)\n",
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "from sklearn.metrics import precision_recall_curve\n",
    "\n",
    "def plot_f_score(ax):\n",
    "    f_scores = np.linspace(0.2, 0.8, num=4)\n",
    "    for f_score in f_scores:\n",
    "        x = np.linspace(0.01, 1)\n",
    "        y = f_score * x / (2 * x - f_score)\n",
    "        l, = ax.plot(x[y >= 0], y[y >= 0], color='gray', alpha=0.2)\n",
    "        ax.annotate('f1={0:0.1f}'.format(f_score), xy=(0.9, y[45] + 0.02))\n",
    "\n",
    "fig, ax = plt.subplots(len(test_ground_truth.keys()), 1, figsize=(12, 24), dpi=80, facecolor='w', edgecolor='k')\n",
    "if len(test_ground_truth.keys())==1:\n",
    "    ax = np.array((ax,)) \n",
    "\n",
    "for idx, characteristic in enumerate(test_ground_truth.keys()):\n",
    "    for label in np.unique(np.asarray(test_ground_truth[characteristic])):\n",
    "        gt = [subelement == label for subelement in test_ground_truth[characteristic]]\n",
    "        prediction = test_probability[characteristic][label]\n",
    "        precision, recall, thresholds = precision_recall_curve(gt, prediction)\n",
    "        ax[idx].plot(recall, precision, label=label)\n",
    "    ax[idx].set_xlabel('Recall')\n",
    "    ax[idx].set_ylabel('Precision')\n",
    "    ax[idx].set_xlim(-0.1,1.1)\n",
    "    ax[idx].set_ylim(-0.1,1.1)\n",
    "    ax[idx].set_title('{}: Precision-Recall curves'.format(characteristic))   \n",
    "    ax[idx].spines[\"top\"].set_visible(False)\n",
    "    ax[idx].spines[\"right\"].set_visible(False)\n",
    "    ax[idx].get_xaxis().tick_bottom()\n",
    "    ax[idx].get_yaxis().tick_left()\n",
    "    ax[idx].legend()\n",
    "    ax[idx].grid()\n",
    "    plot_f_score(ax[idx])\n",
    "\n",
    "fig.show()"
   ]
  },
  {
   "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"
  },
  "toc": {
   "base_numbering": 1.0,
   "nav_menu": {},
   "number_sections": true,
   "sideBar": true,
   "skip_h1_title": false,
   "title_cell": "Table of Contents",
   "title_sidebar": "Contents",
   "toc_cell": false,
   "toc_position": {},
   "toc_section_display": true,
   "toc_window_display": false
  },
  "varInspector": {
   "cols": {
    "lenName": 16.0,
    "lenType": 16.0,
    "lenVar": 40.0
   },
   "kernels_config": {
    "python": {
     "delete_cmd_postfix": "",
     "delete_cmd_prefix": "del ",
     "library": "var_list.py",
     "varRefreshCmd": "print(var_dic_list())"
    },
    "r": {
     "delete_cmd_postfix": ") ",
     "delete_cmd_prefix": "rm(",
     "library": "var_list.r",
     "varRefreshCmd": "cat(var_dic_list()) "
    }
   },
   "types_to_exclude": [
    "module",
    "function",
    "builtin_function_or_method",
    "instance",
    "_Feature"
   ],
   "window_display": false
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}