alg/autoprognosis/tutorial_autoprognosis_api.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## AutoPrognosis API Tutorial\n",
    "\n",
    "A demonstration for AP functionality and operation"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "This tutorial shows how to use [Autoprognosis](https://arxiv.org/abs/1802.07207). We are using the UCI Spam dataset.\n",
    "\n",
    "See [installation instructions](../../doc/install.md) to install the dependencies.\n",
    "\n",
    "Load dataset and show the first five samples:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import initpath_ap\n",
    "initpath_ap.init_sys_path()\n",
    "import utilmlab\n",
    "\n",
    "from sklearn.datasets import load_breast_cancer\n",
    "\n",
    "df = load_breast_cancer()\n",
    "X_ = pd.DataFrame(df.data)\n",
    "Y_ = pd.DataFrame(df.target)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Import the AutoPrognosis library"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Run the model with few iterations"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[ mean, Gradient Boosting ]\n",
      "[ most_frequent, MultinomialNaiveBayes ]\n",
      "[ median, LinearSVM ]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Widget Javascript not detected.  It may not be installed properly. Did you enable the widgetsnbextension? If not, then run \"jupyter nbextension enable --py --sys-prefix widgetsnbextension\"\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[ mean, XGBoost ]\n",
      "[ mean, BernoullinNaiveBayes ]\n",
      "[ most_frequent, LinearSVM ]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Iteration number: 1 3s (3s) (8s), Current pipelines:  [[[ median, XGBoost ]]], [[[ mean, BernoullinNaiveBayes ]]], [[[ median, LinearSVM ]]], BO objective: -0.9891936728238395\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[ median, XGBoost ]\n",
      "[ median, BernoullinNaiveBayes ]\n",
      "[ median, QDA ]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Iteration number: 2 5s (3s) (8s), Current pipelines:  [[[ mean, XGBoost ]]], [[[ mean, BernoullinNaiveBayes ]]], [[[ median, QDA ]]], BO objective: -0.9999999999999997\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[ missForest, Random Forest ]\n",
      "[ mean, Bagging ]\n"
     ]
    }
   ],
   "source": [
    "metric = 'aucprc'\n",
    "acquisition_type = 'MPI' # default and prefered is LCB but this generates excessive warnings, MPI is a good compromise.\n",
    "AP_mdl   = model.AutoPrognosis_Classifier(\n",
    "    metric=metric, CV=5, num_iter=3, kernel_freq=100, ensemble=True,\n",
    "    ensemble_size=3, Gibbs_iter=100, burn_in=50, num_components=3, \n",
    "    acquisition_type=acquisition_type)\n",
    "\n",
    "AP_mdl.fit(X_, Y_)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Computing model predictions"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "##### ~~~First element in the output is the predictions of a single model, the second element is the prediction of the ensemble~~~"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "AP_mdl.predict(X_)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Compute performance via multi-fold cross-validation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "model.evaluate_ens(X_, Y_, AP_mdl, n_folds=5, visualize=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Visualize data..."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "AP_mdl.visualize_data(X_)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Visualize the model..."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "AP_mdl.APReport()"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "widgets": {
   "state": {
    "b36d11ca14b24a118b3c3a295a788faf": {
     "views": [
      {
       "cell_index": 6
      }
     ]
    }
   },
   "version": "1.2.0"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}