diff --git a/krasnodar_catboost_pipeline.ipynb b/krasnodar_catboost_pipeline.ipynb deleted file mode 100644 index 2672b27..0000000 --- a/krasnodar_catboost_pipeline.ipynb +++ /dev/null @@ -1 +0,0 @@ -{"nbformat":4,"nbformat_minor":0,"metadata":{"colab":{"provenance":[],"authorship_tag":"ABX9TyNZctONPU2EVkxxRRjxT6Z6"},"kernelspec":{"name":"python3","display_name":"Python 3"},"language_info":{"name":"python"}},"cells":[{"cell_type":"markdown","source":["# Import"],"metadata":{"id":"RpR8nK-X_hgw"}},{"cell_type":"code","execution_count":1,"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"id":"OsZspeG7-ig9","executionInfo":{"status":"ok","timestamp":1676664405476,"user_tz":480,"elapsed":18621,"user":{"displayName":"Радмир Зосимов","userId":"04742503366191314986"}},"outputId":"a02c974a-660e-45fb-a1d9-c54d8caa2c43"},"outputs":[{"output_type":"stream","name":"stdout","text":["\u001b[2K \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m575.9/575.9 KB\u001b[0m \u001b[31m11.9 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n","\u001b[2K \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m76.6/76.6 MB\u001b[0m \u001b[31m8.8 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n","\u001b[?25h"]}],"source":["!pip install shap catboost -qqq"]},{"cell_type":"code","source":["import pandas as pd\n","import numpy as np\n","import random\n","import os\n","import matplotlib.pyplot as plt\n","import seaborn as sns\n","from sklearn.metrics import mean_absolute_error\n","from catboost import CatBoostRegressor, Pool\n","from sklearn.model_selection import StratifiedKFold, StratifiedGroupKFold\n","from copy import deepcopy \n","import pickle\n","import sklearn.metrics\n","\n","import warnings\n","warnings.filterwarnings('ignore')"],"metadata":{"id":"RP5Hvy3U_B3k","executionInfo":{"status":"ok","timestamp":1676667073297,"user_tz":480,"elapsed":132,"user":{"displayName":"Радмир Зосимов","userId":"04742503366191314986"}}},"execution_count":57,"outputs":[]},{"cell_type":"code","source":["def set_seed(seed: int):\n"," \"\"\"Set a random seed for complete reproducibility.\"\"\"\n"," random.seed(seed)\n"," np.random.seed(seed)\n"," os.environ['PYTHONHASHSEED'] = str(seed)\n","\n","set_seed(0xBACED)"],"metadata":{"id":"fc5OTTsL_KKO","executionInfo":{"status":"ok","timestamp":1676664442131,"user_tz":480,"elapsed":160,"user":{"displayName":"Радмир Зосимов","userId":"04742503366191314986"}}},"execution_count":3,"outputs":[]},{"cell_type":"code","source":["from google.colab import drive\n","drive.mount('/content/drive')\n","%cd /content/drive/MyDrive/krasnodar_hack/\n","!ls"],"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"id":"WZdpl2tz_Qs7","executionInfo":{"status":"ok","timestamp":1676664495035,"user_tz":480,"elapsed":17468,"user":{"displayName":"Радмир Зосимов","userId":"04742503366191314986"}},"outputId":"aed1f89d-88bc-432b-d372-a191dcfb88cd"},"execution_count":4,"outputs":[{"output_type":"stream","name":"stdout","text":["Mounted at /content/drive\n","/content/drive/MyDrive/krasnodar_hack\n","krasnodar_catboost_pipeline.ipynb krasnodar_data.ipynb\n","krasnodar_data_final.csv\n"]}]},{"cell_type":"markdown","source":["# Modeling"],"metadata":{"id":"Evu6SGtcBszQ"}},{"cell_type":"code","source":["class RegressionModel:\n"," def __init__(self, data=None, agg_cols=None, agg_cols_double=None, cat_features=['community', 'liter_num', 'done_date', 'developer', 'floor'],\n"," target_col='price_per_m', n_splits=5, stratify_col='price_per_m', group_col='community', params={}, path_to_save='model_reg.pkl'):\n"," self.data = data.fillna(-1)\n"," self.target_col = target_col\n"," \n"," self.agg_cols = agg_cols\n"," self.agg_cols_double = agg_cols_double\n","\n"," self.cat_features = cat_features \n","\n"," self.params = params\n"," self.params['cat_features'] = self.cat_features\n"," self.path_to_save = path_to_save\n"," self.models = []\n","\n"," self.n_splits = n_splits\n"," self.stratify_col = stratify_col\n"," self.group_col = group_col\n"," if group_col:\n"," self.kfold = StratifiedGroupKFold(n_splits=self.n_splits, shuffle=True, random_state=0xBACED)\n"," else:\n"," self.kfold = StratifiedKFold(n_splits=self.n_splits, shuffle=True, random_state=0xBACED)\n","\n"," def make_model(self):\n"," return CatBoostRegressor(**self.params)\n","\n"," def get_overall_features(self, df_train, df_val=None):\n"," ret_dfs = []\n"," for df in [df_train, df_val]:\n"," df['date'] = pd.to_datetime(df['date'])\n","\n"," df['year'] = df['date'].dt.year\n"," df['month'] = df['date'].dt.month\n","\n"," df['price_per_m'] = np.log(df['price_per_m'])\n","\n"," ret_dfs.append(df)\n"," \n"," return ret_dfs\n","\n"," def split_data(self):\n"," if self.group_col:\n"," self.data[self.group_col] = self.data[self.group_col].astype(str)\n","\n"," splits = self.kfold.split(self.data, self.data[self.stratify_col], self.data[self.group_col])\n"," else:\n"," splits = self.kfold.split(self.data, self.data[self.stratify_col])\n","\n"," for fold, (train_index, val_index) in enumerate(splits):\n"," df_train = self.data.iloc[train_index]\n"," df_val = self.data.iloc[val_index]\n","\n"," df_train, df_val = self.get_overall_features(df_train, df_val)\n","\n"," self.num_features = df_train.columns[~df_train.columns.isin(self.cat_features+['price_per_m', 'date', 'to_group'])].tolist() \n","\n"," cols = self.num_features + self.cat_features\n"," X_train = df_train[cols]\n"," X_val = df_val[cols]\n"," y_val = np.log(df_val[self.target_col])\n"," y_train = np.log(df_train[self.target_col])\n","\n"," print(X_train.shape, X_val.shape)\n"," yield fold, (X_train, X_val, y_train, y_val)\n","\n"," def get_data(self, fold):\n"," for f, (X_train, X_val, y_train, y_val) in self.split_data():\n"," if f == fold:\n"," return X_train, X_val, y_train, y_val\n","\n"," def fit_model(self):\n"," for fold, (X_train, X_val, y_train, y_val) in self.split_data():\n"," print(f\"Training for fold {fold} Stratified by {self.stratify_col}\")\n"," print('-'*30)\n"," cat = self.make_model()\n","\n"," cat.fit(X_train, y_train,\n"," eval_set=(X_val, y_val))\n"," self.models.append(cat)\n","\n"," with open(self.path_to_save, 'wb') as f:\n"," pickle.dump(self.models, f) \n","\n"," metrics = self.evaluate()\n","\n"," return metrics\n","\n"," def predict(self, data):\n"," orig = deepcopy(data)\n"," if not self.models:\n"," print('Train or load models before prediction phase')\n","\n"," self.data, _ = self.get_overall_features(self.data)\n"," # X = merge_val_df(self.data, orig, self.agg_cols)\n","\n"," X = X[self.models[0].feature_names_]\n","\n"," preds = [model.predict(X) for model in self.models]\n"," preds = np.mean(np.array(preds), axis=0)\n"," return preds\n"," \n"," \n"," def evaluate(self):\n"," if not self.models:\n"," print('Train or load models before evaluation phase')\n","\n"," metrics = {'MAE': [], 'RMSE': [], 'MAPE': []}\n","\n"," for fold, (X_train, X_val, y_train, y_val) in self.split_data():\n"," y_pred = np.exp(self.models[fold].predict(X_val[self.models[fold].feature_names_]))\n"," y_val = np.exp(y_val)\n","\n"," mse = sklearn.metrics.mean_squared_error(y_val, y_pred)\n"," mae = sklearn.metrics.mean_absolute_error(y_val, y_pred)\n"," mape = 1-sklearn.metrics.mean_absolute_percentage_error(y_val, y_pred)\n"," rmse = np.sqrt(mse)\n","\n"," metrics['MAE'].append(mae)\n"," metrics['MAPE'].append(mape)\n"," metrics['RMSE'].append(rmse)\n","\n","\n"," print('### Overall metrics ###')\n"," for key, values in metrics.items():\n"," print(f'For {key}: {np.mean(values).round(4)}')\n"," print(len('### Overall metrics ###') * '#')\n"," print()\n","\n"," return metrics\n","\n"," def load_model(self, path):\n"," with open(path, 'rb') as f:\n"," self.models = pickle.load(f) "],"metadata":{"id":"l2T42ahmA3Je","executionInfo":{"status":"ok","timestamp":1676668374033,"user_tz":480,"elapsed":172,"user":{"displayName":"Радмир Зосимов","userId":"04742503366191314986"}}},"execution_count":96,"outputs":[]},{"cell_type":"code","source":["df = pd.read_csv('krasnodar_data_final.csv')\n","df['date'] = pd.to_datetime(df['date'])\n","df['year'] = df['date'].dt.year\n","df['to_group'] = df['community'] + '_' + df['liter_num']\n","\n","regression_params = dict(\n"," iterations=1000, \n"," learning_rate=0.05,\n"," cat_features=None,\n"," verbose=100,\n"," early_stopping_rounds=500,\n"," loss_function='RMSE',\n"," eval_metric='MAE',\n"," task_type='CPU',\n"," # bootstrap_type='Poisson'\n",")\n","\n","model = RegressionModel(\n"," data=df, \n"," agg_cols=[],\n"," target_col='price_per_m', \n"," n_splits=5, \n"," stratify_col='price_per_m', \n"," group_col='to_group',\n"," params=regression_params, \n",")"],"metadata":{"id":"p3VEyhN7GsCG","executionInfo":{"status":"ok","timestamp":1676668374539,"user_tz":480,"elapsed":2,"user":{"displayName":"Радмир Зосимов","userId":"04742503366191314986"}}},"execution_count":97,"outputs":[]},{"cell_type":"code","source":["metrics = model.fit_model()"],"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"id":"TD2srfAHHUFK","executionInfo":{"status":"ok","timestamp":1676668391607,"user_tz":480,"elapsed":15546,"user":{"displayName":"Радмир Зосимов","userId":"04742503366191314986"}},"outputId":"c38db97a-4ebb-4b51-e15e-d3039299bd8a"},"execution_count":98,"outputs":[{"output_type":"stream","name":"stdout","text":["(1178, 9) (111, 9)\n","Training for fold 0 Stratified by price_per_m\n","------------------------------\n","0:\tlearn: 0.0183076\ttest: 0.0153729\tbest: 0.0153729 (0)\ttotal: 3.57ms\tremaining: 3.57s\n","100:\tlearn: 0.0039936\ttest: 0.0052346\tbest: 0.0052346 (100)\ttotal: 311ms\tremaining: 2.76s\n","200:\tlearn: 0.0030294\ttest: 0.0048681\tbest: 0.0048681 (200)\ttotal: 623ms\tremaining: 2.48s\n","300:\tlearn: 0.0025245\ttest: 0.0049576\tbest: 0.0048681 (200)\ttotal: 1.18s\tremaining: 2.73s\n","400:\tlearn: 0.0021508\ttest: 0.0048496\tbest: 0.0048471 (390)\ttotal: 1.94s\tremaining: 2.9s\n","500:\tlearn: 0.0018880\ttest: 0.0047996\tbest: 0.0047642 (485)\ttotal: 2.76s\tremaining: 2.75s\n","600:\tlearn: 0.0016759\ttest: 0.0048038\tbest: 0.0047642 (485)\ttotal: 3.58s\tremaining: 2.38s\n","700:\tlearn: 0.0015244\ttest: 0.0048103\tbest: 0.0047642 (485)\ttotal: 4.42s\tremaining: 1.89s\n","800:\tlearn: 0.0014093\ttest: 0.0047755\tbest: 0.0047577 (768)\ttotal: 4.91s\tremaining: 1.22s\n","900:\tlearn: 0.0013101\ttest: 0.0047158\tbest: 0.0047095 (882)\ttotal: 5.21s\tremaining: 573ms\n","999:\tlearn: 0.0012238\ttest: 0.0046975\tbest: 0.0046771 (943)\ttotal: 5.5s\tremaining: 0us\n","\n","bestTest = 0.004677130625\n","bestIteration = 943\n","\n","Shrink model to first 944 iterations.\n","(1019, 9) (270, 9)\n","Training for fold 1 Stratified by price_per_m\n","------------------------------\n","0:\tlearn: 0.0189938\ttest: 0.0152091\tbest: 0.0152091 (0)\ttotal: 5.49ms\tremaining: 5.49s\n","100:\tlearn: 0.0039330\ttest: 0.0054161\tbest: 0.0053876 (93)\ttotal: 297ms\tremaining: 2.64s\n","200:\tlearn: 0.0029411\ttest: 0.0058529\tbest: 0.0053344 (111)\ttotal: 575ms\tremaining: 2.29s\n","300:\tlearn: 0.0023228\ttest: 0.0058582\tbest: 0.0053344 (111)\ttotal: 842ms\tremaining: 1.96s\n","400:\tlearn: 0.0019468\ttest: 0.0061122\tbest: 0.0053344 (111)\ttotal: 1.11s\tremaining: 1.66s\n","500:\tlearn: 0.0017086\ttest: 0.0063125\tbest: 0.0053344 (111)\ttotal: 1.4s\tremaining: 1.4s\n","600:\tlearn: 0.0015209\ttest: 0.0063740\tbest: 0.0053344 (111)\ttotal: 1.7s\tremaining: 1.13s\n","Stopped by overfitting detector (500 iterations wait)\n","\n","bestTest = 0.005334417346\n","bestIteration = 111\n","\n","Shrink model to first 112 iterations.\n","(859, 9) (430, 9)\n","Training for fold 2 Stratified by price_per_m\n","------------------------------\n","0:\tlearn: 0.0130150\ttest: 0.0292111\tbest: 0.0292111 (0)\ttotal: 3.83ms\tremaining: 3.83s\n","100:\tlearn: 0.0034040\ttest: 0.0293268\tbest: 0.0281034 (16)\ttotal: 243ms\tremaining: 2.16s\n","200:\tlearn: 0.0026375\ttest: 0.0308591\tbest: 0.0281034 (16)\ttotal: 475ms\tremaining: 1.89s\n","300:\tlearn: 0.0021564\ttest: 0.0306285\tbest: 0.0281034 (16)\ttotal: 737ms\tremaining: 1.71s\n","400:\tlearn: 0.0017409\ttest: 0.0308015\tbest: 0.0281034 (16)\ttotal: 999ms\tremaining: 1.49s\n","500:\tlearn: 0.0015137\ttest: 0.0306867\tbest: 0.0281034 (16)\ttotal: 1.25s\tremaining: 1.25s\n","Stopped by overfitting detector (500 iterations wait)\n","\n","bestTest = 0.02810337199\n","bestIteration = 16\n","\n","Shrink model to first 17 iterations.\n","(1090, 9) (199, 9)\n","Training for fold 3 Stratified by price_per_m\n","------------------------------\n","0:\tlearn: 0.0187981\ttest: 0.0140491\tbest: 0.0140491 (0)\ttotal: 3.3ms\tremaining: 3.3s\n","100:\tlearn: 0.0040912\ttest: 0.0050387\tbest: 0.0047003 (74)\ttotal: 292ms\tremaining: 2.6s\n","200:\tlearn: 0.0031168\ttest: 0.0054071\tbest: 0.0047003 (74)\ttotal: 581ms\tremaining: 2.31s\n","300:\tlearn: 0.0025175\ttest: 0.0055865\tbest: 0.0047003 (74)\ttotal: 866ms\tremaining: 2.01s\n","400:\tlearn: 0.0021220\ttest: 0.0056781\tbest: 0.0047003 (74)\ttotal: 1.15s\tremaining: 1.72s\n","500:\tlearn: 0.0017955\ttest: 0.0057054\tbest: 0.0047003 (74)\ttotal: 1.44s\tremaining: 1.44s\n","Stopped by overfitting detector (500 iterations wait)\n","\n","bestTest = 0.004700271121\n","bestIteration = 74\n","\n","Shrink model to first 75 iterations.\n","(1010, 9) (279, 9)\n","Training for fold 4 Stratified by price_per_m\n","------------------------------\n","0:\tlearn: 0.0188050\ttest: 0.0150863\tbest: 0.0150863 (0)\ttotal: 3.61ms\tremaining: 3.6s\n","100:\tlearn: 0.0038160\ttest: 0.0098701\tbest: 0.0093704 (45)\ttotal: 298ms\tremaining: 2.65s\n","200:\tlearn: 0.0028236\ttest: 0.0101171\tbest: 0.0093704 (45)\ttotal: 575ms\tremaining: 2.29s\n","300:\tlearn: 0.0022027\ttest: 0.0102054\tbest: 0.0093704 (45)\ttotal: 855ms\tremaining: 1.98s\n","400:\tlearn: 0.0018134\ttest: 0.0102956\tbest: 0.0093704 (45)\ttotal: 1.14s\tremaining: 1.7s\n","500:\tlearn: 0.0015809\ttest: 0.0102020\tbest: 0.0093704 (45)\ttotal: 1.41s\tremaining: 1.4s\n","Stopped by overfitting detector (500 iterations wait)\n","\n","bestTest = 0.009370404766\n","bestIteration = 45\n","\n","Shrink model to first 46 iterations.\n","(1178, 9) (111, 9)\n","(1019, 9) (270, 9)\n","(859, 9) (430, 9)\n","(1090, 9) (199, 9)\n","(1010, 9) (279, 9)\n","### Overall metrics ###\n","For MAE: 0.1234\n","For RMSE: 0.1568\n","For MAPE: 0.9897\n","#######################\n","\n"]}]},{"cell_type":"code","source":["metrics"],"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"id":"cRw5_XX1HWPX","executionInfo":{"status":"ok","timestamp":1676668391608,"user_tz":480,"elapsed":24,"user":{"displayName":"Радмир Зосимов","userId":"04742503366191314986"}},"outputId":"d13b74b1-4c93-4b16-f5ab-5c617625aad7"},"execution_count":99,"outputs":[{"output_type":"execute_result","data":{"text/plain":["{'MAE': [0.05448504858283975,\n"," 0.062343977097046,\n"," 0.33544665440639787,\n"," 0.0546641594066206,\n"," 0.11008324406993113],\n"," 'RMSE': [0.06927946822197725,\n"," 0.07957608603990474,\n"," 0.42269099504684476,\n"," 0.07031775478421469,\n"," 0.14190741201572488],\n"," 'MAPE': [0.9953292000899124,\n"," 0.9946654467229404,\n"," 0.972440092666718,\n"," 0.9952906658465586,\n"," 0.9906571162083089]}"]},"metadata":{},"execution_count":99}]},{"cell_type":"code","source":["model.load_model('model_reg.pkl')"],"metadata":{"id":"cwVeqmTpM4GV","executionInfo":{"status":"ok","timestamp":1676668391608,"user_tz":480,"elapsed":19,"user":{"displayName":"Радмир Зосимов","userId":"04742503366191314986"}}},"execution_count":100,"outputs":[]},{"cell_type":"code","source":["X_train, X_val, y_train, y_val = model.get_data(0)"],"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"id":"a2J_TlXmM9DK","executionInfo":{"status":"ok","timestamp":1676668041671,"user_tz":480,"elapsed":180,"user":{"displayName":"Радмир Зосимов","userId":"04742503366191314986"}},"outputId":"98f3b201-4706-4ce5-cda4-5e4060cbdd44"},"execution_count":89,"outputs":[{"output_type":"stream","name":"stdout","text":["(1178, 9) (111, 9)\n"]}]},{"cell_type":"code","source":["import shap\n","shap.initjs()\n","\n","explainer = shap.TreeExplainer(model.models[0])\n","shap_values = explainer.shap_values(X_train)\n","\n","shap.summary_plot(shap_values, X_train)"],"metadata":{"colab":{"base_uri":"https://localhost:8080/","height":404},"id":"i1ro6UhnKY6j","executionInfo":{"status":"ok","timestamp":1676668079151,"user_tz":480,"elapsed":1920,"user":{"displayName":"Радмир Зосимов","userId":"04742503366191314986"}},"outputId":"0a39b4f3-0e47-486f-a72a-c9c6a1fb0526"},"execution_count":91,"outputs":[{"output_type":"display_data","data":{"text/plain":[""],"text/html":["
"]},"metadata":{}},{"output_type":"display_data","data":{"text/plain":["
"],"image/png":"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\n"},"metadata":{"needs_background":"light"}}]},{"cell_type":"code","source":[],"metadata":{"id":"UiFF7J9YNFln"},"execution_count":null,"outputs":[]}]} \ No newline at end of file