{ "cells": [ { "cell_type": "markdown", "metadata": { "_cell_guid": "02f999f1-9b67-a535-a483-abcf95c81c5d" }, "source": [ "What combination of location, time, and advertisers are likely to receive clicks?" ] }, { "cell_type": "markdown", "metadata": { "_cell_guid": "75617bb3-ab5a-7fe8-c65b-f7ea014024ec" }, "source": [ "filedir = '../input/'\n", "import pandas as pd\n", "import matplotlib.pyplot as plt\n", "import numpy as np\n", "import datetime\n", "import time\n", "from sklearn import tree\n", "from sklearn import linear_model\n", "from sklearn.model_selection import train_test_split\n", "from sklearn.model_selection import cross_val_score\n", "from sklearn import ensemble\n", "from sklearn import svm\n", "from sklearn import metrics\n", "from sklearn.model_selection import learning_curve\n", "from sklearn.model_selection import ShuffleSplit\n", "\n", "N_ROWS = 1000000\n", "\n", "'''\n", "#load documents csv's\n", "documents_categories_df = pd.read_csv(filedir+'documents_categories.csv')\n", "documents_entities_df = pd.read_csv(filedir+'documents_entities.csv')\n", "documents_meta_df = pd.read_csv(filedir+'documents_meta.csv')\n", "document_topics_df = pd.read_csv(filedir+'documents_topics.csv')\n", "'''\n", "\n", "#load events, clicks_train, promoted_content and page_views_sample csv's\n", "events_df = pd.read_csv(filedir+'events.csv', nrows=N_ROWS)\n", "promoted_content_df = pd.read_csv(filedir+'promoted_content.csv', nrows=N_ROWS)\n", "promoted_content_df = promoted_content_df.rename(columns={'document_id': 'ad_document_id'})\n", "clicks_train_df = pd.read_csv(filedir+'clicks_train.csv', nrows=N_ROWS)\n", "page_views_sample_df = pd.read_csv(filedir+'page_views_sample.csv', usecols=['uuid','document_id', 'timestamp', 'traffic_source'], nrows=N_ROWS)\n", "\n", "\n", "#merge events and clicks_train on display_id value\n", "#in order to find what ad was clicked for each event/display_id\n", "#this links platform and geo_location data to each ad that was clicked\n", "events_x_clicks_train_df = pd.merge(events_df, clicks_train_df, on='display_id', how='outer').dropna()\n", "#print(events_x_clicks_train_df)\n", "\n", "\n", "#convert geo_location to int data\n", "events_x_clicks_train_df['geo_location'] = events_x_clicks_train_df['geo_location'].replace('[^0-9]', '', regex=True).replace('', np.nan, regex=True)\n", "pd.to_numeric(events_x_clicks_train_df['geo_location'])\n", "\n", "\n", "#merge previous with page_views on uuid and document_id\n", "#in order to find traffic source for each event/display_id\n", "all_available_data_for_each_event = pd.merge(events_x_clicks_train_df, page_views_sample_df, on=['uuid', 'document_id','timestamp'])\n", "all_available_data_for_each_event = pd.merge(all_available_data_for_each_event, promoted_content_df, on='ad_id', how='outer').dropna()\n", "all_available_data_for_each_event = all_available_data_for_each_event.sort_values(by='display_id')\n", "#print(all_available_data_for_each_event)\n", "\n", "X = all_available_data_for_each_event[['display_id', 'document_id', 'timestamp', 'platform', 'geo_location', 'ad_id', 'traffic_source', 'ad_document_id', 'campaign_id', 'advertiser_id']]\n", "y = all_available_data_for_each_event[['clicked']]\n", "\n", "\n", "#rectify timestamp column and convert to hour of day clicked\n", "pd.to_numeric(X['timestamp'])\n", "X['true_time'] = X['timestamp']+1465876799998\n", "X['date'] = pd.to_datetime(X['true_time'], unit='ms')\n", "X['hour'] = X['date'].apply(lambda x: x.strftime('%H'))\n", "X = X.drop('date', axis=1)\n", "X = X.drop('timestamp', axis=1)\n", "X = X.drop('true_time', axis=1)\n", "\n", "\n", "#get platform & traffic source columns\n", "X['platform'] = X['platform'].astype(str).convert_objects(convert_numeric=True)\n", "X['traffic_source'] = X['traffic_source'].astype(str).convert_objects(convert_numeric=True)\n", "X['traffic_source'] = X['traffic_source'].astype(int)\n", "\n", "#print(X)\n", "#print(y)\n", "'''\n", "all_document_ids = X.document_id.unique()\n", "all_geo_locations = X.geo_location.unique()\n", "all_ad_ids = X.ad_id.unique()\n", "all_ad_document_ids = X.ad_document_id.unique()\n", "all_campaign_ids = X.campaign_id.unique()\n", "all_advertiser_ids = X.advertiser_id.unique()\n", "print(len(all_geo_locations)+len(all_advertiser_ids))\n", "'''\n", "# Get one hot encoding of columns geo_location, platform, hour, advertiser_id, and traffic_source\n", "# Drop column as it is now encoded, then drop irrelevant columns\n", "X = X.join(pd.get_dummies(X['platform'], prefix='pl_'))\n", "X = X.drop('platform', axis=1)\n", "X = X.join(pd.get_dummies(X['geo_location'], prefix='ge_'))\n", "X = X.drop('geo_location', axis=1)\n", "X = X.join(pd.get_dummies(X['advertiser_id'], prefix='ad_'))\n", "X = X.drop('advertiser_id', axis=1)\n", "X = X.join(pd.get_dummies(X['hour'], prefix='hr_'))\n", "X = X.drop('hour', axis=1)\n", "X = X.join(pd.get_dummies(X['traffic_source'], prefix='ts_'))\n", "X = X.drop('traffic_source', axis=1)\n", "\n", "#X = X.join(pd.get_dummies(X['campaign_id']))\n", "X = X.drop('campaign_id', axis=1)\n", "X = X.drop('ad_id', axis=1)\n", "X = X.drop('document_id', axis=1)\n", "X = X.drop('ad_document_id', axis=1)\n", "X = X.drop('display_id', axis=1)\n", "\n", "print(X)\n", "print(y)\n", "\n", "\n", "#sklearn stuff\n", "\n", "#clf = tree.DecisionTreeClassifier() \n", "#0.78\n", "\n", "#clf = linear_model.LogisticRegression(solver='sag', multi_class='ovr') \n", "#0.82\n", "\n", "#clf = linear_model.LogisticRegressionCV(Cs=100, fit_intercept=True, cv=None, \n", "# dual=False, penalty='l2', \n", "# solver='liblinear', tol=0.0001, max_iter=1000, refit=True, multi_class='ovr')\n", "#0.82\n", "\n", "#clf = linear_model.SGDClassifier() \n", "#0.77\n", "\n", "clf = ensemble.RandomForestClassifier(n_estimators=50, criterion='gini', max_depth=None, \n", " min_samples_split=20, min_samples_leaf=1, min_weight_fraction_leaf=0.0, \n", " max_features='auto', max_leaf_nodes=None, min_impurity_split=1e-02, bootstrap=True, \n", " class_weight=None)\n", "#0.82\n", "\n", "#tree.export_graphviz(clf, out_file='tree.dot') \n", "\n", "#clf = svm.SVC(C=1.0, cache_size=200, class_weight=None, coef0=0.0,\n", "# decision_function_shape=None, degree=3, gamma='auto', kernel='rbf',\n", "# max_iter=-1, probability=False, random_state=None, shrinking=True,\n", "# tol=0.001, verbose=False)\n", "#0.82\n", "\n", "#clf = svm.LinearSVC(penalty='l1', loss='squared_hinge', dual=False, \n", "# tol=0.00001, C=1.0, multi_class='crammer_singer', fit_intercept=False, \n", "# intercept_scaling=1, max_iter=2000)\n", "#0.81\n", "\n", "#clf = svm.NuSVC(nu=0.5, kernel='rbf', degree=3, gamma='auto', coef0=0.1, shrinking=True, \n", "# probability=False, tol=0.0001, cache_size=20,\n", "# max_iter=1000, decision_function_shape='ovr', random_state=None)\n", "#0.60\n", "\n", "#clf = linear_model.PassiveAggressiveClassifier(C=1.0, fit_intercept=True, n_iter=5, \n", "# shuffle=True, verbose=0, loss='hinge', n_jobs=1, random_state=None, \n", "# warm_start=False, class_weight=None)\n", "#0.81\n", "\n", "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=2)\n", "clf.fit(X_train.as_matrix(), y_train.as_matrix())\n", "print(clf.score(X_test, y_test))\n", "np.set_printoptions(threshold=np.nan)\n", "#print(clf.predict(X_test))\n", "#print(y_test)\n", "print(metrics.f1_score(y_test, clf.predict(X_test)))\n", "\n", "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.3, random_state=3)\n", "clf.fit(X_train.as_matrix(), y_train.as_matrix())\n", "print(clf.score(X_test, y_test))\n", "print(metrics.f1_score(y_test, clf.predict(X_test)))\n", "\n", "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.4, random_state=4)\n", "clf.fit(X_train.as_matrix(), y_train.as_matrix())\n", "print(clf.score(X_test, y_test))\n", "print(metrics.f1_score(y_test, clf.predict(X_test)))\n", "\n", "\n", "def plot_learning_curve(estimator, title, X, y, ylim=None, cv=None, n_jobs=1, train_sizes=np.linspace(.1, 1.0, 5)):\n", " \"\"\"\n", " Generate a simple plot of the test and training learning curve.\n", "\n", " Parameters\n", " ----------\n", " estimator : object type that implements the \"fit\" and \"predict\" methods\n", " An object of that type which is cloned for each validation.\n", "\n", " title : string\n", " Title for the chart.\n", "\n", " X : array-like, shape (n_samples, n_features)\n", " Training vector, where n_samples is the number of samples and\n", " n_features is the number of features.\n", "\n", " y : array-like, shape (n_samples) or (n_samples, n_features), optional\n", " Target relative to X for classification or regression;\n", " None for unsupervised learning.\n", "\n", " ylim : tuple, shape (ymin, ymax), optional\n", " Defines minimum and maximum yvalues plotted.\n", "\n", " cv : int, cross-validation generator or an iterable, optional\n", " Determines the cross-validation splitting strategy.\n", " Possible inputs for cv are:\n", " - None, to use the default 3-fold cross-validation,\n", " - integer, to specify the number of folds.\n", " - An object to be used as a cross-validation generator.\n", " - An iterable yielding train/test splits.\n", "\n", " For integer/None inputs, if ``y`` is binary or multiclass,\n", " :class:`StratifiedKFold` used. If the estimator is not a classifier\n", " or if ``y`` is neither binary nor multiclass, :class:`KFold` is used.\n", "\n", " Refer :ref:`User Guide ` for the various\n", " cross-validators that can be used here.\n", "\n", " n_jobs : integer, optional\n", " Number of jobs to run in parallel (default 1).\n", " \"\"\"\n", " plt.figure()\n", " plt.title(title)\n", " if ylim is not None:\n", " plt.ylim(*ylim)\n", " plt.xlabel(\"Training examples\")\n", " plt.ylabel(\"Score\")\n", " train_sizes, train_scores, test_scores = learning_curve(estimator, X, y, cv=cv, n_jobs=n_jobs, train_sizes=train_sizes)\n", " train_scores_mean = np.mean(train_scores, axis=1)\n", " train_scores_std = np.std(train_scores, axis=1)\n", " test_scores_mean = np.mean(test_scores, axis=1)\n", " test_scores_std = np.std(test_scores, axis=1)\n", " plt.grid()\n", "\n", " plt.fill_between(train_sizes, train_scores_mean - train_scores_std, train_scores_mean + train_scores_std, alpha=0.1, color=\"r\")\n", " plt.fill_between(train_sizes, test_scores_mean - test_scores_std, test_scores_mean + test_scores_std, alpha=0.1, color=\"g\")\n", " plt.plot(train_sizes, train_scores_mean, 'o-', color=\"r\", label=\"Training score\")\n", " plt.plot(train_sizes, test_scores_mean, 'o-', color=\"g\", label=\"Cross-validation score\")\n", " plt.legend(loc=\"best\")\n", " plt.show()\n", " return plt\n", "\n", "title = \"Learning Curves (Random Forest)\"\n", "# Cross validation with 100 iterations to get smoother mean test and train\n", "# score curves, each time with 20% data randomly selected as a validation set.\n", "cvx = ShuffleSplit(n_splits=100, test_size=0.2, random_state=0)\n", "y = y.as_matrix().ravel()\n", "plot_learning_curve(clf, title, X.as_matrix(), y, ylim=(0.7, 1.01), cv=cvx, n_jobs=4)\n", "\n", "#title = \"Learning Curves (SVM, RBF kernel, $\\gamma=0.001$)\"\n", "# SVC is more expensive so we do a lower number of CV iterations:\n", "#cv = ShuffleSplit(n_splits=10, test_size=0.2, random_state=0)\n", "#estimator = SVC(gamma=0.001)\n", "#plot_learning_curve(estimator, title, X, y, (0.7, 1.01), cv=cv, n_jobs=4)" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "_cell_guid": "7a1c7365-2822-0b32-2417-bea2c6b96850" }, "outputs": [], "source": [ "I seem to get an accuracy of about 0.82 no matter which algorithm I use." ] } ], "metadata": { "_change_revision": 0, "_is_fork": 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": "3.5.2" } }, "nbformat": 4, "nbformat_minor": 0 }