examples/pzmmModelImportExample.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "Collapsed": "false"
   },
   "source": [
    "Copyright © 2021, SAS Institute Inc., Cary, NC, USA.  All Rights Reserved.\n",
    "SPDX-License-Identifier: Apache-2.0"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "Collapsed": "false"
   },
   "source": [
    "# HMEQ Dataset : Build and Import Trained Models into SAS Model Manager\n",
    "\n",
    "This notebook provides an example of how to build and train a Python model and then import the model into SAS Model Manager using the fleet maintenance data set. Lines of code that must be modified by the user, such as directory paths are noted with the comment \"_Changes required by user._\".\n",
    "\n",
    "_**Note:** If you download only this notebook and not the rest of the repository, you must also download the hmeq.csv file from the data folder in the examples directory. These files are used when executing this notebook example._\n",
    "\n",
    "Here are the steps shown in this notebook:\n",
    "\n",
    "1. Import and review data and preprocess for model training.\n",
    "2. Build, train, and access a decision tree, random forest, and gradient boosting model.\n",
    "3. Serialize the models into separate pickle files.\n",
    "4. Write the metadata JSON files needed for importing into SAS Model Manager as well as optional files for fit statistics and ROC/Lift charts.\n",
    "4. Write a score code Python file for model scoring.\n",
    "5. Zip the pickle, JSON, and score code files into an archive file.\n",
    "6. Import the ZIP archive file to SAS Model Manager via the Session object and relevant function call."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "Collapsed": "false"
   },
   "source": [
    "### Python Package Imports"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "Collapsed": "false"
   },
   "outputs": [],
   "source": [
    "# Dataframes for data manipulations\n",
    "import pandas as pd\n",
    "pd.options.mode.chained_assignment = None  # default='warn'\n",
    "# Mathematical calculations and array handling\n",
    "import numpy as np\n",
    "\n",
    "# Data partitioning for TRAIN and TEST data sets\n",
    "from sklearn.model_selection import train_test_split\n",
    "# Decision tree, random forest, and gradient boosting models\n",
    "from sklearn.tree import DecisionTreeClassifier\n",
    "from sklearn.ensemble import RandomForestClassifier, GradientBoostingClassifier\n",
    "# Model assessments \n",
    "from sklearn.metrics import classification_report, confusion_matrix\n",
    "\n",
    "# Embedded plotting\n",
    "import matplotlib.pyplot as plt \n",
    "plt.rc(\"font\", size=14)\n",
    "\n",
    "# Pathing support\n",
    "from pathlib import Path\n",
    "\n",
    "# sasctl interface for importing models\n",
    "import sasctl.pzmm as pzmm\n",
    "from sasctl import Session"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "Collapsed": "false"
   },
   "source": [
    "### Import and Review Data Set"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "Collapsed": "false"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(5960, 13)"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "hmeqData = pd.read_csv('data/hmeq.csv',sep= ',')\n",
    "hmeqData.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "Collapsed": "false"
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>BAD</th>\n",
       "      <th>LOAN</th>\n",
       "      <th>MORTDUE</th>\n",
       "      <th>VALUE</th>\n",
       "      <th>REASON</th>\n",
       "      <th>JOB</th>\n",
       "      <th>YOJ</th>\n",
       "      <th>DEROG</th>\n",
       "      <th>DELINQ</th>\n",
       "      <th>CLAGE</th>\n",
       "      <th>NINQ</th>\n",
       "      <th>CLNO</th>\n",
       "      <th>DEBTINC</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>26800</td>\n",
       "      <td>46236.0</td>\n",
       "      <td>62711.0</td>\n",
       "      <td>DebtCon</td>\n",
       "      <td>Office</td>\n",
       "      <td>17.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>175.075058</td>\n",
       "      <td>1.0</td>\n",
       "      <td>22.0</td>\n",
       "      <td>33.059934</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0</td>\n",
       "      <td>26900</td>\n",
       "      <td>74982.0</td>\n",
       "      <td>126972.0</td>\n",
       "      <td>DebtCon</td>\n",
       "      <td>Office</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>315.818911</td>\n",
       "      <td>0.0</td>\n",
       "      <td>23.0</td>\n",
       "      <td>38.325990</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0</td>\n",
       "      <td>26900</td>\n",
       "      <td>67144.0</td>\n",
       "      <td>92923.0</td>\n",
       "      <td>DebtCon</td>\n",
       "      <td>Other</td>\n",
       "      <td>16.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>89.112173</td>\n",
       "      <td>1.0</td>\n",
       "      <td>17.0</td>\n",
       "      <td>32.791478</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0</td>\n",
       "      <td>26900</td>\n",
       "      <td>45763.0</td>\n",
       "      <td>73797.0</td>\n",
       "      <td>DebtCon</td>\n",
       "      <td>Other</td>\n",
       "      <td>23.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.0</td>\n",
       "      <td>291.591681</td>\n",
       "      <td>1.0</td>\n",
       "      <td>29.0</td>\n",
       "      <td>39.370858</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0</td>\n",
       "      <td>27000</td>\n",
       "      <td>144901.0</td>\n",
       "      <td>178093.0</td>\n",
       "      <td>DebtCon</td>\n",
       "      <td>ProfExe</td>\n",
       "      <td>7.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>331.113972</td>\n",
       "      <td>0.0</td>\n",
       "      <td>34.0</td>\n",
       "      <td>40.566552</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   BAD   LOAN   MORTDUE     VALUE   REASON      JOB   YOJ  DEROG  DELINQ  \\\n",
       "0    0  26800   46236.0   62711.0  DebtCon   Office  17.0    0.0     0.0   \n",
       "1    0  26900   74982.0  126972.0  DebtCon   Office   0.0    0.0     0.0   \n",
       "2    0  26900   67144.0   92923.0  DebtCon    Other  16.0    0.0     0.0   \n",
       "3    0  26900   45763.0   73797.0  DebtCon    Other  23.0    NaN     0.0   \n",
       "4    0  27000  144901.0  178093.0  DebtCon  ProfExe   7.0    0.0     0.0   \n",
       "\n",
       "        CLAGE  NINQ  CLNO    DEBTINC  \n",
       "0  175.075058   1.0  22.0  33.059934  \n",
       "1  315.818911   0.0  23.0  38.325990  \n",
       "2   89.112173   1.0  17.0  32.791478  \n",
       "3  291.591681   1.0  29.0  39.370858  \n",
       "4  331.113972   0.0  34.0  40.566552  "
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "hmeqData.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "Collapsed": "false"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 1080x1080 with 16 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "hmeqData.hist(figsize=(15,15), layout=(4, 4));"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "Collapsed": "false"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Index(['BAD', 'LOAN', 'MORTDUE', 'VALUE', 'REASON', 'JOB', 'YOJ', 'DEROG',\n",
       "       'DELINQ', 'CLAGE', 'NINQ', 'CLNO', 'DEBTINC'],\n",
       "      dtype='object')"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "hmeqData.columns"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "Collapsed": "false"
   },
   "source": [
    "### Preprocess Data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "Collapsed": "false"
   },
   "outputs": [],
   "source": [
    "predictorColumns = ['LOAN', 'MORTDUE', 'VALUE', 'YOJ', 'DEROG', 'DELINQ', 'CLAGE', 'NINQ', 'CLNO', 'DEBTINC']\n",
    "\n",
    "targetColumn = 'BAD'\n",
    "x = hmeqData[predictorColumns]\n",
    "y = hmeqData[targetColumn]\n",
    "\n",
    "xTrain, xTest, yTrain, yTest = train_test_split(x, y, test_size=0.3, random_state=42)\n",
    "\n",
    "# For missing values, impute the data set's mean value\n",
    "xTest.fillna(xTest.mean(), inplace=True)\n",
    "xTrain.fillna(xTrain.mean(), inplace=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "Collapsed": "false"
   },
   "source": [
    "### Create, Train, and Assess Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "Collapsed": "false"
   },
   "outputs": [],
   "source": [
    "treeModel = DecisionTreeClassifier(random_state=42)\n",
    "treeModel = treeModel.fit(xTrain, yTrain)\n",
    "\n",
    "forestModel = RandomForestClassifier(random_state=42)\n",
    "forestModel = forestModel.fit(xTrain, yTrain)\n",
    "\n",
    "gradientModel = GradientBoostingClassifier(random_state=42)\n",
    "gradientModel = gradientModel.fit(xTrain, yTrain)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "Collapsed": "false"
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>DecisionTree</th>\n",
       "      <th>RandomForest</th>\n",
       "      <th>GradientBoosting</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>DEBTINC</th>\n",
       "      <td>36.16%</td>\n",
       "      <td>26.14%</td>\n",
       "      <td>51.18%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>CLAGE</th>\n",
       "      <td>12.6%</td>\n",
       "      <td>10.87%</td>\n",
       "      <td>7.76%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DELINQ</th>\n",
       "      <td>12.18%</td>\n",
       "      <td>9.82%</td>\n",
       "      <td>14.68%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>VALUE</th>\n",
       "      <td>7.16%</td>\n",
       "      <td>9.42%</td>\n",
       "      <td>4.77%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>LOAN</th>\n",
       "      <td>6.97%</td>\n",
       "      <td>8.98%</td>\n",
       "      <td>3.54%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>YOJ</th>\n",
       "      <td>6.75%</td>\n",
       "      <td>6.88%</td>\n",
       "      <td>2.04%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>MORTDUE</th>\n",
       "      <td>6.15%</td>\n",
       "      <td>8.46%</td>\n",
       "      <td>2.03%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>CLNO</th>\n",
       "      <td>5.97%</td>\n",
       "      <td>8.16%</td>\n",
       "      <td>4.2%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DEROG</th>\n",
       "      <td>3.46%</td>\n",
       "      <td>6.47%</td>\n",
       "      <td>7.65%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>NINQ</th>\n",
       "      <td>2.59%</td>\n",
       "      <td>4.8%</td>\n",
       "      <td>2.15%</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        DecisionTree RandomForest GradientBoosting\n",
       "DEBTINC       36.16%       26.14%           51.18%\n",
       "CLAGE          12.6%       10.87%            7.76%\n",
       "DELINQ        12.18%        9.82%           14.68%\n",
       "VALUE          7.16%        9.42%            4.77%\n",
       "LOAN           6.97%        8.98%            3.54%\n",
       "YOJ            6.75%        6.88%            2.04%\n",
       "MORTDUE        6.15%        8.46%            2.03%\n",
       "CLNO           5.97%        8.16%             4.2%\n",
       "DEROG          3.46%        6.47%            7.65%\n",
       "NINQ           2.59%         4.8%            2.15%"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "def sortFeatureImportance(model, xData):\n",
    "    features = {}\n",
    "    for importance, name in sorted(zip(model.feature_importances_, xData.columns), reverse=True):\n",
    "        features[name] = str(np.round(importance*100, 2)) + '%'\n",
    "    return features\n",
    "\n",
    "importances = pd.DataFrame.from_dict(sortFeatureImportance(treeModel, xTrain), orient='index').rename(columns={0: 'DecisionTree'})\n",
    "importances['RandomForest'] = pd.DataFrame.from_dict(sortFeatureImportance(forestModel, xTrain), orient='index')\n",
    "importances['GradientBoosting'] = pd.DataFrame.from_dict(sortFeatureImportance(gradientModel, xTrain), orient='index')\n",
    "importances"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "Collapsed": "false"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[1396   45]\n",
      " [ 218  129]]\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.86      0.97      0.91      1441\n",
      "           1       0.74      0.37      0.50       347\n",
      "\n",
      "    accuracy                           0.85      1788\n",
      "   macro avg       0.80      0.67      0.70      1788\n",
      "weighted avg       0.84      0.85      0.83      1788\n",
      "\n",
      "Decision Tree Model Accuracy = 85.29%\n"
     ]
    }
   ],
   "source": [
    "yTreePredict = treeModel.predict(xTest)\n",
    "yTreeProba = treeModel.predict_proba(xTest)\n",
    "print(confusion_matrix(yTest, yTreePredict))\n",
    "print(classification_report(yTest, yTreePredict))\n",
    "print('Decision Tree Model Accuracy = ' + str(np.round(treeModel.score(xTest, yTest)*100,2)) + '%')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "Collapsed": "false"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[1438    3]\n",
      " [ 218  129]]\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.87      1.00      0.93      1441\n",
      "           1       0.98      0.37      0.54       347\n",
      "\n",
      "    accuracy                           0.88      1788\n",
      "   macro avg       0.92      0.68      0.73      1788\n",
      "weighted avg       0.89      0.88      0.85      1788\n",
      "\n",
      "Random Forest Model Accuracy = 87.64%\n"
     ]
    }
   ],
   "source": [
    "yForestPredict = forestModel.predict(xTest)\n",
    "yForestProba = forestModel.predict_proba(xTest)\n",
    "print(confusion_matrix(yTest, yForestPredict))\n",
    "print(classification_report(yTest, yForestPredict))\n",
    "print('Random Forest Model Accuracy = ' + str(np.round(forestModel.score(xTest, yTest)*100,2)) + '%')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[1440    1]\n",
      " [ 251   96]]\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.85      1.00      0.92      1441\n",
      "           1       0.99      0.28      0.43       347\n",
      "\n",
      "    accuracy                           0.86      1788\n",
      "   macro avg       0.92      0.64      0.68      1788\n",
      "weighted avg       0.88      0.86      0.83      1788\n",
      "\n",
      "Gradient Boosting Model Accuracy = 85.91%\n"
     ]
    }
   ],
   "source": [
    "yGradientPredict = gradientModel.predict(xTest)\n",
    "yGradientProba = gradientModel.predict_proba(xTest)\n",
    "print(confusion_matrix(yTest, yGradientPredict))\n",
    "print(classification_report(yTest, yGradientPredict))\n",
    "print('Gradient Boosting Model Accuracy = ' + str(np.round(gradientModel.score(xTest, yTest)*100,2)) + '%')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "Collapsed": "false"
   },
   "source": [
    "### Register Model in SAS Model Manager with pzmm"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "Collapsed": "false"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model DecisionTreeClassifier was successfully pickled and saved to c:\\Users\\sclind\\Documents\\Python Scripts\\GitHub\\sassoftware\\python-sasctl\\examples\\data\\hmeqModels\\DecisionTreeClassifier\\DecisionTreeClassifier.pickle.\n",
      "Model RandomForest was successfully pickled and saved to c:\\Users\\sclind\\Documents\\Python Scripts\\GitHub\\sassoftware\\python-sasctl\\examples\\data\\hmeqModels\\RandomForest\\RandomForest.pickle.\n",
      "Model GradientBoosting was successfully pickled and saved to c:\\Users\\sclind\\Documents\\Python Scripts\\GitHub\\sassoftware\\python-sasctl\\examples\\data\\hmeqModels\\GradientBoosting\\GradientBoosting.pickle.\n"
     ]
    }
   ],
   "source": [
    "modelPrefix = ['DecisionTreeClassifier', 'RandomForest', 'GradientBoosting']\n",
    "zipFolder = [Path.cwd() / 'data/hmeqModels/DecisionTreeClassifier/',\n",
    "             Path.cwd() / 'data/hmeqModels/RandomForest/',\n",
    "             Path.cwd() / 'data/hmeqModels/GradientBoosting'] # User created directories\n",
    "model = [treeModel, forestModel, gradientModel]\n",
    "\n",
    "for (m, prefix, path) in zip(model, modelPrefix, zipFolder):\n",
    "    pzmm.PickleModel.pickleTrainedModel(_, m, prefix, path)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "Collapsed": "false"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "inputVar.json was successfully written and saved to c:\\Users\\sclind\\Documents\\Python Scripts\\GitHub\\sassoftware\\python-sasctl\\examples\\data\\hmeqModels\\DecisionTreeClassifier\\inputVar.json\n",
      "outputVar.json was successfully written and saved to c:\\Users\\sclind\\Documents\\Python Scripts\\GitHub\\sassoftware\\python-sasctl\\examples\\data\\hmeqModels\\DecisionTreeClassifier\\outputVar.json\n",
      "ModelProperties.json was successfully written and saved to c:\\Users\\sclind\\Documents\\Python Scripts\\GitHub\\sassoftware\\python-sasctl\\examples\\data\\hmeqModels\\DecisionTreeClassifier\\ModelProperties.json\n",
      "fileMetaData.json was successfully written and saved to c:\\Users\\sclind\\Documents\\Python Scripts\\GitHub\\sassoftware\\python-sasctl\\examples\\data\\hmeqModels\\DecisionTreeClassifier\\fileMetaData.json\n",
      "inputVar.json was successfully written and saved to c:\\Users\\sclind\\Documents\\Python Scripts\\GitHub\\sassoftware\\python-sasctl\\examples\\data\\hmeqModels\\RandomForest\\inputVar.json\n",
      "outputVar.json was successfully written and saved to c:\\Users\\sclind\\Documents\\Python Scripts\\GitHub\\sassoftware\\python-sasctl\\examples\\data\\hmeqModels\\RandomForest\\outputVar.json\n",
      "ModelProperties.json was successfully written and saved to c:\\Users\\sclind\\Documents\\Python Scripts\\GitHub\\sassoftware\\python-sasctl\\examples\\data\\hmeqModels\\RandomForest\\ModelProperties.json\n",
      "fileMetaData.json was successfully written and saved to c:\\Users\\sclind\\Documents\\Python Scripts\\GitHub\\sassoftware\\python-sasctl\\examples\\data\\hmeqModels\\RandomForest\\fileMetaData.json\n",
      "inputVar.json was successfully written and saved to c:\\Users\\sclind\\Documents\\Python Scripts\\GitHub\\sassoftware\\python-sasctl\\examples\\data\\hmeqModels\\GradientBoosting\\inputVar.json\n",
      "outputVar.json was successfully written and saved to c:\\Users\\sclind\\Documents\\Python Scripts\\GitHub\\sassoftware\\python-sasctl\\examples\\data\\hmeqModels\\GradientBoosting\\outputVar.json\n",
      "ModelProperties.json was successfully written and saved to c:\\Users\\sclind\\Documents\\Python Scripts\\GitHub\\sassoftware\\python-sasctl\\examples\\data\\hmeqModels\\GradientBoosting\\ModelProperties.json\n",
      "fileMetaData.json was successfully written and saved to c:\\Users\\sclind\\Documents\\Python Scripts\\GitHub\\sassoftware\\python-sasctl\\examples\\data\\hmeqModels\\GradientBoosting\\fileMetaData.json\n"
     ]
    }
   ],
   "source": [
    "def writeJSONFiles(data, predict, target, zipFolder, modelPrefix):\n",
    "    J = pzmm.JSONFiles()\n",
    "    \n",
    "    # Write input variable mapping to a json file\n",
    "    J.writeVarJSON(data[predict], isInput=True, jPath=zipFolder)\n",
    "    \n",
    "    # Set output variables and assign an event threshold, then write output variable mapping\n",
    "    outputVar = pd.DataFrame(columns=['EM_EVENTPROBABILITY', 'EM_CLASSIFICATION'], data=[[0.5, 'A']]) # data argument includes example expected types for columns\n",
    "    J.writeVarJSON(outputVar, isInput=False, jPath=zipFolder)\n",
    "    \n",
    "    # Write model properties to a json file\n",
    "    J.writeModelPropertiesJSON(modelName=modelPrefix,\n",
    "                               modelDesc='',\n",
    "                               targetVariable=target,\n",
    "                               modelType='',\n",
    "                               modelPredictors=predict,\n",
    "                               targetEvent=1,\n",
    "                               numTargetCategories=1,\n",
    "                               eventProbVar='EM_EVENTPROBABILITY',\n",
    "                               jPath=zipFolder,\n",
    "                               modeler='sasdemo')\n",
    "    \n",
    "    # Write model metadata to a json file\n",
    "    J.writeFileMetadataJSON(modelPrefix, jPath=zipFolder)\n",
    "\n",
    "for (prefix, path) in zip(modelPrefix, zipFolder):\n",
    "    writeJSONFiles(hmeqData, predictorColumns, targetColumn, path, prefix)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "Collapsed": "false"
   },
   "outputs": [],
   "source": [
    "import getpass\n",
    "def writeModelStats(xTrain, yTrain, testProba, yTest, model, target, zipFolder, conn):\n",
    "    J = pzmm.JSONFiles()\n",
    "    \n",
    "    # Calculate train predictions\n",
    "    trainProba = model.predict_proba(xTrain)\n",
    "    \n",
    "    # Assign data to lists of actual and predicted values\n",
    "    trainData = pd.concat([yTrain.reset_index(drop=True), pd.Series(data=trainProba[:,1])], axis=1)\n",
    "    testData = pd.concat([yTest.reset_index(drop=True), pd.Series(data=testProba[:,1])], axis=1)\n",
    "    \n",
    "    # Calculate the model statistics and write to json files\n",
    "    J.calculateFitStat(trainData=trainData, testData=testData, jPath=zipFolder)\n",
    "    J.generateROCLiftStat(target, 1, conn, trainData=trainData, testData=testData, jPath=zipFolder)\n",
    "    \n",
    "username = getpass.getpass()\n",
    "password = getpass.getpass()\n",
    "host = 'demo.sas.com'\n",
    "sess = Session(host, username, password, protocol='http')\n",
    "conn = sess.as_swat()\n",
    "\n",
    "testProba = [yTreeProba, yForestProba, yGradientProba]\n",
    "for (m, proba, path) in zip(model, testProba, zipFolder):\n",
    "    writeModelStats(xTrain, yTrain, proba, yTest, m, targetColumn, path, conn)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "Collapsed": "false"
   },
   "outputs": [],
   "source": [
    "I = pzmm.ImportModel()\n",
    "for (prefix, path) in zip(modelPrefix, zipFolder):\n",
    "    I.pzmmImportModel(path, prefix, 'HMEQModels', x, y, '{}.predict({})', force=True)"
   ]
  }
 ],
 "metadata": {
  "interpreter": {
   "hash": "f9708d3f38eeab835578f0695c8890716ee809285281a28db6e379a5abca1310"
  },
  "kernelspec": {
   "display_name": "Python 3.8.3 64-bit ('dev-sasctl': conda)",
   "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": "3.8.3"
  },
  "latex_envs": {
   "LaTeX_envs_menu_present": true,
   "autoclose": true,
   "autocomplete": true,
   "bibliofile": "biblio.bib",
   "cite_by": "apalike",
   "current_citInitial": 1,
   "eqLabelWithNumbers": true,
   "eqNumInitial": 1,
   "hotkeys": {
    "equation": "Ctrl-E",
    "itemize": "Ctrl-I"
   },
   "labels_anchors": false,
   "latex_user_defs": false,
   "report_style_numbering": false,
   "user_envs_cfg": false
  },
  "varInspector": {
   "cols": {
    "lenName": 16,
    "lenType": 16,
    "lenVar": 40
   },
   "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
}