4_4_XGB_implementation.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "4.4. XGB implementation.ipynb",
      "version": "0.3.2",
      "provenance": [],
      "collapsed_sections": [],
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python2",
      "display_name": "Python 2"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "[View in Colaboratory](https://colab.research.google.com/github/adowaconan/Deep_learning_fMRI/blob/master/4_4_XGB_implementation.ipynb)"
      ]
    },
    {
      "metadata": {
        "id": "isvnwR348yRI",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 635
        },
        "outputId": "d79b894c-714f-413d-a2e6-9ac9402cbd72"
      },
      "cell_type": "code",
      "source": [
        "!pip install -U -q PyDrive\n",
        "!pip install tqdm\n",
        "!apt-get install swig\n",
        "!pip install -U pymvpa2\n",
        "!pip install -U seaborn\n",
        "\n",
        "# this lets you access to googel drive shared files\n",
        "from pydrive.auth import GoogleAuth\n",
        "from pydrive.drive import GoogleDrive\n",
        "from google.colab import auth\n",
        "from oauth2client.client import GoogleCredentials\n",
        "\n",
        "# 1. Authenticate and create the PyDrive client.\n",
        "auth.authenticate_user()\n",
        "gauth = GoogleAuth()\n",
        "gauth.credentials = GoogleCredentials.get_application_default()\n",
        "drive = GoogleDrive(gauth)\n",
        "\n",
        "# why not numpy?\n",
        "import numpy as np\n",
        "import zipfile\n",
        "from sklearn.metrics import roc_auc_score\n",
        "\n",
        "\n",
        "from glob import glob # useless\n",
        "from tqdm import tqdm # maybe useful\n",
        "import pickle # very useful\n",
        "from collections import Counter # I'm a counter\n",
        "from mvpa2.mappers.fx import mean_group_sample # pymvpa\n",
        "from mvpa2.generators.partition import NFoldPartitioner # pymvpa\n",
        "from mvpa2.base.node import ChainNode # pymvpa\n",
        "from mvpa2.generators.resampling import Balancer # pymvpa\n",
        "from mvpa2.generators.base import Sifter # pymvpa\n",
        "from mvpa2.mappers.base import Mapper # pymvpa\n",
        "from keras.layers import Dense,Dropout,BatchNormalization\n",
        "from keras.layers import Input, AveragePooling3D,Reshape\n",
        "from keras.layers import Conv2D,Conv2DTranspose\n",
        "from keras.layers import GaussianNoise # add noise to our data\n",
        "from keras.layers import Flatten,Dense,LeakyReLU,Conv3DTranspose\n",
        "from keras.models import Model\n",
        "import keras\n",
        "from keras.callbacks import ModelCheckpoint,TensorBoard\n",
        "from keras import regularizers\n",
        "from sklearn import metrics\n",
        "import pandas as pd\n",
        "import os\n",
        "from sklearn.model_selection import StratifiedKFold\n",
        "\n",
        "import seaborn as sns\n",
        "sns.set_context('poster')\n",
        "sns.set_style('whitegrid')\n",
        "import matplotlib.pyplot as plt\n",
        "\n",
        "from sklearn.preprocessing import MinMaxScaler\n",
        "\n",
        "data_id = 'https://drive.google.com/open?id=1sMQh_8qhCwN7N1yM8mTEBT9m8q7l_0iv'.split('id=')[-1]\n",
        "zip_import = drive.CreateFile({'id':data_id})\n",
        "zip_import.GetContentFile('{}'.format('all_bin.pkl'))"
      ],
      "execution_count": 1,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Requirement already satisfied: tqdm in /usr/local/lib/python2.7/dist-packages (4.25.0)\n",
            "Reading package lists... Done\n",
            "Building dependency tree       \n",
            "Reading state information... Done\n",
            "swig is already the newest version (3.0.10-1.2).\n",
            "0 upgraded, 0 newly installed, 0 to remove and 0 not upgraded.\n",
            "Requirement already up-to-date: pymvpa2 in /usr/local/lib/python2.7/dist-packages (2.6.5)\n",
            "Requirement already satisfied, skipping upgrade: scipy in /usr/local/lib/python2.7/dist-packages (from pymvpa2) (0.19.1)\n",
            "Requirement already satisfied, skipping upgrade: nibabel in /usr/local/lib/python2.7/dist-packages (from pymvpa2) (2.3.0)\n",
            "Requirement already satisfied, skipping upgrade: numpy>=1.8.2 in /usr/local/lib/python2.7/dist-packages (from scipy->pymvpa2) (1.14.5)\n",
            "Collecting seaborn\n",
            "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/7a/bf/04cfcfc9616cedd4b5dd24dfc40395965ea9f50c1db0d3f3e52b050f74a5/seaborn-0.9.0.tar.gz (198kB)\n",
            "\u001b[K    100% |████████████████████████████████| 204kB 21.5MB/s \n",
            "\u001b[?25hRequirement already satisfied, skipping upgrade: numpy>=1.9.3 in /usr/local/lib/python2.7/dist-packages (from seaborn) (1.14.5)\n",
            "Requirement already satisfied, skipping upgrade: scipy>=0.14.0 in /usr/local/lib/python2.7/dist-packages (from seaborn) (0.19.1)\n",
            "Requirement already satisfied, skipping upgrade: pandas>=0.15.2 in /usr/local/lib/python2.7/dist-packages (from seaborn) (0.22.0)\n",
            "Requirement already satisfied, skipping upgrade: matplotlib>=1.4.3 in /usr/local/lib/python2.7/dist-packages (from seaborn) (2.1.2)\n",
            "Requirement already satisfied, skipping upgrade: pytz>=2011k in /usr/local/lib/python2.7/dist-packages (from pandas>=0.15.2->seaborn) (2018.5)\n",
            "Requirement already satisfied, skipping upgrade: python-dateutil in /usr/local/lib/python2.7/dist-packages (from pandas>=0.15.2->seaborn) (2.5.3)\n",
            "Requirement already satisfied, skipping upgrade: cycler>=0.10 in /usr/local/lib/python2.7/dist-packages (from matplotlib>=1.4.3->seaborn) (0.10.0)\n",
            "Requirement already satisfied, skipping upgrade: backports.functools-lru-cache in /usr/local/lib/python2.7/dist-packages (from matplotlib>=1.4.3->seaborn) (1.5)\n",
            "Requirement already satisfied, skipping upgrade: subprocess32 in /usr/local/lib/python2.7/dist-packages (from matplotlib>=1.4.3->seaborn) (3.5.2)\n",
            "Requirement already satisfied, skipping upgrade: six>=1.10 in /usr/local/lib/python2.7/dist-packages (from matplotlib>=1.4.3->seaborn) (1.11.0)\n",
            "Requirement already satisfied, skipping upgrade: pyparsing!=2.0.4,!=2.1.2,!=2.1.6,>=2.0.1 in /usr/local/lib/python2.7/dist-packages (from matplotlib>=1.4.3->seaborn) (2.2.0)\n",
            "Building wheels for collected packages: seaborn\n",
            "  Running setup.py bdist_wheel for seaborn ... \u001b[?25l-\b \bdone\n",
            "\u001b[?25h  Stored in directory: /root/.cache/pip/wheels/fc/1c/74/c8f80a532c06a789599b8659b117ec7d7574cac4a06f7dabfe\n",
            "Successfully built seaborn\n",
            "Installing collected packages: seaborn\n",
            "  Found existing installation: seaborn 0.7.1\n",
            "    Uninstalling seaborn-0.7.1:\n",
            "      Successfully uninstalled seaborn-0.7.1\n",
            "Successfully installed seaborn-0.9.0\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "Using TensorFlow backend.\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "metadata": {
        "id": "S0ESzur4_PLw",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "# Read condition\n",
        "\n",
        "## Experiment:\n",
        "\n",
        "### Subjects in the sanner, and were instructed to read the word repeatedly as it was present. The words can be divided into 2 semantic categories: living animals and non-living tools\n",
        "\n",
        "## words:\n",
        "\n",
        "'llave', 'lapiz', 'tijera', 'aguja', 'pinza', 'sierra', 'clavo',\n",
        "'pincel', 'alicate', 'tuerca', 'navaja', 'cepillo', 'taladro',\n",
        "'soplete', 'tornillo', 'cuchara', 'martillo', 'cuchillo', 'tigre',\n",
        "'gallo', 'perro', 'oveja', 'cerdo', 'gorila', 'burro', 'yegua',\n",
        "'ardilla', 'conejo', 'gallina', 'caballo', 'ballena', 'pantera',\n",
        "'camello', 'tortuga', 'elefante', 'canguro'\n",
        "<br />**can you divide them?**"
      ]
    },
    {
      "metadata": {
        "id": "fjXmhHtxAkQU",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "from IPython.display import Image\n",
        "image_id = 'https://drive.google.com/open?id=1qZfXGiqtQ9g5DJsCtb_Injy_tSCEJ3dG'.split('id=')[-1]\n",
        "image = drive.CreateFile({'id':image_id})\n",
        "image.GetContentFile('{}'.format('words.png'))\n"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "7DmbM_p99kUp",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "dataset_ = pickle.load(open('all_bin.pkl','rb'))\n",
        "dataset = dataset_[dataset_.sa.context == 'read']"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "li0Pv-xG-CIZ",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 635
        },
        "outputId": "1bd85638-8e35-4188-a614-1ed2e8d3cb13"
      },
      "cell_type": "code",
      "source": [
        "print(dataset.summary())"
      ],
      "execution_count": 4,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Dataset: 1544x16130@float32, <sa: categories,chunks,context,id,targets,time_coords,time_indices,trials,words>, <fa: voxel_indices>, <a: imgaffine,imghdr,imgtype,mapper,voxel_dim,voxel_eldim>\n",
            "stats: mean=8.95748e-10 std=0.993167 var=0.98638 min=-6.82599 max=6.5985\n",
            "\n",
            "Counts of targets in each chunk:\n",
            "  chunks\\targets animal tool\n",
            "                   ---   ---\n",
            "        0          111   111\n",
            "        2          110   111\n",
            "        4          110   110\n",
            "        6          108   115\n",
            "        8          109   109\n",
            "       10          108   110\n",
            "       12          111   111\n",
            "\n",
            "Summary for targets across chunks\n",
            "  targets mean  std min max #chunks\n",
            "  animal   110 1.18 108 111    7\n",
            "   tool    111 1.77 109 115    7\n",
            "\n",
            "Summary for chunks across targets\n",
            "  chunks mean std min max #targets\n",
            "    0     111  0  111 111     2\n",
            "    2     110 0.5 110 111     2\n",
            "    4     110  0  110 110     2\n",
            "    6     112 3.5 108 115     2\n",
            "    8     109  0  109 109     2\n",
            "   10     109  1  108 110     2\n",
            "   12     111  0  111 111     2\n",
            "Sequence statistics for 1544 entries from set ['animal', 'tool']\n",
            "Counter-balance table for orders up to 2:\n",
            "Targets/Order  O1      |   O2      |\n",
            "   animal:    702  65  |  637 130  |\n",
            "    tool:      64 712  |  128 647  |\n",
            "Correlations: min=-0.19 max=0.83 mean=-0.00065 sum(abs)=71\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "87KDWiYKBExj",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "import xgboost as xgb\n"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "hDM-CgLo0mWy",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "[A Gentle Introduction to XGBoost for Applied Machine Learning](https://machinelearningmastery.com/gentle-introduction-xgboost-applied-machine-learning/)\n",
        "\n",
        "\n",
        "https://www.kaggle.com/datacanary/xgboost-example-python"
      ]
    },
    {
      "metadata": {
        "id": "j93ipAfIulf9",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "from sklearn.svm import SVC\n",
        "from sklearn.preprocessing import MinMaxScaler\n",
        "from sklearn.decomposition import PCA\n",
        "from sklearn.pipeline import make_pipeline\n",
        "from sklearn.model_selection import cross_val_score,StratifiedShuffleSplit\n",
        "from sklearn.metrics import roc_auc_score,roc_curve\n",
        "from sklearn.metrics import (classification_report,\n",
        "                             matthews_corrcoef,\n",
        "                             confusion_matrix,\n",
        "                             f1_score,\n",
        "                             log_loss)\n",
        "\n",
        "# helper funtion\n",
        "def get_blocks_metasema(dataset__,label_map):\n",
        "    \"\"\"\n",
        "    # use ids, chunks,and labels to make unique blocks of the pre-average dataset, because I don't want to \n",
        "    # average the dataset until I actually want to, but at the same time, I want to balance the data for \n",
        "    # both the training and test set.\n",
        "    \n",
        "    Preparation for cross validation functions used in scit-kit learning\n",
        "    \"\"\"\n",
        "    ids                 = dataset__.sa.id.astype(int)\n",
        "    chunks              = dataset__.sa.chunks\n",
        "    labels              = np.array([label_map[item] for item in dataset__.targets])[:,-1]\n",
        "    sample_indecies     = np.arange(len(labels))\n",
        "    blocks              = [np.array([ids[ids             == target],\n",
        "                                     chunks[ids          == target], \n",
        "                                     labels[ids          == target],\n",
        "                                     sample_indecies[ids == target]\n",
        "                                     ]) for target in np.unique(ids)]\n",
        "    block_labels        = np.array([np.unique(ll[-2]) for ll in blocks]).ravel()\n",
        "    return blocks,block_labels\n",
        "def Find_Optimal_Cutoff(target, predicted):\n",
        "    \"\"\" Find the optimal probability cutoff point for a classification model related to event rate\n",
        "    Parameters\n",
        "    ----------\n",
        "    target : Matrix with dependent or target data, where rows are observations\n",
        "\n",
        "    predicted : Matrix with predicted data, where rows are observations\n",
        "\n",
        "    Returns\n",
        "    -------     \n",
        "    list type, with optimal cutoff value\n",
        "\n",
        "    \"\"\"\n",
        "    fpr, tpr, threshold         = roc_curve(target, predicted)\n",
        "    i                           = np.arange(len(tpr)) \n",
        "    roc                         = pd.DataFrame({'tf' : pd.Series(tpr-(1-fpr), index=i), 'threshold' : pd.Series(threshold, index=i)})\n",
        "    roc_t                       = roc.iloc[(roc.tf-0).abs().argsort()[:1]]\n",
        "\n",
        "    return list(roc_t['threshold']) "
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "KytHcoy7u4Z9",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 4296
        },
        "outputId": "7340243e-0bef-467a-cde1-70b08ff1acbf"
      },
      "cell_type": "code",
      "source": [
        "average = True # average the instances before training the baseline model\n",
        "\n",
        "scores = []\n",
        "cfm    = []\n",
        "weight = []\n",
        "label_map = dict(animal  = [0,1],\n",
        "                 tool    = [1,0])\n",
        "blocks, block_labels = get_blocks_metasema(dataset,label_map)\n",
        "cv = StratifiedShuffleSplit(n_splits     = 30,# random partition 30 times, but ideally, we need 100 or more\n",
        "                            test_size    = 0.2, # 20 % as the test data\n",
        "                            random_state = 12345 # for reproducibility purpose\n",
        "                           )\n",
        "for fold, (train,test) in enumerate(cv.split(blocks,block_labels)):\n",
        "    idx_train = np.concatenate([blocks[tr_element][-1] for tr_element in train])\n",
        "    idx_test  = np.concatenate([blocks[te_element][-1] for te_element in test ])\n",
        "    scaler    = MinMaxScaler()\n",
        "    \n",
        "    ##############################################################################\n",
        "    \n",
        "    \n",
        "    # select the training and testing split again to make sure we do the right cv\n",
        "    tr                    = dataset[idx_train]\n",
        "    te                    = dataset[idx_test]\n",
        "    # average the train data for baseline to improve the signal to noise ratio\n",
        "    if average:\n",
        "        tr                = tr.get_mapped(mean_group_sample(['chunks', 'trials'],order = 'occurrence'))\n",
        "    \n",
        "    te                    = te.get_mapped(mean_group_sample(['chunks', 'trials'],order = 'occurrence'))\n",
        "    y_train               = np.array([label_map[item] for item in tr.targets])[:,-1]\n",
        "    # transform the test labels as well\n",
        "    y_test                = np.array([label_map[item] for item in te.targets])[:,-1]\n",
        "    \n",
        "    X_train               = scaler.fit_transform(tr.samples.astype('float32'))\n",
        "    X_test                = scaler.transform(te.samples.astype('float32'))\n",
        "    \n",
        "    bst                   = xgb.XGBClassifier(#max_depth = 7, \n",
        "                                              n_estimators = 1200, \n",
        "                                              learning_rate = 0.05,\n",
        "                                              subsample = 0.9,\n",
        "                                              random_state = 12345,\n",
        "                                              reg_alpha = 1e-4, \n",
        "                                              reg_lambda = 1e-4, \n",
        "                                              booster = 'gbtree',\n",
        "                                              objective = 'binary:logistic')\n",
        "    bst.fit(X_train,y_train)\n",
        "    \n",
        "    \n",
        "    # make the probabilistic prediction\n",
        "    pred_                 = bst.predict_proba(X_test)[:,-1]\n",
        "    # find the optimal cut-off probability in the roc curve\n",
        "    score_baseline1       = roc_auc_score(y_test,pred_)\n",
        "    threshold_            = Find_Optimal_Cutoff(y_test,pred_)[0]\n",
        "    score_baseline2       = matthews_corrcoef(y_test,pred_>threshold_) # MCC: conservative\n",
        "    score_baseline3       = f1_score(y_test,pred_>threshold_) # f1 score, based on precision and recall, biased to positive class\n",
        "    score_baseline4       = log_loss(y_test,bst.predict_proba(X_test))# an unbiased metric, but hard to interpret\n",
        "    tn, fp, fn, tp        = confusion_matrix(y_test,pred_>threshold_).ravel()# unbiased metric, but hard to plot in complex experiments like this with multiple factors\n",
        "    print('fold {}'.format(fold + 1))\n",
        "    print(classification_report(y_test,pred_>threshold_))\n",
        "    scores.append([score_baseline1,score_baseline2,score_baseline3,score_baseline4])\n",
        "    cfm.append(   [tn,fp,fn,tp])"
      ],
      "execution_count": 7,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "WARNING: There were no samples for combination {'chunks': 0, 'trials': '16'}. It might be a sign of a disbalanced dataset <Dataset: 1230x16130@float32, <sa: categories,chunks,context,id,targets,time_coords,time_indices,trials,words>, <fa: voxel_indices>, <a: imgaffine,imghdr,imgtype,mapper,voxel_dim,voxel_eldim>>.\n",
            " * Please note: warnings are printed only once, but underlying problem might occur many times *\n",
            "fold 1\n",
            "             precision    recall  f1-score   support\n",
            "\n",
            "          0       0.58      0.58      0.58        26\n",
            "          1       0.56      0.56      0.56        25\n",
            "\n",
            "avg / total       0.57      0.57      0.57        51\n",
            "\n",
            "fold 2\n",
            "             precision    recall  f1-score   support\n",
            "\n",
            "          0       0.56      0.58      0.57        26\n",
            "          1       0.54      0.52      0.53        25\n",
            "\n",
            "avg / total       0.55      0.55      0.55        51\n",
            "\n",
            "fold 3\n",
            "             precision    recall  f1-score   support\n",
            "\n",
            "          0       0.59      0.62      0.60        26\n",
            "          1       0.58      0.56      0.57        25\n",
            "\n",
            "avg / total       0.59      0.59      0.59        51\n",
            "\n",
            "fold 4\n",
            "             precision    recall  f1-score   support\n",
            "\n",
            "          0       0.58      0.58      0.58        26\n",
            "          1       0.56      0.56      0.56        25\n",
            "\n",
            "avg / total       0.57      0.57      0.57        51\n",
            "\n",
            "fold 5\n",
            "             precision    recall  f1-score   support\n",
            "\n",
            "          0       0.46      0.52      0.49        25\n",
            "          1       0.48      0.42      0.45        26\n",
            "\n",
            "avg / total       0.47      0.47      0.47        51\n",
            "\n",
            "fold 6\n",
            "             precision    recall  f1-score   support\n",
            "\n",
            "          0       0.58      0.60      0.59        25\n",
            "          1       0.60      0.58      0.59        26\n",
            "\n",
            "avg / total       0.59      0.59      0.59        51\n",
            "\n",
            "fold 7\n",
            "             precision    recall  f1-score   support\n",
            "\n",
            "          0       0.48      0.50      0.49        26\n",
            "          1       0.46      0.44      0.45        25\n",
            "\n",
            "avg / total       0.47      0.47      0.47        51\n",
            "\n",
            "fold 8\n",
            "             precision    recall  f1-score   support\n",
            "\n",
            "          0       0.62      0.64      0.63        25\n",
            "          1       0.64      0.62      0.63        26\n",
            "\n",
            "avg / total       0.63      0.63      0.63        51\n",
            "\n",
            "fold 9\n",
            "             precision    recall  f1-score   support\n",
            "\n",
            "          0       0.50      0.50      0.50        26\n",
            "          1       0.48      0.48      0.48        25\n",
            "\n",
            "avg / total       0.49      0.49      0.49        51\n",
            "\n",
            "fold 10\n",
            "             precision    recall  f1-score   support\n",
            "\n",
            "          0       0.65      0.68      0.67        25\n",
            "          1       0.68      0.65      0.67        26\n",
            "\n",
            "avg / total       0.67      0.67      0.67        51\n",
            "\n",
            "fold 11\n",
            "             precision    recall  f1-score   support\n",
            "\n",
            "          0       0.52      0.56      0.54        25\n",
            "          1       0.54      0.50      0.52        26\n",
            "\n",
            "avg / total       0.53      0.53      0.53        51\n",
            "\n",
            "fold 12\n",
            "             precision    recall  f1-score   support\n",
            "\n",
            "          0       0.58      0.60      0.59        25\n",
            "          1       0.60      0.58      0.59        26\n",
            "\n",
            "avg / total       0.59      0.59      0.59        51\n",
            "\n",
            "fold 13\n",
            "             precision    recall  f1-score   support\n",
            "\n",
            "          0       0.56      0.56      0.56        25\n",
            "          1       0.58      0.58      0.58        26\n",
            "\n",
            "avg / total       0.57      0.57      0.57        51\n",
            "\n",
            "fold 14\n",
            "             precision    recall  f1-score   support\n",
            "\n",
            "          0       0.56      0.58      0.57        26\n",
            "          1       0.54      0.52      0.53        25\n",
            "\n",
            "avg / total       0.55      0.55      0.55        51\n",
            "\n",
            "fold 15\n",
            "             precision    recall  f1-score   support\n",
            "\n",
            "          0       0.54      0.58      0.56        26\n",
            "          1       0.52      0.48      0.50        25\n",
            "\n",
            "avg / total       0.53      0.53      0.53        51\n",
            "\n",
            "fold 16\n",
            "             precision    recall  f1-score   support\n",
            "\n",
            "          0       0.56      0.58      0.57        26\n",
            "          1       0.54      0.52      0.53        25\n",
            "\n",
            "avg / total       0.55      0.55      0.55        51\n",
            "\n",
            "fold 17\n",
            "             precision    recall  f1-score   support\n",
            "\n",
            "          0       0.59      0.62      0.60        26\n",
            "          1       0.58      0.56      0.57        25\n",
            "\n",
            "avg / total       0.59      0.59      0.59        51\n",
            "\n",
            "fold 18\n",
            "             precision    recall  f1-score   support\n",
            "\n",
            "          0       0.59      0.62      0.60        26\n",
            "          1       0.58      0.56      0.57        25\n",
            "\n",
            "avg / total       0.59      0.59      0.59        51\n",
            "\n",
            "fold 19\n",
            "             precision    recall  f1-score   support\n",
            "\n",
            "          0       0.52      0.56      0.54        25\n",
            "          1       0.54      0.50      0.52        26\n",
            "\n",
            "avg / total       0.53      0.53      0.53        51\n",
            "\n",
            "fold 20\n",
            "             precision    recall  f1-score   support\n",
            "\n",
            "          0       0.69      0.72      0.71        25\n",
            "          1       0.72      0.69      0.71        26\n",
            "\n",
            "avg / total       0.71      0.71      0.71        51\n",
            "\n",
            "fold 21\n",
            "             precision    recall  f1-score   support\n",
            "\n",
            "          0       0.65      0.65      0.65        26\n",
            "          1       0.64      0.64      0.64        25\n",
            "\n",
            "avg / total       0.65      0.65      0.65        51\n",
            "\n",
            "fold 22\n",
            "             precision    recall  f1-score   support\n",
            "\n",
            "          0       0.50      0.54      0.52        26\n",
            "          1       0.48      0.44      0.46        25\n",
            "\n",
            "avg / total       0.49      0.49      0.49        51\n",
            "\n",
            "fold 23\n",
            "             precision    recall  f1-score   support\n",
            "\n",
            "          0       0.62      0.62      0.62        26\n",
            "          1       0.60      0.60      0.60        25\n",
            "\n",
            "avg / total       0.61      0.61      0.61        51\n",
            "\n",
            "fold 24\n",
            "             precision    recall  f1-score   support\n",
            "\n",
            "          0       0.52      0.50      0.51        26\n",
            "          1       0.50      0.52      0.51        25\n",
            "\n",
            "avg / total       0.51      0.51      0.51        51\n",
            "\n",
            "fold 25\n",
            "             precision    recall  f1-score   support\n",
            "\n",
            "          0       0.61      0.65      0.63        26\n",
            "          1       0.61      0.56      0.58        25\n",
            "\n",
            "avg / total       0.61      0.61      0.61        51\n",
            "\n",
            "fold 26\n",
            "             precision    recall  f1-score   support\n",
            "\n",
            "          0       0.68      0.73      0.70        26\n",
            "          1       0.70      0.64      0.67        25\n",
            "\n",
            "avg / total       0.69      0.69      0.69        51\n",
            "\n",
            "fold 27\n",
            "             precision    recall  f1-score   support\n",
            "\n",
            "          0       0.65      0.65      0.65        26\n",
            "          1       0.64      0.64      0.64        25\n",
            "\n",
            "avg / total       0.65      0.65      0.65        51\n",
            "\n",
            "fold 28\n",
            "             precision    recall  f1-score   support\n",
            "\n",
            "          0       0.65      0.65      0.65        26\n",
            "          1       0.64      0.64      0.64        25\n",
            "\n",
            "avg / total       0.65      0.65      0.65        51\n",
            "\n",
            "fold 29\n",
            "             precision    recall  f1-score   support\n",
            "\n",
            "          0       0.52      0.50      0.51        26\n",
            "          1       0.50      0.52      0.51        25\n",
            "\n",
            "avg / total       0.51      0.51      0.51        51\n",
            "\n",
            "fold 30\n",
            "             precision    recall  f1-score   support\n",
            "\n",
            "          0       0.61      0.65      0.63        26\n",
            "          1       0.61      0.56      0.58        25\n",
            "\n",
            "avg / total       0.61      0.61      0.61        51\n",
            "\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "SgHTeQ9cxThQ",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "scores = np.array(scores)\n",
        "cfm    = np.array(cfm)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "pRr3YYi10S2c",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 52
        },
        "outputId": "f85c22bc-b800-4bd5-ffcd-15600cce9e9e"
      },
      "cell_type": "code",
      "source": [
        "print('ROC AUC score = {:.3f} +/- {:.3f}\\nf1 score = {:.3f} +/- {:.3f}'.format(\n",
        "        scores.mean(0)[0],scores.std(0)[0],scores.mean(0)[-2],scores.std(0)[-2]\n",
        "        ))"
      ],
      "execution_count": 9,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "ROC AUC score = 0.603 +/- 0.060\n",
            "f1 score = 0.564 +/- 0.065\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "lsRCsI4z0UTp",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "cm = cfm.mean(0).reshape(2,2)\n",
        "cm = cm / cm.sum(axis=1)[:,np.newaxis]\n",
        "cm = pd.DataFrame(cm,columns=['tool','animal'],index=['tool','animal'])"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "7Mr9TdvM0VuW",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 675
        },
        "outputId": "5b2baac4-8779-4c2d-97b3-48cb82d0a747"
      },
      "cell_type": "code",
      "source": [
        "fig, ax = plt.subplots(figsize=(8,8))\n",
        "sns.heatmap(cm,ax=ax,vmin=0.4,vmax=0.65)\n",
        "ax.set(ylabel = 'True', xlabel = 'Predicted',title = 'Extreem Boosting')"
      ],
      "execution_count": 11,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python2.7/dist-packages/seaborn/matrix.py:143: DeprecationWarning: elementwise == comparison failed; this will raise an error in the future.\n",
            "  self.xticks = \"auto\"\n",
            "/usr/local/lib/python2.7/dist-packages/seaborn/matrix.py:151: DeprecationWarning: elementwise == comparison failed; this will raise an error in the future.\n",
            "  self.yticklabels = []\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[Text(41.8,0.5,u'True'),\n",
              " Text(0.5,41.8,u'Predicted'),\n",
              " Text(0.5,1,u'Extreem Boosting')]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 11
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAfkAAAIDCAYAAAAHV2cGAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nIErkASBK5AEgSuQBIErkASBK5AEgah2gK/6U33ZsAAAAAElFTkSuQmCC\n",
            "text/plain": [
              "<matplotlib.figure.Figure at 0x7f23bfc671d0>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "metadata": {
        "id": "8osshDVW0ZPf",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 123
        },
        "outputId": "85c57b44-f1bf-4f33-c1ff-a9760220007c"
      },
      "cell_type": "code",
      "source": [
        "bst                   = xgb.XGBClassifier(max_depth = 7, \n",
        "                                              n_estimators = 200, \n",
        "                                              learning_rate = 0.05,\n",
        "                                              subsample = 0.9,\n",
        "                                              random_state = 12345,\n",
        "                                              reg_alpha = 1e-4, \n",
        "                                              reg_lambda = 1e-4, \n",
        "                                              booster = 'gbtree',\n",
        "                                              objective = 'binary:logistic')\n",
        "full_data = MinMaxScaler().fit_transform(dataset.samples.astype('float32'))\n",
        "full_labels = np.array([label_map[item] for item in dataset.sa.targets])[:,-1]\n",
        "bst.fit(full_data,full_labels)"
      ],
      "execution_count": 12,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "XGBClassifier(base_score=0.5, booster='gbtree', colsample_bylevel=1,\n",
              "       colsample_bytree=1, gamma=0, learning_rate=0.05, max_delta_step=0,\n",
              "       max_depth=7, min_child_weight=1, missing=None, n_estimators=200,\n",
              "       n_jobs=1, nthread=None, objective='binary:logistic',\n",
              "       random_state=12345, reg_alpha=0.0001, reg_lambda=0.0001,\n",
              "       scale_pos_weight=1, seed=None, silent=True, subsample=0.9)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 12
        }
      ]
    },
    {
      "metadata": {
        "id": "iuIm3QLIoKJ6",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "feature_importance_norm = (bst.feature_importances_ - bst.feature_importances_.mean()) / bst.feature_importances_.std()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "FW3EqiqD03Mg",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 574
        },
        "outputId": "25826a21-3e65-4a27-d87c-20b4cf52e542"
      },
      "cell_type": "code",
      "source": [
        "plt.plot(feature_importance_norm)\n",
        "plt.xlabel('Voxel Indices')\n",
        "plt.ylabel('Importance')"
      ],
      "execution_count": 14,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "Text(0,0.5,u'Importance')"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 14
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAwoAAAIbCAYAAABLxWmrAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nEwUiIiIiInLw/wGMd/DeTI7uLwAAAABJRU5ErkJggg==\n",
            "text/plain": [
              "<matplotlib.figure.Figure at 0x7f23bfcb6550>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "metadata": {
        "id": "VTjrNvia1D97",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 90
        },
        "outputId": "7adc9fb1-8398-4f3d-9737-690b6cf7ef00"
      },
      "cell_type": "code",
      "source": [
        "bst.score(full_data,full_labels)"
      ],
      "execution_count": 15,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python2.7/dist-packages/sklearn/preprocessing/label.py:151: DeprecationWarning: The truth value of an empty array is ambiguous. Returning False, but in future this will result in an error. Use `array.size > 0` to check that an array is not empty.\n",
            "  if diff:\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "1.0"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 15
        }
      ]
    },
    {
      "metadata": {
        "id": "RXi7C_FZ1XNV",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    }
  ]
}