18-EX-Boosting-v1.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "Creative Commons CC BY 4.0 Lynd Bacon & Associates, Ltd. Not warranted to be suitable for any particular purpose. (You're on your own!)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Boosting\n",
    "\n",
    "Boosting algorithms are [_ensemble_ methods](https://scikit-learn.org/stable/modules/ensemble.html).  They apply a _sequence_ of algorithms, usually regression or classification learners, with the objective of improving predictive accuracy. Each learner in the sequence is trained using the results from the training of the previous learner.  Boosting ca reduce bias.\n",
    "\n",
    "\n",
    "Those results used, learner by learner, depend on the nature of the boosting algorithm.  They may be the residuals resulting from the fit provided by the previous learner (as in _Gradient Boosting_), or they may be the data weighted by the prediction errors resulting from the fit of the preceding learner(_AdaBoost_).\n",
    "\n",
    "The \"learners\" in the sequence are often decision trees, like CART, and they are often constrained to be small, with just a few features. \n",
    "\n",
    "On boosting: \n",
    "\n",
    "[Introduction to Statistical Learning Chapter 8](http://www-bcf.usc.edu/~gareth/ISL/)  \n",
    "[A Gentle Introduction to Boosting](https://machinelearningmastery.com/gentle-introduction-gradient-boosting-algorithm-machine-learning/)  \n",
    "\n",
    "In the following we apply gradient boosting and AdaBoost."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Getting Packages\n",
    "\n",
    "(There may be things here that we don't use.  Apologies.)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "from IPython.core.interactiveshell import InteractiveShell\n",
    "InteractiveShell.ast_node_interactivity = \"all\"\n",
    "from IPython.display import display, Markdown\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "import scikitplot as skplt\n",
    "from sklearn import linear_model\n",
    "from sklearn.model_selection import KFold, StratifiedShuffleSplit\n",
    "from sklearn.model_selection import cross_val_predict, train_test_split\n",
    "from sklearn.model_selection import GridSearchCV\n",
    "from sklearn import preprocessing\n",
    "from sklearn.metrics import accuracy_score, confusion_matrix, roc_auc_score\n",
    "from sklearn.metrics import roc_curve, precision_score, recall_score\n",
    "from sklearn.metrics import classification_report\n",
    "from sklearn.naive_bayes import GaussianNB\n",
    "from sklearn.ensemble import GradientBoostingClassifier, AdaBoostClassifier\n",
    "import pickle\n",
    "import os\n",
    "import shelve\n",
    "%matplotlib inline\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Gradient Boosting: Tree-Based Classification\n",
    "\n",
    "Like almost all other ML methods, Boosting methods have parameters that can be used for tuning performance.  When trees are boosted, the parameters include number of trees, a parameter that controls how quickly learning occurs across the trees, and the number of splits a tree can have.  The latter is often small, and may even be 1, in which case the tree is a _stump_.  \n",
    "\n",
    "In the following we're going to use the cervical CA data again. We'll apply [scikit-learn's GradientBoostingClassifier](https://scikit-learn.org/stable/modules/generated/sklearn.ensemble.GradientBoostingClassifier.html#sklearn.ensemble.GradientBoostingClassifier)."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Getting the Data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "# A DataFrame from the shelve db.  Adjust path as may be required.\n",
    "with shelve.open('DATA/ML/cervical') as sdb:\n",
    "    cervical2=sdb['cervical2']  # this should be, too"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "X=cervical2.iloc[:,1:].to_numpy()\n",
    "y=cervical2.iloc[:,0].to_numpy()\n",
    "feat_nams=cervical2.columns.to_list()[1:]  # could need these"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Simple Train, Test Split"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "## Simple Validation Split.  We'll try to balance the target variable labels in train and test.\n",
    "\n",
    "skf1=StratifiedShuffleSplit(n_splits=1,random_state=99) # Note, just a single random split here\n",
    "for trainNdx,testNdx in skf1.split(X,y):\n",
    "    Xtrain=X[trainNdx]\n",
    "    ytrain=y[trainNdx]\n",
    "    Xtest=X[testNdx]\n",
    "    ytest=y[testNdx]\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Select Some Parameter Values\n",
    "\n",
    "Let's use \"tiny\" trees which will be \"weak\" learners. Maximum depth of 1 split.  300 \"stumps.\"  \n",
    "Amongst the parameters of this boosting classifier, note the _warm_start_ option.  Setting it to True will allow adding additional trees to a trained classifier."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "treeBoosts = GradientBoostingClassifier(n_estimators=300, learning_rate=0.1, max_depth=1, random_state=99)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.93      1.00      0.96        63\n",
      "           1       0.00      0.00      0.00         5\n",
      "\n",
      "    accuracy                           0.93        68\n",
      "   macro avg       0.46      0.50      0.48        68\n",
      "weighted avg       0.86      0.93      0.89        68\n",
      "\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/lynd/anaconda3/lib/python3.6/site-packages/sklearn/metrics/classification.py:1437: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples.\n",
      "  'precision', 'predicted', average, warn_for)\n"
     ]
    }
   ],
   "source": [
    "treeBoostsF=treeBoosts.fit(Xtrain, ytrain)\n",
    "ypred = treeBoostsF.predict(Xtest)\n",
    "print(classification_report(ytest, ypred))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "See [Classification Report](https://scikit-learn.org/stable/modules/generated/sklearn.metrics.classification_report.html#sklearn.metrics.classification_report) for details about the above metrics."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# UDU: Graph the Feature Importances for the Classifier Above"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# UDU: Grid Search for a Better Boosted Tree Classifier\n",
    "\n",
    "Do a grid search to see if  you can get a better model than the above.  Metrics to explore include the learning rate (try varying above and below 0.1), max_depth (try increasing to 2 and 3)."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# UDU: Early Stopping\n",
    "\n",
    "One way to avoid overfitting is to stop training when an algorithm's fit on test data is no longer improving. The algorithm used above is one of those that has a feature that will end training when a criterion is reached, which are several.  For example, the \"n_iter_no_change\" parameter can be used to specify that training should stop when the validation score being used isn't changing for this many iterations.  \n",
    "\n",
    "Try setting n_iter_no_change to some small integer, like 10, and then run and evaluate the above classifer."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Gradient Boosting: Tree-Based Classification Take 2: Different Data\n",
    "\n",
    "Let's use the data set `dataset_37_diabetes.csv`.  It has one binary target, `class`, coded \"tested_positive\" and \"tested_negative\" for disease, and eight \"features.\""
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Getting the Data, Munging It"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "diabetes=pd.read_csv('DATA/ML/dataset_37_diabetes.csv')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "diabetes.isna().sum().sum()   # number of missing values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array(['tested_positive', 'tested_negative'], dtype=object)"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# What are those class labels?\n",
    "\n",
    "diabetes['class'].unique()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Weird.\n",
    "\n",
    "Let's recode the target variable into a 0,1 variable we'll call \"target\".  It's currently str data type"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([1, 0])"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "diabetes['target']=np.where(diabetes['class']==\"tested_positive\",1,0)\n",
    "diabetes.target.unique()\n",
    "diabetes.drop('class',axis=1,inplace=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "collapsed": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "preg        int64\n",
       "plas        int64\n",
       "pres        int64\n",
       "skin        int64\n",
       "insu        int64\n",
       "mass      float64\n",
       "pedi      float64\n",
       "age         int64\n",
       "target      int64\n",
       "dtype: object"
      ]
     },
     "execution_count": 26,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Let's check the other dtypes:\n",
    "\n",
    "diabetes.dtypes"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0    500\n",
       "1    268\n",
       "Name: target, dtype: int64"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Target label counts\n",
    "\n",
    "diabetes.target.value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(768, 8)"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "(768,)"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "## Convert to np arrays\n",
    "\n",
    "X=diabetes.iloc[:,:8].to_numpy()\n",
    "y=diabetes.iloc[:,8].to_numpy()\n",
    "X.shape\n",
    "y.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "## Training and test split\n",
    "skf1=StratifiedShuffleSplit(n_splits=1,random_state=99) # Note, just a single random split here\n",
    "for trainNdx,testNdx in skf1.split(X,y):\n",
    "    Xtrain=X[trainNdx]\n",
    "    ytrain=y[trainNdx]\n",
    "    Xtest=X[testNdx]\n",
    "    ytest=y[testNdx]\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [],
   "source": [
    "treeBoosts = GradientBoostingClassifier(n_estimators=300, learning_rate=0.1, max_depth=1, random_state=99)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {
    "collapsed": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.78      0.90      0.83        50\n",
      "           1       0.74      0.52      0.61        27\n",
      "\n",
      "    accuracy                           0.77        77\n",
      "   macro avg       0.76      0.71      0.72        77\n",
      "weighted avg       0.76      0.77      0.75        77\n",
      "\n"
     ]
    }
   ],
   "source": [
    "treeBoostsFit=treeBoosts.fit(Xtrain, ytrain)\n",
    "ypred = treeBoostsFit.predict(Xtest)\n",
    "print(classification_report(ytest, ypred))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## ROC, AUC"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [],
   "source": [
    "ypredProb=treeBoostsFit.predict_proba(Xtest)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {
    "collapsed": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 504x504 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "skplt.metrics.plot_roc(ytest, ypredProb, title=\"Gradient Boosting Test Data ROC\",\n",
    "                      figsize=(7,7),\n",
    "                      plot_micro=False,\n",
    "                      classes_to_plot=1)\n",
    "plt.show();"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# UDU: Plot the Features Importances for the Above Boosted Classifier"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# UDU: Gradient Boosting Regression\n",
    "\n",
    "Try training a Boosted regression model using the _radon_ data.\n",
    "\n",
    "Here's where you can find the scikit-learn documentation to do it: [GradientBoostingRegressor](https://scikit-learn.org/stable/modules/generated/sklearn.ensemble.GradientBoostingRegressor.html#sklearn.ensemble.GradientBoostingRegressor). "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# AdaBoost Classifier\n",
    "\n",
    "Let's give AdaBoost a try using the diabetes data.  Here's where the documentation is pm the scikit-learn site:\n",
    "\n",
    "[AdaBoostClassifier](https://scikit-learn.org/stable/modules/generated/sklearn.ensemble.AdaBoostClassifier.html#sklearn.ensemble.AdaBoostClassifier)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {
    "collapsed": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.79      0.88      0.83        50\n",
      "           1       0.71      0.56      0.63        27\n",
      "\n",
      "    accuracy                           0.77        77\n",
      "   macro avg       0.75      0.72      0.73        77\n",
      "weighted avg       0.76      0.77      0.76        77\n",
      "\n"
     ]
    }
   ],
   "source": [
    "AdaBoosts=AdaBoostClassifier(n_estimators=300,random_state=99,\n",
    "                             learning_rate=0.1)\n",
    "AdaBoostsFit=AdaBoosts.fit(Xtrain, ytrain)\n",
    "ypred = AdaBoostsFit.predict(Xtest)\n",
    "print(classification_report(ytest, ypred))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {
    "collapsed": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "AdaBoost Feature Importances:\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "array([0.04666667, 0.22666667, 0.05666667, 0.01      , 0.05666667,\n",
       "       0.17666667, 0.27666667, 0.15      ])"
      ]
     },
     "execution_count": 36,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "print('AdaBoost Feature Importances:')\n",
    "\n",
    "AdaBoostsFit.feature_importances_"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Looks about the same as the gradient boosting results?  Here's the ROC:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [],
   "source": [
    "ypredProb=AdaBoostsFit.predict_proba(Xtest)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {
    "collapsed": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 504x504 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "skplt.metrics.plot_roc(ytest, ypredProb, title=\"ADaBoosted Test Data ROC\",\n",
    "                      figsize=(7,7),\n",
    "                      plot_micro=False,\n",
    "                      classes_to_plot=1)\n",
    "plt.show();"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# UDU: AdaBoosted Regression\n",
    "\n",
    "Try training a Boosted regression model using the _radon_ data.\n",
    "\n",
    "Here's where you can find the scikit-learn documentation to do it: [AdaBoostRegressor](https://scikit-learn.org/stable/modules/generated/sklearn.ensemble.AdaBoostRegressor.html#sklearn.ensemble.AdaBoostRegressor). "
   ]
  }
 ],
 "metadata": {
  "hide_input": false,
  "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": "3.6.8"
  },
  "latex_envs": {
   "LaTeX_envs_menu_present": true,
   "autoclose": false,
   "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
  },
  "toc": {
   "base_numbering": 1,
   "nav_menu": {},
   "number_sections": false,
   "sideBar": true,
   "skip_h1_title": false,
   "title_cell": "Table of Contents",
   "title_sidebar": "Contents",
   "toc_cell": false,
   "toc_position": {
    "height": "calc(100% - 180px)",
    "left": "10px",
    "top": "150px",
    "width": "315.8px"
   },
   "toc_section_display": true,
   "toc_window_display": false
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}