{ "cells": [ { "cell_type": "markdown", "metadata": { "_cell_guid": "ec35aa40-97b0-526b-68d4-202c8df37a80" }, "source": [ "## Introduction ##\n", "Hello! This is my first Kaggle notebook. I hope you like it. I'm from Argentina and I want to get some experience while I have some fun with March Madness. So let's start." ] }, { "cell_type": "code", "execution_count": null, "metadata": { "_cell_guid": "ce39e796-2e2e-39e0-39ec-4f77dd148c2e" }, "outputs": [], "source": [ "#Added by me\n", "import numpy as np\n", "import pandas as pd\n", "import random\n", "import statsmodels.api as sm\n", "import matplotlib.pyplot as plt\n", "from patsy import dmatrices\n", "from sklearn.linear_model import LogisticRegression\n", "from sklearn import metrics\n", "from sklearn import preprocessing\n", "\n", "# This Python 3 environment comes with many helpful analytics libraries installed\n", "# It is defined by the kaggle/python docker image: https://github.com/kaggle/docker-python\n", "# For example, here's several helpful packages to load in \n", "\n", "import numpy as np # linear algebra\n", "import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)\n", "\n", "#Agregados m\u00edos\n", "from keras.wrappers.scikit_learn import KerasRegressor\n", "from sklearn.model_selection import cross_val_score\n", "from sklearn.ensemble import RandomForestClassifier\n", "from sklearn.model_selection import KFold\n", "from sklearn.neighbors import KNeighborsClassifier\n", "from sklearn.preprocessing import StandardScaler\n", "from sklearn.pipeline import Pipeline\n", "from keras.models import Sequential\n", "from keras.layers import Dense\n", "\n", "# Input data files are available in the \"../input/\" directory.\n", "# For example, running this (by clicking run or pressing Shift+Enter) will list the files in the input directory\n", "\n", "from subprocess import check_output\n", "print(check_output([\"ls\", \"../input\"]).decode(\"utf8\"))\n", "\n", "# Any results you write to the current directory are saved as output." ] }, { "cell_type": "markdown", "metadata": { "_cell_guid": "57b7d6e1-734b-3e34-49c9-f577417a4808" }, "source": [ "## Processing data ##\n", "### Data comparison ###" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "_cell_guid": "9db33d3b-0088-ed8e-1674-bc14dbe163ff" }, "outputs": [], "source": [ "tourney_data=pd.read_csv(\"../input/TourneyDetailedResults.csv\")\n", "tourney_data.shape" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "_cell_guid": "06224a13-5098-42c7-3d37-85042b684dff" }, "outputs": [], "source": [ "%matplotlib inline" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "_cell_guid": "1e545e3d-2d58-d96d-bc96-de030a4d4416" }, "outputs": [], "source": [ "tourney_data=tourney_data.groupby(['Season']).mean()\n", "\n", "tourney_data.plot(y='Wscore')\n", "tourney_data.plot(y='Wfga')" ] }, { "cell_type": "markdown", "metadata": { "_cell_guid": "db9c46b7-caf7-2f86-663d-5a9c5b097be7" }, "source": [ "# Feature Engineering #" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "_cell_guid": "d54f9650-b76d-9385-955d-393ce35ba597" }, "outputs": [], "source": [ "detailed_data=pd.read_csv(\"../input/TourneyDetailedResults.csv\")\n", "\n", "training_data = pd.DataFrame()\n", "training_data[[\"Season\",\"team1\", \"team2\"]] =detailed_data[[\"Season\",\"Wteam\", \"Lteam\"]].copy()\n", "training_data[\"pred\"] = 1\n", "\n", "for index, row in training_data.iterrows():\n", " if random.randint(0,1)==1:\n", " temp=row['team1']\n", " training_data.set_value(index,'team1',row['team2'])\n", " training_data.set_value(index,'team2',temp)\n", " training_data.set_value(index,'pred',0)\n", "\n", "def loadInTraining(name,data):\n", " new_training_data=pd.merge(training_data,data,how=\"inner\",left_on=[\"Season\",\"team1\"], right_on=[\"Season\",\"team\"])\n", " new_training_data=new_training_data.drop(\"team\",axis=1)\n", " new_training_data=new_training_data.rename(columns={name: name+\"1\"})\n", " \n", " new_training_data=pd.merge(new_training_data,data,how=\"inner\",left_on=[\"Season\",\"team2\"], right_on=[\"Season\",\"team\"])\n", " new_training_data=new_training_data.drop(\"team\",axis=1)\n", " new_training_data=new_training_data.rename(columns={name: name+\"2\"})\n", " \n", " return new_training_data" ] }, { "cell_type": "markdown", "metadata": { "_cell_guid": "fa96adf3-aa00-82a5-45c9-959f05754456" }, "source": [ "**PPG - Points Per Game**" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "_cell_guid": "e0831f56-2ef7-4f6a-06c6-ce2d47b482f5" }, "outputs": [], "source": [ "season_data=pd.read_csv(\"../input/RegularSeasonDetailedResults.csv\")\n", "\n", "ppg_data_a=pd.DataFrame()\n", "ppg_data_a[[\"Season\",\"team\",\"score\"]]=season_data[[\"Season\",\"Wteam\",\"Wscore\"]]\n", "ppg_data_b=pd.DataFrame()\n", "ppg_data_b[[\"Season\",\"team\",\"score\"]]=season_data[[\"Season\",\"Lteam\",\"Lscore\"]]\n", "\n", "ppg_data=pd.concat((ppg_data_a,ppg_data_b),axis=0)\n", "ppg_data=ppg_data.groupby([\"Season\",\"team\"])[\"score\"].agg(['sum','count']).reset_index()\n", "ppg_data[\"PPG\"]=ppg_data[\"sum\"]/ppg_data[\"count\"]\n", "\n", "ppg_data=ppg_data[[\"Season\",\"team\",\"PPG\"]]\n", "training_data=loadInTraining(\"PPG\",ppg_data)\n", "training_data.head()\n" ] }, { "cell_type": "markdown", "metadata": { "_cell_guid": "fae51698-c4fe-9963-4e76-ec82fc1ceee5" }, "source": [ "**RPG - Received Per Game**" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "_cell_guid": "f69a99d2-8228-c19e-1f8d-f95c0632d4bc" }, "outputs": [], "source": [ "season_data=pd.read_csv(\"../input/RegularSeasonDetailedResults.csv\")\n", "\n", "rpg_data_a=pd.DataFrame()\n", "rpg_data_a[[\"Season\",\"team\",\"score\"]]=season_data[[\"Season\",\"Wteam\",\"Lscore\"]]\n", "rpg_data_b=pd.DataFrame()\n", "rpg_data_b[[\"Season\",\"team\",\"score\"]]=season_data[[\"Season\",\"Lteam\",\"Wscore\"]]\n", "\n", "rpg_data=pd.concat((rpg_data_a,rpg_data_b),axis=0)\n", "rpg_data=rpg_data.groupby([\"Season\",\"team\"])[\"score\"].agg(['sum','count']).reset_index()\n", "rpg_data[\"RPG\"]=rpg_data[\"sum\"]/rpg_data[\"count\"]\n", "\n", "rpg_data=rpg_data[[\"Season\",\"team\",\"RPG\"]]\n", "training_data=loadInTraining(\"RPG\",rpg_data)\n", "training_data.head()\n" ] }, { "cell_type": "markdown", "metadata": { "_cell_guid": "ef736d77-02dc-ae86-f4f3-51fbd418c5ef" }, "source": [ "**REC - Record (W/L)**" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "_cell_guid": "0d6f1f41-1d22-b72d-b16d-a1d8931a10d1" }, "outputs": [], "source": [ "season_data=pd.read_csv(\"../input/RegularSeasonDetailedResults.csv\")\n", "\n", "rec_data_w=pd.DataFrame()\n", "rec_data_w[[\"Season\",\"team\",\"W\"]]=season_data[[\"Season\",\"Wteam\",\"Wscore\"]]\n", "rec_data_l=pd.DataFrame()\n", "rec_data_l[[\"Season\",\"team\",\"L\"]]=season_data[[\"Season\",\"Lteam\",\"Lscore\"]]\n", "\n", "rec_data_w=rec_data_w.groupby([\"Season\",\"team\"])[\"W\"].count().reset_index()\n", "rec_data_l=rec_data_l.groupby([\"Season\",\"team\"])[\"L\"].count().reset_index()\n", "rec_data=pd.merge(rec_data_w,rec_data_l,how=\"outer\",on=[\"Season\",\"team\"])\n", "rec_data=rec_data.fillna(0)\n", "rec_data[\"REC\"]=rec_data[\"W\"]/(rec_data[\"W\"]+rec_data[\"L\"])\n", "\n", "rec_data=rec_data[[\"Season\",\"team\",\"REC\"]]\n", "\n", "training_data=loadInTraining(\"REC\",rec_data)\n", "training_data.head()" ] }, { "cell_type": "markdown", "metadata": { "_cell_guid": "e148ef4d-18b8-e8d8-ca39-e057260915a6" }, "source": [ "**eFG% - Effective Field Goal Percentage**\n", "\n", "Four Factors - Dean Oliver" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "_cell_guid": "9ecc2cc7-4bc0-94b3-708d-7d971c77ea5e" }, "outputs": [], "source": [ "season_data=pd.read_csv(\"../input/RegularSeasonDetailedResults.csv\")\n", "\n", "efg_data_w=pd.DataFrame()\n", "efg_data_w[[\"Season\",\"team\",\"fga\",\"fg\",\"3p\"]]=season_data[[\"Season\",\"Wteam\",\"Wfga\",\"Wfgm\",\"Wfgm3\"]]\n", "efg_data_l=pd.DataFrame()\n", "efg_data_l[[\"Season\",\"team\",\"fga\",\"fg\",\"3p\"]]=season_data[[\"Season\",\"Lteam\",\"Lfga\",\"Lfgm\",\"Lfgm3\"]]\n", "\n", "efg_data=pd.concat((efg_data_w,efg_data_l),axis=0)\n", "efg_data=efg_data.groupby([\"Season\",\"team\"]).sum().reset_index()\n", "efg_data[\"eFG%\"]=(efg_data[\"fg\"]+0.5*efg_data[\"3p\"])/efg_data[\"fga\"]\n", "\n", "efg_data=efg_data[[\"Season\",\"team\",\"eFG%\"]]\n", "training_data=loadInTraining(\"eFG%\",efg_data)\n", "training_data.head()" ] }, { "cell_type": "markdown", "metadata": { "_cell_guid": "559cdb52-243b-c2fb-4fb0-e6f036355770" }, "source": [ "**dFG% - Effective Defensive Field Goal Percentage**" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "_cell_guid": "2aaf75eb-8662-71dd-1bb3-66513edb191a" }, "outputs": [], "source": [ "season_data=pd.read_csv(\"../input/RegularSeasonDetailedResults.csv\")\n", "\n", "efg_data_w=pd.DataFrame()\n", "efg_data_w[[\"Season\",\"team\",\"fga\",\"fg\",\"3p\"]]=season_data[[\"Season\",\"Wteam\",\"Lfga\",\"Lfgm\",\"Lfgm3\"]]\n", "efg_data_l=pd.DataFrame()\n", "efg_data_l[[\"Season\",\"team\",\"fga\",\"fg\",\"3p\"]]=season_data[[\"Season\",\"Lteam\",\"Wfga\",\"Wfgm\",\"Wfgm3\"]]\n", "\n", "efg_data=pd.concat((efg_data_w,efg_data_l),axis=0)\n", "efg_data=efg_data.groupby([\"Season\",\"team\"]).sum().reset_index()\n", "efg_data[\"dFG%\"]=(efg_data[\"fg\"]+0.5*efg_data[\"3p\"])/efg_data[\"fga\"]\n", "\n", "efg_data=efg_data[[\"Season\",\"team\",\"dFG%\"]]\n", "training_data=loadInTraining(\"dFG%\",efg_data)\n", "training_data.head()" ] }, { "cell_type": "markdown", "metadata": { "_cell_guid": "8fcab66e-a867-c896-0326-2e576e04a46b" }, "source": [ "**TOV% - Turnover Percentage**" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "_cell_guid": "ca18c3bd-9da0-1e35-403d-d34b404d2e6c" }, "outputs": [], "source": [ "season_data=pd.read_csv(\"../input/RegularSeasonDetailedResults.csv\")\n", "\n", "tov_data_w=pd.DataFrame()\n", "tov_data_w[[\"Season\",\"team\",\"fga\",\"fta\",\"tov\"]]=season_data[[\"Season\",\"Wteam\",\"Wfga\",\"Wfta\",\"Wto\"]]\n", "tov_data_l=pd.DataFrame()\n", "tov_data_l[[\"Season\",\"team\",\"fga\",\"fta\",\"tov\"]]=season_data[[\"Season\",\"Lteam\",\"Lfga\",\"Lfta\",\"Lto\"]]\n", "\n", "tov_data=pd.concat((tov_data_w,tov_data_l),axis=0)\n", "tov_data=tov_data.groupby([\"Season\",\"team\"]).sum().reset_index()\n", "tov_data[\"TOV%\"]=tov_data[\"tov\"]/(tov_data[\"fga\"]+0.44*tov_data[\"fta\"]+tov_data[\"tov\"])\n", "\n", "tov_data=tov_data[[\"Season\",\"team\",\"TOV%\"]]\n", "training_data=loadInTraining(\"TOV%\",tov_data)\n", "training_data.head()" ] }, { "cell_type": "markdown", "metadata": { "_cell_guid": "71a0105c-ca02-4b89-2e0e-9bde4781de39" }, "source": [ "**dTO% - Defensive Turnover Percentage**" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "_cell_guid": "8d12dc63-f8a3-1627-1dfb-c0567844639a" }, "outputs": [], "source": [ "season_data=pd.read_csv(\"../input/RegularSeasonDetailedResults.csv\")\n", "\n", "tov_data_w=pd.DataFrame()\n", "tov_data_w[[\"Season\",\"team\",\"fga\",\"fta\",\"tov\"]]=season_data[[\"Season\",\"Wteam\",\"Lfga\",\"Lfta\",\"Lto\"]]\n", "tov_data_l=pd.DataFrame()\n", "tov_data_l[[\"Season\",\"team\",\"fga\",\"fta\",\"tov\"]]=season_data[[\"Season\",\"Lteam\",\"Wfga\",\"Wfta\",\"Wto\"]]\n", "\n", "tov_data=pd.concat((tov_data_w,tov_data_l),axis=0)\n", "tov_data=tov_data.groupby([\"Season\",\"team\"]).sum().reset_index()\n", "tov_data[\"dTO%\"]=tov_data[\"tov\"]/(tov_data[\"fga\"]+0.44*tov_data[\"fta\"]+tov_data[\"tov\"])\n", "\n", "tov_data=tov_data[[\"Season\",\"team\",\"dTO%\"]]\n", "training_data=loadInTraining(\"dTO%\",tov_data)\n", "training_data.head()" ] }, { "cell_type": "markdown", "metadata": { "_cell_guid": "2313651c-f318-aa4c-467b-f93374c3367c" }, "source": [ "**ORB% - Offensive Rebound Percentage**" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "_cell_guid": "7113f043-fb32-9489-207e-ec59dbafa992" }, "outputs": [], "source": [ "season_data=pd.read_csv(\"../input/RegularSeasonDetailedResults.csv\")\n", "\n", "orb_data_w=pd.DataFrame()\n", "orb_data_w[[\"Season\",\"team\",\"or\",\"odr\"]]=season_data[[\"Season\",\"Wteam\",\"Wor\",\"Ldr\"]]\n", "orb_data_l=pd.DataFrame()\n", "orb_data_l[[\"Season\",\"team\",\"or\",\"odr\"]]=season_data[[\"Season\",\"Lteam\",\"Lor\",\"Wdr\"]]\n", "\n", "orb_data=pd.concat((orb_data_w,orb_data_l),axis=0)\n", "orb_data=orb_data.groupby([\"Season\",\"team\"]).sum().reset_index()\n", "orb_data[\"ORB%\"]=orb_data[\"or\"]/(orb_data[\"or\"]+orb_data[\"odr\"])\n", "\n", "orb_data=orb_data[[\"Season\",\"team\",\"ORB%\"]]\n", "training_data=loadInTraining(\"ORB%\",orb_data)\n", "training_data.head()" ] }, { "cell_type": "markdown", "metadata": { "_cell_guid": "3b7ac165-1cdf-ad57-a583-14c93368c9f6" }, "source": [ "**DRB% - Defensive Rebound Percentage**" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "_cell_guid": "56d5a609-f3f8-bad5-8d15-b4cba0be56d4" }, "outputs": [], "source": [ "season_data=pd.read_csv(\"../input/RegularSeasonDetailedResults.csv\")\n", "\n", "drb_data_w=pd.DataFrame()\n", "drb_data_w[[\"Season\",\"team\",\"dr\",\"oor\"]]=season_data[[\"Season\",\"Wteam\",\"Wdr\",\"Lor\"]]\n", "drb_data_l=pd.DataFrame()\n", "drb_data_l[[\"Season\",\"team\",\"dr\",\"oor\"]]=season_data[[\"Season\",\"Lteam\",\"Ldr\",\"Wor\"]]\n", "\n", "drb_data=pd.concat((drb_data_w,drb_data_l),axis=0)\n", "drb_data=drb_data.groupby([\"Season\",\"team\"]).sum().reset_index()\n", "drb_data[\"DRB%\"]=drb_data[\"dr\"]/(drb_data[\"dr\"]+drb_data[\"oor\"])\n", "\n", "drb_data=drb_data[[\"Season\",\"team\",\"DRB%\"]]\n", "training_data=loadInTraining(\"DRB%\",drb_data)\n", "training_data.head()" ] }, { "cell_type": "markdown", "metadata": { "_cell_guid": "36c7fa35-04cd-ff6b-598f-a36227a13e7a" }, "source": [ "**FTR - Free Throw Rate**" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "_cell_guid": "78e07857-9410-cfb4-2b80-b6cd9c6ff21e" }, "outputs": [], "source": [ "season_data=pd.read_csv(\"../input/RegularSeasonDetailedResults.csv\")\n", "\n", "ftr_data_w=pd.DataFrame()\n", "ftr_data_w[[\"Season\",\"team\",\"ft\",\"fga\"]]=season_data[[\"Season\",\"Wteam\",\"Wftm\",\"Wfga\"]]\n", "ftr_data_l=pd.DataFrame()\n", "ftr_data_l[[\"Season\",\"team\",\"ft\",\"fga\"]]=season_data[[\"Season\",\"Lteam\",\"Lftm\",\"Lfga\"]]\n", "\n", "ftr_data=pd.concat((ftr_data_w,ftr_data_l),axis=0)\n", "ftr_data=ftr_data.groupby([\"Season\",\"team\"]).sum().reset_index()\n", "ftr_data[\"FTR\"]=ftr_data[\"ft\"]/ftr_data[\"fga\"]\n", "\n", "ftr_data=ftr_data[[\"Season\",\"team\",\"FTR\"]]\n", "training_data=loadInTraining(\"FTR\",ftr_data)\n", "training_data.head()\n" ] }, { "cell_type": "markdown", "metadata": { "_cell_guid": "f9cff636-1c9a-41c5-00eb-560f2e4eda90" }, "source": [ "**SdN - Seed Number**" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "_cell_guid": "707befe9-d758-43c6-63a4-2808083b24d0" }, "outputs": [], "source": [ "seed_data=pd.read_csv(\"../input/TourneySeeds.csv\")\n", "\n", "seed_data[\"SdN\"]=[int(str(x)[1:3]) for x in seed_data[\"Seed\"]]\n", "seed_data[\"team\"]=seed_data[\"Team\"]\n", "\n", "seed_data=seed_data[[\"Season\",\"team\",\"SdN\"]]\n", "training_data=loadInTraining(\"SdN\",seed_data)\n", "training_data.head()" ] }, { "cell_type": "markdown", "metadata": { "_cell_guid": "e89dde96-3d06-a5b4-dfc3-c7a288038a18" }, "source": [ "**dFT - Defensive Free Throw Rate**" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "_cell_guid": "08bf48c7-18d8-d4a1-1490-01aaa1a6729e" }, "outputs": [], "source": [ "season_data=pd.read_csv(\"../input/RegularSeasonDetailedResults.csv\")\n", "\n", "ftr_data_w=pd.DataFrame()\n", "ftr_data_w[[\"Season\",\"team\",\"ft\",\"fga\"]]=season_data[[\"Season\",\"Wteam\",\"Lftm\",\"Lfga\"]]\n", "ftr_data_l=pd.DataFrame()\n", "ftr_data_l[[\"Season\",\"team\",\"ft\",\"fga\"]]=season_data[[\"Season\",\"Lteam\",\"Wftm\",\"Wfga\"]]\n", "\n", "ftr_data=pd.concat((ftr_data_w,ftr_data_l),axis=0)\n", "ftr_data=ftr_data.groupby([\"Season\",\"team\"]).sum().reset_index()\n", "ftr_data[\"dFT\"]=ftr_data[\"ft\"]/ftr_data[\"fga\"]\n", "\n", "ftr_data=ftr_data[[\"Season\",\"team\",\"dFT\"]]\n", "training_data=loadInTraining(\"dFT\",ftr_data)\n", "training_data.head()" ] }, { "cell_type": "markdown", "metadata": { "_cell_guid": "4fde03a5-b403-e287-93dc-90417c2efab6" }, "source": [ "## Model sketch ##" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "_cell_guid": "7482edd0-2056-0705-bff6-39d14bbba801" }, "outputs": [], "source": [ "X = pd.DataFrame()\n", "X = training_data.loc[training_data[\"Season\"]<2013].copy()\n", "X = X.reindex_axis(sorted(X.columns), axis=1)\n", "\n", "\n", "Xval=pd.DataFrame()\n", "Xval = X[:323]\n", "X=X[323:]\n", "\n", "Xtest=pd.DataFrame()\n", "Xtest=training_data.loc[training_data[\"Season\"]>=2013].copy()\n", "Xtest = Xtest.reindex_axis(sorted(Xtest.columns), axis=1)\n", "\n", "y = pd.DataFrame()\n", "y[\"pred\"]=X[\"pred\"]\n", "\n", "yval = pd.DataFrame()\n", "yval[\"pred\"]=Xval[\"pred\"]\n", "\n", "ytest = pd.DataFrame()\n", "ytest[\"pred\"]=Xtest[\"pred\"]\n", "yseeds=ytest.copy()\n", "\n", "X = X.drop([\"Season\",\"team1\",\"team2\",\"pred\"],axis=1)\n", "Xval = Xval.drop([\"Season\",\"team1\",\"team2\",\"pred\"],axis=1)\n", "Xseeds=Xtest.copy().drop([\"pred\"],axis=1)\n", "Xtest = Xtest.drop([\"Season\",\"team1\",\"team2\",\"pred\"],axis=1)\n", "\n", "X.head()" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "_cell_guid": "d9131a93-29a4-a846-b119-c083c71c41d5" }, "outputs": [], "source": [ "def normalize(df):\n", " result = df.copy()\n", " for feature_name in df.columns:\n", " max_value = df[feature_name].max()\n", " min_value = df[feature_name].min()\n", " if max_value>1:\n", " result[feature_name] = (df[feature_name] - min_value) / (max_value - min_value)\n", " return result\n", "\n", "X=normalize(X)\n", "Xval=normalize(Xval)\n", "Xtest=normalize(Xtest)\n", "Xval.head()" ] }, { "cell_type": "markdown", "metadata": { "_cell_guid": "b2556f9e-223d-a860-c577-7bb915a80409" }, "source": [ "**Feature reduction**" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "_cell_guid": "acc2bcc3-e0b0-84c2-2180-f9edca235463" }, "outputs": [], "source": [ "Xforest=X.copy()\n", "yforest=y.copy()\n", "forest=RandomForestClassifier(n_estimators=30)\n", "forest.fit(Xforest,yforest)\n", "\n", "table=pd.concat((pd.DataFrame(Xforest.iloc[:, 0:].columns, columns = ['variable']), \n", " pd.DataFrame(forest.feature_importances_, columns = ['importance'])), \n", " axis = 1).sort_values(by='importance', ascending = False)[:20].reset_index()\n", "table=table.drop(\"index\",axis=1)\n", "table=table.groupby([table.variable.str[:3]]).sum().reset_index().sort_values(by='importance',ascending=False).reset_index()\n", "table=table.drop(\"index\",axis=1)\n", "table" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "_cell_guid": "914341e8-aa96-f3f1-ef15-17713ba9d00d" }, "outputs": [], "source": [ "def reduction(X,features,cant):\n", " delete=features.shape[0]-cant\n", " features=features.sort_values(by=\"importance\",ascending=True).reset_index().drop(\"index\",axis=1)\n", " for i in range(0,delete):\n", " pattern=features.loc[i][\"variable\"]\n", " col=[col for col in X.columns if str(pattern) in col]\n", " X=X.drop(col[0],1)\n", " X=X.drop(col[1],1)\n", " return X\n", " \n", "reduct_to=7\n", "#X=reduction(X,table,reduct_to)\n", "#Xval=reduction(Xval,table,reduct_to)\n", "#Xtest=reduction(Xtest,table,reduct_to)" ] }, { "cell_type": "markdown", "metadata": { "_cell_guid": "d680e110-cab5-5d2f-2fc1-f20958b4ad98" }, "source": [ "**Knn**" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "_cell_guid": "1e7fbbd0-031b-1f24-571a-6d5ce8f96f43" }, "outputs": [], "source": [ "n_neighbors=135\n", "knn=KNeighborsClassifier(n_neighbors=n_neighbors, p=2)\n", "knn=knn.fit(X,y)\n", "\n", "knn.score(X,y)" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "_cell_guid": "d1242963-3154-0daa-8bec-5f2de61f2f05" }, "outputs": [], "source": [ "predict=knn.predict(Xval)" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "_cell_guid": "38419127-d64d-cb02-3895-0f5c79fdc99b" }, "outputs": [], "source": [ "accuracy=metrics.accuracy_score(yval, predict)\n", "accuracy\n", "#print (metrics.classification_report(yval, predict))" ] }, { "cell_type": "markdown", "metadata": { "_cell_guid": "4ba5218e-3099-c69c-4305-60f43a1b1635" }, "source": [ "## Test set ##" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "_cell_guid": "234d52f1-922b-6d24-27d0-b0ede72b21a7" }, "outputs": [], "source": [ "predict_test=knn.predict(Xtest)\n", "probs = knn.predict_proba(Xtest)" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "_cell_guid": "df02350f-a5ff-63e2-ef76-bc3c37a3b0b8" }, "outputs": [], "source": [ "xtr=pd.concat((probs,predic_test,ytest),axis=0)\n", "xtr=xtr.loc[xtr[]]\n", "\n", "#accuracy_xtreme=metrics.accuracy_score(ytest,probs)\n", "#accuracy_xtreme" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "_cell_guid": "162d2f50-dac4-8322-1a0c-420e54f4f00c" }, "outputs": [], "source": [ "accuracy_test=metrics.accuracy_score(ytest, predict_test)\n", "accuracy_test\n", "\n", "#print (metrics.classification_report(ytest, predict))" ] }, { "cell_type": "markdown", "metadata": { "_cell_guid": "d72c9665-f5c9-0349-e26f-97a2ad38c70e" }, "source": [ "#Create submission file#" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "_cell_guid": "10a93929-fc90-f81f-c32d-d54f4118cf41" }, "outputs": [], "source": [ "bestScore=pd.Series.from_csv(\"../working/BestCase.csv\")\n", "if bestScore.values[0]>accuracy_test:\n", " grabar=False\n", "else:\n", " grabar=True\n", " scoreTest=pd.Series(accuracy_test,index=[\"accuracy\"])\n", " scoreTest.to_csv(\"../working/BestCase.csv\")" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "_cell_guid": "321689e0-9b10-e354-d621-ea609553edcb" }, "outputs": [], "source": [ "if grabar:\n", " tourney_seeds=pd.read_csv(\"../input/TourneySeeds.csv\")\n", " tourney_seeds.head()\n", " model=knn\n", " submission=pd.DataFrame()\n", " Xseeds.head()" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "_cell_guid": "612d4893-d646-615a-9196-c5e1d02303f6" }, "outputs": [], "source": [ "def getValues(seeds,team,year):\n", " seeds_season=seeds.copy().reset_index().drop(\"index\",1)\n", " seeds_season=seeds_season.loc[seeds_season[\"Season\"]==year]\n", " seeds_season=seeds_season.loc[(seeds_season[\"team1\"]==team) | (seeds_season[\"team2\"]==team)].reset_index()\n", " if seeds_season.loc[0][\"team1\"]==team:\n", " values=seeds_season.loc[0].drop([\"Season\"])\n", " values=values.drop([col for col in X.columns if str(\"2\") in col])\n", " else:\n", " values=seeds_season.loc[0].drop([\"Season\"])\n", " values=values.drop([col for col in X.columns if str(\"1\") in col])\n", " values=values.drop([\"index\",\"team1\",\"team2\"])\n", " return values" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "_cell_guid": "0543d699-646f-9003-f0fb-e23c07bd1758" }, "outputs": [], "source": [ "def getDataFrame(one,two):\n", " one=pd.DataFrame({'variable':one.index, 'value':one.values})\n", " one=one.set_index([\"variable\"]).transpose()\n", " one=one.rename(columns= lambda x: str(x)[:-1])\n", " one=one.rename(columns= lambda x: str(x+\"1\"))\n", " \n", " two=pd.DataFrame({'variable':two.index, 'value':two.values})\n", " two=two.set_index([\"variable\"]).transpose()\n", " two=two.rename(columns= lambda x: str(x)[:-1])\n", " two=two.rename(columns= lambda x: str(x+\"2\"))\n", "\n", " frame=pd.concat((one,two),axis=1)\n", " frame = frame.reindex_axis(sorted(frame.columns), axis=1)\n", " frame=frame.reset_index().drop(\"index\",axis=1)\n", " frame= frame.rename_axis(None,axis=1)\n", " return frame" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "_cell_guid": "a0802e43-9f5e-2c80-dca8-7309933b2560" }, "outputs": [], "source": [ "#2013\n", "if grabar:\n", " seeds_2013=tourney_seeds.loc[tourney_seeds[\"Season\"] == 2013]\n", " seeds_2013=seeds_2013.sort_values(\"Team\", ascending=[1])\n", " seeds_2013.head()\n", "\n", " submission_2013=pd.DataFrame()\n", " submission_2013 = pd.DataFrame(columns=('id', 'pred'))\n", " for index in range(0,seeds_2013.shape[0]):\n", " for index2 in range(0,seeds_2013.shape[0]):\n", " if seeds_2013.iloc[index][\"Team\"]