Contest Brel

gistfile1.txt

{
 "metadata": {
  "name": "contest"
 },
 "nbformat": 3,
 "nbformat_minor": 0,
 "worksheets": [
  {
   "cells": [
    {
     "cell_type": "code",
     "collapsed": false,
     "input": "import pandas as pd",
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 73
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": "def dummify2(df, columns):\n    for column in columns:\n        dummies = pd.get_dummies(df[column], prefix=column)\n        df = df.drop(column, axis=1)\n        df = df.join(dummies)\n    return df",
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 74
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": "def convert_class(cls):\n    if cls.find('+') >=0:\n        return 1\n    else:\n        return 0",
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 75
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": "df = pd.read_csv('income_train.csv')",
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 76
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": "df.CLASS = df.CLASS.apply(convert_class)",
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 77
    },
    {
     "cell_type": "markdown",
     "metadata": {},
     "source": "\u0421\u043e\u0441\u0442\u0430\u0432\u0438\u043c \u0441\u043f\u0438\u0441\u043e\u043a \u043a\u0430\u0442\u0435\u0433\u043e\u0440\u0438\u0430\u043b\u044c\u043d\u044b\u0445 \u043a\u043e\u043b\u043e\u043d\u043e\u043a"
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": "categorial = []\ncnt = 0\nfor t in df.dtypes:\n    if (t == 'object'):\n       categorial.append(df.columns[cnt])\n    cnt += 1",
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 78
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": "cat_not_str = []\nfor column in df.columns:\n    if column not in categorial and column != 'CLASS':\n        desc = df.loc[:,[column]].describe()\n        if len(pd.unique(df.loc[:,[column]].values)) <= 7:\n            cat_not_str.append(column)\n            ",
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 79
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": "df2 = dummify2(df, categorial + cat_not_str)",
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 82
    },
    {
     "cell_type": "markdown",
     "metadata": {},
     "source": "\u041f\u043e\u0434\u0431\u0435\u0440\u0435\u043c \u043f\u0430\u0440\u0430\u043c\u0435\u0442\u0440 max_depth \u0434\u043b\u044f \u043e\u0431\u0443\u0447\u0430\u044e\u0449\u0435\u0433\u043e \u0434\u0435\u0440\u0435\u0432\u0430"
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": "from sklearn import tree\nfrom sklearn import cross_validation\nfrom sklearn.preprocessing import scale\ndef preidict_with_tree(max_depth):\n    X_train, X_test, Y_train, Y_test = cross_validation.train_test_split(scale(df2.drop('CLASS', axis=1).values), df2.CLASS, test_size=1/4)\n    clf = tree.DecisionTreeClassifier(max_depth=max_depth)\n    clf = clf.fit(X_train, Y_train)\n    Y_predicted = clf.predict_proba(X_test)\n    from sklearn.metrics import mean_absolute_error as MAE\n    e = MAE(Y_test, Y_predicted[:, 1])\n    print(\"Error: {0}\".format(e))",
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 239
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": "from sklearn.grid_search import GridSearchCV\nsearch = GridSearchCV(tree.DecisionTreeClassifier(), param_grid=[{'max_depth' : list(range(7, 25, 1))} ])",
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 305
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": "search.fit(X_train, Y_train)",
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "metadata": {},
       "output_type": "pyout",
       "prompt_number": 306,
       "text": "GridSearchCV(cv=None,\n       estimator=DecisionTreeClassifier(compute_importances=None, criterion='gini',\n            max_depth=None, max_features=None, min_density=None,\n            min_samples_leaf=1, min_samples_split=2, random_state=None,\n            splitter='best'),\n       fit_params={}, iid=True, loss_func=None, n_jobs=1,\n       param_grid=[{'max_depth': [7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24]}],\n       pre_dispatch='2*n_jobs', refit=True, score_func=None, scoring=None,\n       verbose=0)"
      }
     ],
     "prompt_number": 306
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": "search.best_params_",
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "metadata": {},
       "output_type": "pyout",
       "prompt_number": 307,
       "text": "{'max_depth': 8}"
      }
     ],
     "prompt_number": 307
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": "preidict_with_tree(15)\n# preidict_with_tree(15)\n# preidict_with_tree(17)",
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": "Error: 0.1781860220338418\n"
      }
     ],
     "prompt_number": 304
    },
    {
     "cell_type": "markdown",
     "metadata": {},
     "source": "GridSearchCV \u0433\u043e\u0432\u043e\u0440\u0438\u0442, \u0447\u0442\u043e 8, \u043d\u043e \u044f \u043f\u043e\u0442\u044b\u043a\u0430\u043b \u0440\u0443\u043a\u0430\u043c\u0438 \u0438 \u0443 \u043c\u0435\u043d\u044f \u043f\u043e\u043b\u0443\u0447\u0438\u043b\u043e\u0441\u044c, \u0447\u0442\u043e 15 \u043b\u0443\u0447\u0448\u0435 (WTF?)"
    },
    {
     "cell_type": "markdown",
     "metadata": {},
     "source": "\u041f\u043e\u0441\u043c\u043e\u0442\u0440\u0438\u043c, \u043b\u0443\u0447\u0448\u0435 \u043b\u0438 KNN"
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": "from sklearn.neighbors import KNeighborsClassifier as KNN\nfrom sklearn.grid_search import GridSearchCV\nsearch = GridSearchCV(KNN(weights='distance'),param_grid=[{'n_neighbors' : list(range(1, 25, 5))} ])\nsearch = search.fit(X_train, Y_train)",
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 83
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": "search.best_params_",
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "metadata": {},
       "output_type": "pyout",
       "prompt_number": 84,
       "text": "{'n_neighbors': 21}"
      }
     ],
     "prompt_number": 84
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": "search = GridSearchCV(KNN(weights='distance'),param_grid=[{'n_neighbors' : list(range(17, 25, 3))} ])\nsearch.fit(X_train, Y_train)",
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "metadata": {},
       "output_type": "pyout",
       "prompt_number": 85,
       "text": "GridSearchCV(cv=None,\n       estimator=KNeighborsClassifier(algorithm='auto', leaf_size=30, metric='minkowski',\n           n_neighbors=5, p=2, weights='distance'),\n       fit_params={}, iid=True, loss_func=None, n_jobs=1,\n       param_grid=[{'n_neighbors': [21, 23]}], pre_dispatch='2*n_jobs',\n       refit=True, score_func=None, scoring=None, verbose=0)"
      }
     ],
     "prompt_number": 85
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": "search.best_params_",
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "metadata": {},
       "output_type": "pyout",
       "prompt_number": 86,
       "text": "{'n_neighbors': 23}"
      }
     ],
     "prompt_number": 86
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": "search = GridSearchCV(KNN(weights='distance'),param_grid=[{'n_neighbors' : list(range(21, 23, 1))} ])\nsearch.fit(X_train, Y_train)",
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "metadata": {},
       "output_type": "pyout",
       "prompt_number": 87,
       "text": "GridSearchCV(cv=None,\n       estimator=KNeighborsClassifier(algorithm='auto', leaf_size=30, metric='minkowski',\n           n_neighbors=5, p=2, weights='distance'),\n       fit_params={}, iid=True, loss_func=None, n_jobs=1,\n       param_grid=[{'n_neighbors': [21, 22]}], pre_dispatch='2*n_jobs',\n       refit=True, score_func=None, scoring=None, verbose=0)"
      }
     ],
     "prompt_number": 87
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": "search.best_params_",
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "metadata": {},
       "output_type": "pyout",
       "prompt_number": 88,
       "text": "{'n_neighbors': 22}"
      }
     ],
     "prompt_number": 88
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": "from sklearn import cross_validation\ndef preidict_with_knn(nn):\n    X_train, X_test, Y_train, Y_test = cross_validation.train_test_split(scale(df2.drop('CLASS', axis=1).values), df2.CLASS, test_size=1/4)\n    clf = KNN(n_neighbors=nn, weights='distance')\n    clf = clf.fit(X_train, Y_train)\n    Y_predicted = clf.predict_proba(X_test)\n    from sklearn.metrics import mean_absolute_error as MAE\n    e = MAE(Y_test, Y_predicted[:, 1])\n    print(\"Error: {0}\".format(e))",
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 93
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": "preidict_with_knn(25)",
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": "Error: 0.2713978826288885\n"
      }
     ],
     "prompt_number": 94
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": "preidict_with_knn(22)",
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": "Error: 0.26926506682476214\n"
      }
     ],
     "prompt_number": 96
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": "preidict_with_knn(19)",
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": "Error: 0.2730720793439653\n"
      }
     ],
     "prompt_number": 99
    },
    {
     "cell_type": "markdown",
     "metadata": {},
     "source": "\u0417\u0434\u0435\u0441\u044c \u0443\u0436\u0435 \u0431\u043b\u0438\u0436\u0435 \u043a \u0438\u0441\u0442\u0438\u043d\u0435. \u041d\u043e \u0434\u0435\u0440\u0435\u0432\u044c\u044f \u043b\u0443\u0447\u0448\u0435. \u041f\u043e\u044d\u0442\u043e\u043c\u0443, \u043f\u0440\u0438\u043c\u0435\u043d\u044f\u0435\u043c \u0434\u0435\u0440\u0435\u0432\u044c\u044f."
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": "df_train = pd.read_csv('income_train.csv')",
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 61
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": "df_test = pd.read_csv('income_test.csv')",
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 62
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": "df_joined = pd.concat([df_train.drop('CLASS', axis=1), df_test.drop('ID', axis=1)]) ",
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 63
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": "def dummify3(df, columns):\n    for column in columns:\n        dummies = pd.get_dummies(df[column], prefix=column)\n        df = pd.concat([df, dummies], axis=1)\n        del df[column]\n    return df",
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 64
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": "X_joined = dummify3(df_joined, categorial + cat_not_str)",
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 65
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": "X_joined.shape",
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "metadata": {},
       "output_type": "pyout",
       "prompt_number": 66,
       "text": "(14800, 386)"
      }
     ],
     "prompt_number": 66
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": "X_joined = scale(X_joined)",
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 67
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": "X_train = X_joined[:len(df_train), :]\nY_train = df_train.CLASS.apply(convert_class).values.astype(np.float32)\nX_test = X_joined[len(df_train):, :]",
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 68
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": "clf = tree.DecisionTreeClassifier(max_depth=15)\nclf = clf.fit(X_train, Y_train)\n",
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 69
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": "Y_predicted = clf.predict_proba(X_test)",
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 70
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": "df_test['Prediction'] = Y_predicted[:, 1]\ndf_test.loc[:, ['ID', 'Prediction']].to_csv('prediction.csv', index=False)",
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stderr",
       "text": "/Library/Frameworks/Python.framework/Versions/3.3/lib/python3.3/site-packages/pandas/core/internals.py:1190: DeprecationWarning: using a non-integer number instead of an integer will result in an error in the future\n  values = values[:, slicer]\n/Library/Frameworks/Python.framework/Versions/3.3/lib/python3.3/site-packages/pandas/core/internals.py:552: DeprecationWarning: using a non-integer number instead of an integer will result in an error in the future\n  values = values[:, slicer]\n/Library/Frameworks/Python.framework/Versions/3.3/lib/python3.3/site-packages/pandas/core/index.py:624: DeprecationWarning: using a non-integer number instead of an integer will result in an error in the future\n  result = arr_idx[key]\n"
      }
     ],
     "prompt_number": 71
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": "",
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 71
    }
   ],
   "metadata": {}
  }
 ]
}