ExtraTrees.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "from hyperopt import fmin, tpe, hp\n",
    "import ipynb.fs.full.utils as utils\n",
    "import ipynb.fs.full.features as features\n",
    "import ipynb.fs.full.features_distancias as f_distancias\n",
    "\n",
    "\n",
    "df_train = pd.read_csv('./data/train_filtrado.csv')\n",
    "# Para usarse con el submit a Kaggle\n",
    "df_eval = pd.read_csv('./data/test.csv')\n",
    "\n",
    "df_train, df_eval = features.features_de_csvs(df_train, df_eval)\n",
    "\n",
    "df_train_idf = pd.read_csv('./data/train_idf.csv')\n",
    "df_eval_idf = pd.read_csv('./data/test_idf.csv')\n",
    "\n",
    "df_train = pd.merge(df_train, df_train_idf, on= 'id', how= 'left')\n",
    "df_eval = pd.merge(df_eval, df_eval_idf, on= 'id', how= 'left')\n",
    "\n",
    "df_train, df_test = utils.dividir_df_testeo(df_train, test_size=0.25)\n",
    "\n",
    "df_test = features.llenar_nulls(df_test, hgb_mean=True, df_fill=df_train)\n",
    "df_train = features.llenar_nulls(df_train, hgb_mean=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "# df_train = df_train.sample(frac=1).reset_index(drop=True)\n",
    "\n",
    "df_test_f = features.features_independientes_precio(df_test)\n",
    "df_test_f = features.features_dependientes_precio(df_test_f, df_train)\n",
    "\n",
    "df_train_f = features.features_independientes_precio(df_train)\n",
    "df_train_f = features.features_dependientes_precio(df_train_f, df_train)\n",
    "\n",
    "df_test_f, cols_tipodepropiedad_ohe = features.columna_a_ohe(df_test_f, 'tipodepropiedad', N=100, df_aux=df_train, devolver_cols=True)\n",
    "df_test_f, cols_provincia_ohe = features.columna_a_ohe(df_test_f, 'provincia', N=100, df_aux=df_train, devolver_cols=True)\n",
    "df_test_f, cols_zona_ohe = features.columna_a_ohe(df_test_f, 'zona', df_aux=df_train_f, devolver_cols=True)\n",
    "\n",
    "df_train_f = features.columna_a_ohe(df_train_f, 'tipodepropiedad', N=100, df_aux=df_test)\n",
    "df_train_f = features.columna_a_ohe(df_train_f, 'provincia', N=100, df_aux=df_test)\n",
    "df_train_f = features.columna_a_ohe(df_train_f, 'zona', df_aux=df_test_f)\n",
    "\n",
    "\n",
    "df_train_f['fecha'] = pd.to_datetime(df_train_f['fecha']).astype(int)\n",
    "df_test_f['fecha'] = pd.to_datetime(df_test_f['fecha']).astype(int)\n",
    "\n",
    "# df_train_f = df_train_f.sample(frac=1).reset_index(drop=True)\n",
    "\n",
    "df_train_f = f_distancias.feature_distancias(df_train_f)\n",
    "df_test_f = f_distancias.feature_distancias(df_test_f, df_train_f)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## GridSearchCV"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.ensemble import ExtraTreesRegressor\n",
    "\n",
    "from sklearn.model_selection import GridSearchCV\n",
    "\n",
    "features = ['habitaciones', 'garages','banos','antiguedad', 'metroscubiertos',  'metrostotales','lat_norm', \n",
    "           'lng_norm', 'gimnasio', 'usosmultiples', 'piscina']\n",
    "\n",
    "features_test = ['top_provincia', 'promedio_precio_ciudad', 'anio', 'promedio_id_zona', \n",
    "                 'promedio_precio_tipo_propiedad', 'count_id_zona', 'count_ciudad', 'puntaje', \n",
    "               'count_tipo_propiedad_ciudad', 'promedio_precio_tipo_propiedad_ciudad_gen','count_id_zona'\n",
    "           'dias_desde_datos','meses_desde_datos','porcentaje_metros','distancia_ciudad_centrica', 'puntaje', 'distancia_centro_mexico']\n",
    "\n",
    "features += features_test\n",
    "\n",
    "extratrees_params = { \n",
    "    \"n_estimators\": [10],\n",
    "    \"max_features\": [\"auto\", \"sqrt\", \"log2\"],\n",
    "    \"min_samples_split\": [2,4,8],\n",
    "    \"min_impurity_decrease\": [0, 0.1, 0.2],\n",
    "    \"bootstrap\": [False]\n",
    "}\n",
    "\n",
    "x_train, x_test, y_train, y_test = utils.dividir_dataset(df_train_f, 'precio', features, test_size=1)\n",
    "\n",
    "rs_cv = GridSearchCV(estimator=ExtraTreesRegressor(), \n",
    "                           param_grid=extratrees_params, \n",
    "                           cv=4, scoring='neg_mean_absolute_error', verbose=1)\n",
    "\n",
    "rs_cv.fit(x_train, y_train)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Evaluación modelo final"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=1)]: Using backend SequentialBackend with 1 concurrent workers.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "building tree 1 of 20\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=1)]: Done   1 out of   1 | elapsed:    0.3s remaining:    0.0s\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "building tree 2 of 20\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=1)]: Done   2 out of   2 | elapsed:    0.7s remaining:    0.0s\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "building tree 3 of 20\n",
      "building tree 4 of 20\n",
      "building tree 5 of 20\n",
      "building tree 6 of 20\n",
      "building tree 7 of 20\n",
      "building tree 8 of 20\n",
      "building tree 9 of 20\n",
      "building tree 10 of 20\n",
      "building tree 11 of 20\n",
      "building tree 12 of 20\n",
      "building tree 13 of 20\n",
      "building tree 14 of 20\n",
      "building tree 15 of 20\n",
      "building tree 16 of 20\n",
      "building tree 17 of 20\n",
      "building tree 18 of 20\n",
      "building tree 19 of 20\n",
      "building tree 20 of 20\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=1)]: Done  20 out of  20 | elapsed:    8.0s finished\n",
      "[Parallel(n_jobs=1)]: Using backend SequentialBackend with 1 concurrent workers.\n",
      "[Parallel(n_jobs=1)]: Done   1 out of   1 | elapsed:    0.2s remaining:    0.0s\n",
      "[Parallel(n_jobs=1)]: Done   2 out of   2 | elapsed:    0.3s remaining:    0.0s\n",
      "[Parallel(n_jobs=1)]: Done  20 out of  20 | elapsed:    1.8s finished\n",
      "[Parallel(n_jobs=1)]: Using backend SequentialBackend with 1 concurrent workers.\n",
      "[Parallel(n_jobs=1)]: Done   1 out of   1 | elapsed:    0.0s remaining:    0.0s\n",
      "[Parallel(n_jobs=1)]: Done   2 out of   2 | elapsed:    0.0s remaining:    0.0s\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "MAE AdaBoost (train): 142320.43568\n",
      "MAE AdaBoost (eval): 531897.23870\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=1)]: Done  20 out of  20 | elapsed:    0.6s finished\n"
     ]
    }
   ],
   "source": [
    "from sklearn.ensemble import ExtraTreesRegressor\n",
    "\n",
    "\n",
    "params = {'criterion': 'mse',\n",
    " 'max_features': 'sqrt',\n",
    " 'min_samples_split': 4}\n",
    "\n",
    "extratrees = ExtraTreesRegressor(n_estimators=20, verbose=3, **params)\n",
    "\n",
    "features = ['habitaciones', 'garages','banos','antiguedad', 'metroscubiertos',  'metrostotales','lat_norm', \n",
    "           'lng_norm', 'gimnasio', 'usosmultiples', 'piscina']\n",
    "\n",
    "features_test = ['top_provincia', 'promedio_precio_ciudad', 'anio', 'promedio_id_zona', \n",
    "                 'promedio_precio_tipo_propiedad', 'count_id_zona', 'count_ciudad', 'puntaje', \n",
    "               'count_tipo_propiedad_ciudad', 'promedio_precio_tipo_propiedad_ciudad_gen','count_id_zona'\n",
    "           'dias_desde_datos','meses_desde_datos','porcentaje_metros','distancia_ciudad_centrica', \n",
    "                'distancia_centro_mexico', 'distancia_ciudad_cara', 'label']\n",
    "\n",
    "features += features_test\n",
    "\n",
    "x_train = utils.filtrar_features(df_train_f, features)\n",
    "y_train = df_train_f['precio']\n",
    "x_eval = utils.filtrar_features(df_test_f, features)\n",
    "y_eval = df_test_f['precio']\n",
    "\n",
    "extratrees.fit(x_train, y_train)\n",
    "\n",
    "y_pred_train = extratrees.predict(x_train)\n",
    "y_pred_eval = extratrees.predict(x_eval)\n",
    "\n",
    "mae_train = utils.MAE(y_train, y_pred_train)\n",
    "mae_eval = utils.MAE(y_eval, y_pred_eval)\n",
    "\n",
    "print(f\"MAE AdaBoost (train): {mae_train:.5f}\")\n",
    "print(f\"MAE AdaBoost (eval): {mae_eval:.5f}\")"
   ]
  }
 ],
 "metadata": {
  "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": "3.7.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}