<\/span><\/h3>\n\n\n\nimport pandas as pd\nimport numpy as np\nfrom sklearn.model_selection import train_test_split, TimeSeriesSplit, cross_val_score\nfrom sklearn.linear_model import LinearRegression\nfrom sklearn.metrics import mean_squared_error\nfrom xgboost import XGBRegressor\nfrom lightgbm import LGBMRegressor\nfrom sklearn.ensemble import RandomForestRegressor\nimport shap\n\n# \u521b\u5efa\u793a\u4f8b\u65f6\u95f4\u5e8f\u5217\u6570\u636e\ndates = pd.date_range(start='2023-01-01', end='2025-06-30', freq='D')\nnp.random.seed(42)\nsales_data = np.cumsum(np.random.normal(loc=100, scale=20, size=len(dates)))\n\ndata = {\n 'order_date': dates,\n 'product_id': np.random.choice(['P{}'.format(i) for i in range(1, 101)], len(dates)),\n 'category_code': np.random.choice(['C{}'.format(i) for i in range(1, 11)], len(dates)),\n 'amount': sales_data,\n 'quantity': np.random.randint(1, 5, size=len(dates)),\n 'promotion': np.random.choice([True, False], len(dates))\n}\n\ndf = pd.DataFrame(data)\n\n# \u8bbe\u7f6e\u65f6\u95f4\u4e3a\u7d22\u5f15\ndf.set_index('order_date', inplace=True)\n\nprint(\"\u539f\u59cb\u6570\u636e:\")\nprint(df.head())<\/code><\/pre>\n\n\n\n<\/span>1. \u51c6\u5907\u7528\u4e8e\u9884\u6d4b\u7684\u7279\u5f81\u5de5\u7a0b\uff08\u65f6\u95f4\u3001\u4fc3\u9500\u3001\u8282\u5047\u65e5\u7b49\uff09<\/span><\/h3>\n\n\n\n# \u6dfb\u52a0\u65f6\u95f4\u7279\u5f81\ndf['year'] = df.index.year\ndf['month'] = df.index.month\ndf['day_of_week'] = df.index.dayofweek\n\n# \u6dfb\u52a0\u8282\u5047\u65e5\u6807\u5fd7\nholidays = pd.to_datetime([\n '2023-01-01', '2023-02-22', '2023-04-05', '2023-05-01', '2023-10-01',\n '2024-01-01', '2024-02-10', '2024-04-04', '2024-05-01', '2024-10-01',\n '2025-01-01', '2025-02-19', '2025-04-04', '2025-05-01', '2025-10-01'\n])\n\ndf['is_holiday'] = df.index.isin(holidays).astype(int)\n\n# \u5c06\u4fc3\u9500\u72b6\u6001\u8f6c\u6362\u4e3a\u6570\u503c\ndf['promotion_numeric'] = df['promotion'].astype(int)\n\nprint(\"\\n\u6dfb\u52a0\u7279\u5f81\u540e\u7684\u6570\u636e:\")\nprint(df.head())<\/code><\/pre>\n\n\n\n<\/span>2. \u5bf9\u5206\u7c7b\u53d8\u91cf\u8fdb\u884c One-Hot \u7f16\u7801<\/span><\/h3>\n\n\n\n# \u5bf9\u7c7b\u522b\u4ee3\u7801\u548c\u4ea7\u54c1ID\u8fdb\u884cOne-Hot\u7f16\u7801\ndf_encoded = pd.get_dummies(df, columns=['category_code', 'product_id'])\n\nprint(\"\\nOne-Hot\u7f16\u7801\u540e\u7684\u6570\u636e:\")\nprint(df_encoded.head())<\/code><\/pre>\n\n\n\n<\/span>3. \u5212\u5206\u8bad\u7ec3\u96c6\u4e0e\u6d4b\u8bd5\u96c6\uff08\u6309\u65f6\u95f4\u5207\u7247\uff09<\/span><\/h3>\n\n\n\n# \u6309\u65f6\u95f4\u987a\u5e8f\u5212\u5206\u8bad\u7ec3\u96c6\u548c\u6d4b\u8bd5\u96c6\ntrain_size = int(len(df_encoded) * 0.8)\ntrain, test = df_encoded.iloc[:train_size], df_encoded.iloc[train_size:]\n\nX_train = train.drop(columns=['amount'])\ny_train = train['amount']\nX_test = test.drop(columns=['amount'])\ny_test = test['amount']\n\nprint(\"\\n\u8bad\u7ec3\u96c6\u548c\u6d4b\u8bd5\u96c6\u7684\u5f62\u72b6:\")\nprint(X_train.shape, y_train.shape, X_test.shape, y_test.shape)<\/code><\/pre>\n\n\n\n<\/span>4. \u4f7f\u7528\u7ebf\u6027\u56de\u5f52\u9884\u6d4b\u5355\u54c1\u9500\u91cf<\/span><\/h3>\n\n\n\n# \u7ebf\u6027\u56de\u5f52\u6a21\u578b\nlr = LinearRegression()\nlr.fit(X_train, y_train)\n\n# \u9884\u6d4b\ny_pred_lr = lr.predict(X_test)\n\n# \u8ba1\u7b97\u5747\u65b9\u8bef\u5dee\nmse_lr = mean_squared_error(y_test, y_pred_lr)\n\nprint(\"\\n\u7ebf\u6027\u56de\u5f52\u6a21\u578b MSE:\", mse_lr)<\/code><\/pre>\n\n\n\n<\/span>5. \u4f7f\u7528 XGBoost \u9884\u6d4b\u4e0d\u540c\u54c1\u7c7b\u7684\u589e\u957f\u8d8b\u52bf<\/span><\/h3>\n\n\n\n# XGBoost\u6a21\u578b\nxgb = XGBRegressor(objective='reg:squarederror', random_state=42)\nxgb.fit(X_train, y_train)\n\n# \u9884\u6d4b\ny_pred_xgb = xgb.predict(X_test)\n\n# \u8ba1\u7b97\u5747\u65b9\u8bef\u5dee\nmse_xgb = mean_squared_error(y_test, y_pred_xgb)\n\nprint(\"\\nXGBoost\u6a21\u578b MSE:\", mse_xgb)<\/code><\/pre>\n\n\n\n<\/span>6. \u4f7f\u7528 LightGBM \u6784\u5efa\u9ad8\u7ef4\u7279\u5f81\u7684\u9500\u91cf\u9884\u6d4b\u6a21\u578b<\/span><\/h3>\n\n\n\n# LightGBM\u6a21\u578b\nlgb = LGBMRegressor(random_state=42)\nlgb.fit(X_train, y_train)\n\n# \u9884\u6d4b\ny_pred_lgb = lgb.predict(X_test)\n\n# \u8ba1\u7b97\u5747\u65b9\u8bef\u5dee\nmse_lgb = mean_squared_error(y_test, y_pred_lgb)\n\nprint(\"\\nLightGBM\u6a21\u578b MSE:\", mse_lgb)<\/code><\/pre>\n\n\n\n<\/span>7. \u4f7f\u7528\u968f\u673a\u68ee\u6797\u9884\u6d4b\u65b0\u54c1\u4e0a\u5e02\u540e\u7684\u8868\u73b0<\/span><\/h3>\n\n\n\n# \u968f\u673a\u68ee\u6797\u6a21\u578b\nrf = RandomForestRegressor(random_state=42)\nrf.fit(X_train, y_train)\n\n# \u9884\u6d4b\ny_pred_rf = rf.predict(X_test)\n\n# \u8ba1\u7b97\u5747\u65b9\u8bef\u5dee\nmse_rf = mean_squared_error(y_test, y_pred_rf)\n\nprint(\"\\n\u968f\u673a\u68ee\u6797\u6a21\u578b MSE:\", mse_rf)<\/code><\/pre>\n\n\n\n<\/span>8. \u6784\u5efa\u591a\u8f93\u51fa\u6a21\u578b\u540c\u65f6\u9884\u6d4b\u591a\u4e2a\u5546\u54c1\u7684\u9500\u91cf<\/span><\/h3>\n\n\n\n\u8fd9\u91cc\u6211\u4eec\u5047\u8bbe\u6bcf\u4e2a\u5546\u54c1\u7684\u9500\u91cf\u53ef\u4ee5\u4f5c\u4e3a\u4e00\u4e2a\u72ec\u7acb\u7684\u76ee\u6807\u53d8\u91cf\u3002\u4e3a\u4e86\u7b80\u5316\u793a\u4f8b\uff0c\u6211\u4eec\u5c06\u53ea\u9009\u62e9\u51e0\u4e2a\u5546\u54c1\u8fdb\u884c\u9884\u6d4b\u3002<\/p>\n\n\n\n
# \u9009\u62e9\u51e0\u4e2a\u5546\u54c1\u4f5c\u4e3a\u76ee\u6807\u53d8\u91cf\ntarget_products = ['P1', 'P2', 'P3']\ntargets = df[df['product_id'].isin(target_products)].pivot_table(index=df.index, columns='product_id', values='amount', fill_value=0)\n\n# \u5408\u5e76\u7279\u5f81\u548c\u76ee\u6807\u53d8\u91cf\ndf_multi_output = pd.concat([df_encoded.drop(columns=['amount']), targets], axis=1)\n\n# \u6309\u65f6\u95f4\u987a\u5e8f\u5212\u5206\u8bad\u7ec3\u96c6\u548c\u6d4b\u8bd5\u96c6\ntrain_size = int(len(df_multi_output) * 0.8)\ntrain, test = df_multi_output.iloc[:train_size], df_multi_output.iloc[train_size:]\n\nX_train_multi = train.drop(columns=target_products)\ny_train_multi = train[target_products]\nX_test_multi = test.drop(columns=target_products)\ny_test_multi = test[target_products]\n\n# \u591a\u8f93\u51fa\u968f\u673a\u68ee\u6797\u6a21\u578b\nrf_multi = RandomForestRegressor(random_state=42)\nrf_multi.fit(X_train_multi, y_train_multi)\n\n# \u9884\u6d4b\ny_pred_rf_multi = rf_multi.predict(X_test_multi)\n\n# \u8ba1\u7b97\u5747\u65b9\u8bef\u5dee\nmse_rf_multi = mean_squared_error(y_test_multi, y_pred_rf_multi)\n\nprint(\"\\n\u591a\u8f93\u51fa\u968f\u673a\u68ee\u6797\u6a21\u578b MSE:\", mse_rf_multi)<\/code><\/pre>\n\n\n\n<\/span>9. \u4f7f\u7528\u4ea4\u53c9\u9a8c\u8bc1\u8bc4\u4f30\u6a21\u578b\u6548\u679c\uff08\u65f6\u95f4\u5e8f\u5217\u4e13\u7528\uff09<\/span><\/h3>\n\n\n\n# \u65f6\u95f4\u5e8f\u5217\u4ea4\u53c9\u9a8c\u8bc1\ntscv = TimeSeriesSplit(n_splits=5)\n\n# \u4f7f\u7528XGBoost\u6a21\u578b\u8fdb\u884c\u4ea4\u53c9\u9a8c\u8bc1\ncv_scores = cross_val_score(xgb, X_train, y_train, cv=tscv, scoring='neg_mean_squared_error')\n\n# \u8f6c\u6362\u4e3a\u6b63\u6570\u5e76\u8ba1\u7b97\u5e73\u5747\u503c\navg_cv_mse_xgb = -np.mean(cv_scores)\n\nprint(\"\\nXGBoost\u6a21\u578b\u4ea4\u53c9\u9a8c\u8bc1 MSE \u5e73\u5747\u503c:\", avg_cv_mse_xgb)<\/code><\/pre>\n\n\n\n<\/span>10. \u4f7f\u7528SHAP\u89e3\u91ca\u6a21\u578b\u9884\u6d4b\u7ed3\u679c\uff08\u53ef\u89e3\u91caAI\uff09<\/span><\/h3>\n\n\n\n# \u4f7f\u7528SHAP\u89e3\u91caXGBoost\u6a21\u578b\nexplainer = shap.Explainer(xgb)\nshap_values = explainer.shap_values(X_test)\n\n# \u7ed8\u5236SHAP\u6458\u8981\u56fe\nshap.summary_plot(shap_values, X_test)<\/code><\/pre>\n\n\n\n\u7efc\u5408\u4ee5\u4e0a\u6b65\u9aa4\uff0c\u6700\u7ec8\u7684\u673a\u5668\u5b66\u4e60\u9884\u6d4b\u7ed3\u679c\u5982\u4e0b\uff1a<\/p>\n\n\n\n
\u8fd9\u6bb5\u4ee3\u7801\u5c55\u793a\u4e86\u4ece\u539f\u59cb\u6570\u636e\u5230\u7ecf\u8fc7\u5168\u9762\u673a\u5668\u5b66\u4e60\u9884\u6d4b\u7684\u7ed3\u679c\u7684\u8fc7\u7a0b\u3002\u4f60\u53ef\u4ee5\u6839\u636e\u5b9e\u9645\u9700\u6c42\u8c03\u6574\u6bcf\u4e00\u6b65\u7684\u64cd\u4f5c\u3002<\/p>\n\n\n\n
import pandas as pd\r\nimport numpy as np\r\nfrom sklearn.model_selection import train_test_split, TimeSeriesSplit, cross_val_score\r\nfrom sklearn.linear_model import LinearRegression\r\nfrom sklearn.metrics import mean_squared_error\r\nfrom xgboost import XGBRegressor\r\nfrom lightgbm import LGBMRegressor\r\nfrom sklearn.ensemble import RandomForestRegressor\r\nimport shap\r\nimport matplotlib.pyplot as plt\r\n\r\n# \u521b\u5efa\u793a\u4f8b\u65f6\u95f4\u5e8f\u5217\u6570\u636e\r\ndates = pd.date_range(start='2023-01-01', end='2025-06-30', freq='D')\r\nnp.random.seed(42)\r\nsales_data = np.cumsum(np.random.normal(loc=100, scale=20, size=len(dates)))\r\n\r\ndata = {\r\n 'order_date': dates,\r\n 'product_id': np.random.choice(['P{}'.format(i) for i in range(1, 101)], len(dates)),\r\n 'category_code': np.random.choice(['C{}'.format(i) for i in range(1, 11)], len(dates)),\r\n 'amount': sales_data,\r\n 'quantity': np.random.randint(1, 5, size=len(dates)),\r\n 'promotion': np.random.choice([True, False], len(dates))\r\n}\r\n\r\ndf = pd.DataFrame(data)\r\n\r\n# \u8bbe\u7f6e\u65f6\u95f4\u4e3a\u7d22\u5f15\r\ndf.set_index('order_date', inplace=True)\r\n\r\n# \u6dfb\u52a0\u65f6\u95f4\u7279\u5f81\r\ndf['year'] = df.index.year\r\ndf['month'] = df.index.month\r\ndf['day_of_week'] = df.index.dayofweek\r\n\r\n# \u6dfb\u52a0\u8282\u5047\u65e5\u6807\u5fd7\r\nholidays = pd.to_datetime([\r\n '2023-01-01', '2023-02-22', '2023-04-05', '2023-05-01', '2023-10-01',\r\n '2024-01-01', '2024-02-10', '2024-04-04', '2024-05-01', '2024-10-01',\r\n '2025-01-01', '2025-02-19', '2025-04-04', '2025-05-01', '2025-10-01'\r\n])\r\n\r\ndf['is_holiday'] = df.index.isin(holidays).astype(int)\r\n\r\n# \u5c06\u4fc3\u9500\u72b6\u6001\u8f6c\u6362\u4e3a\u6570\u503c\r\ndf['promotion_numeric'] = df['promotion'].astype(int)\r\n\r\n# \u5bf9\u7c7b\u522b\u4ee3\u7801\u548c\u4ea7\u54c1ID\u8fdb\u884cOne-Hot\u7f16\u7801\r\ndf_encoded = pd.get_dummies(df, columns=['category_code', 'product_id'])\r\n\r\n# \u6309\u65f6\u95f4\u987a\u5e8f\u5212\u5206\u8bad\u7ec3\u96c6\u548c\u6d4b\u8bd5\u96c6\r\ntrain_size = int(len(df_encoded) * 0.8)\r\ntrain, test = df_encoded.iloc[:train_size], df_encoded.iloc[train_size:]\r\n\r\nX_train = train.drop(columns=['amount'])\r\ny_train = train['amount']\r\nX_test = test.drop(columns=['amount'])\r\ny_test = test['amount']\r\n\r\n# \u7ebf\u6027\u56de\u5f52\u6a21\u578b\r\nlr = LinearRegression()\r\nlr.fit(X_train, y_train)\r\ny_pred_lr = lr.predict(X_test)\r\nmse_lr = mean_squared_error(y_test, y_pred_lr)\r\nprint(\"\\n\u7ebf\u6027\u56de\u5f52\u6a21\u578b MSE:\", mse_lr)\r\n\r\n# XGBoost\u6a21\u578b\r\nxgb = XGBRegressor(objective='reg:squarederror', random_state=42)\r\nxgb.fit(X_train, y_train)\r\ny_pred_xgb = xgb.predict(X_test)\r\nmse_xgb = mean_squared_error(y_test, y_pred_xgb)\r\nprint(\"\\nXGBoost\u6a21\u578b MSE:\", mse_xgb)\r\n\r\n# LightGBM\u6a21\u578b\r\nlgb = LGBMRegressor(random_state=42)\r\nlgb.fit(X_train, y_train)\r\ny_pred_lgb = lgb.predict(X_test)\r\nmse_lgb = mean_squared_error(y_test, y_pred_lgb)\r\nprint(\"\\nLightGBM\u6a21\u578b MSE:\", mse_lgb)\r\n\r\n# \u968f\u673a\u68ee\u6797\u6a21\u578b\r\nrf = RandomForestRegressor(random_state=42)\r\nrf.fit(X_train, y_train)\r\ny_pred_rf = rf.predict(X_test)\r\nmse_rf = mean_squared_error(y_test, y_pred_rf)\r\nprint(\"\\n\u968f\u673a\u68ee\u6797\u6a21\u578b MSE:\", mse_rf)\r\n\r\n# \u9009\u62e9\u51e0\u4e2a\u5546\u54c1\u4f5c\u4e3a\u76ee\u6807\u53d8\u91cf\r\ntarget_products = ['P1', 'P2', 'P3']\r\ntargets = df[df['product_id'].isin(target_products)].pivot_table(index=df.index, columns='product_id', values='amount', fill_value=0)\r\n\r\n# \u5408\u5e76\u7279\u5f81\u548c\u76ee\u6807\u53d8\u91cf\r\ndf_multi_output = pd.concat([df_encoded.drop(columns=['amount']), targets], axis=1)\r\n\r\n# \u6309\u65f6\u95f4\u987a\u5e8f\u5212\u5206\u8bad\u7ec3\u96c6\u548c\u6d4b\u8bd5\u96c6\r\ntrain_size = int(len(df_multi_output) * 0.8)\r\ntrain, test = df_multi_output.iloc[:train_size], df_multi_output.iloc[train_size:]\r\n\r\nX_train_multi = train.drop(columns=target_products)\r\ny_train_multi = train[target_products]\r\nX_test_multi = test.drop(columns=target_products)\r\ny_test_multi = test[target_products]\r\n\r\n# \u591a\u8f93\u51fa\u968f\u673a\u68ee\u6797\u6a21\u578b\r\nrf_multi = RandomForestRegressor(random_state=42)\r\nrf_multi.fit(X_train_multi, y_train_multi)\r\ny_pred_rf_multi = rf_multi.predict(X_test_multi)\r\nmse_rf_multi = mean_squared_error(y_test_multi, y_pred_rf_multi)\r\nprint(\"\\n\u591a\u8f93\u51fa\u968f\u673a\u68ee\u6797\u6a21\u578b MSE:\", mse_rf_multi)\r\n\r\n# \u65f6\u95f4\u5e8f\u5217\u4ea4\u53c9\u9a8c\u8bc1\r\ntscv = TimeSeriesSplit(n_splits=5)\r\ncv_scores = cross_val_score(xgb, X_train, y_train, cv=tscv, scoring='neg_mean_squared_error')\r\navg_cv_mse_xgb = -np.mean(cv_scores)\r\nprint(\"\\nXGBoost\u6a21\u578b\u4ea4\u53c9\u9a8c\u8bc1 MSE \u5e73\u5747\u503c:\", avg_cv_mse_xgb)\r\n\r\n# \u4f7f\u7528SHAP\u89e3\u91caXGBoost\u6a21\u578b\r\nexplainer = shap.Explainer(xgb)\r\nshap_values = explainer.shap_values(X_test)\r\n\r\n# \u7ed8\u5236SHAP\u6458\u8981\u56fe\r\nshap.summary_plot(shap_values, X_test)\r\n\r\n\r\n\r\n<\/code><\/pre>\n","protected":false},"excerpt":{"rendered":"\u5de5\u4f5c\u5e38\u7528\u7684\u673a\u5668\u5b66\u4e60\u9884\u6d4b\u6848\u4f8b\uff1a \u597d\u7684\uff0c\u6211\u4eec\u5c06\u5c55\u793a\u5982\u4f55\u8fdb\u884c\u8fd9\u4e9b\u5e38\u89c1\u7684\u673a\u5668\u5b66\u4e60\u9884\u6d4b\u4efb\u52a1\u3002\u4e3a\u4e86\u6f14\u793a\u8fd9\u4e9b\u65b9\u6cd5\uff0c… \u7ee7\u7eed\u9605\u8bfb \u3010Python10\u5e74\u7ecf\u9a8c\u603b\u7ed3\u3011\u7b2c\u516b\u8bfe \u7535\u5546\u5e73\u53f0\u9500\u552e\u6570\u636e\u5206\u6790\u5b9e\u8df5 -\u673a\u5668\u5b66\u4e60\u9884\u6d4b\uff08Machine Learning Forecasting\uff09<\/span><\/i><\/a><\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"closed","ping_status":"open","sticky":false,"template":"","format":"standard","meta":[],"categories":[1],"tags":[28],"views":1092,"_links":{"self":[{"href":"http:\/\/viplao.com\/index.php\/wp-json\/wp\/v2\/posts\/3550"}],"collection":[{"href":"http:\/\/viplao.com\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"http:\/\/viplao.com\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"http:\/\/viplao.com\/index.php\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"http:\/\/viplao.com\/index.php\/wp-json\/wp\/v2\/comments?post=3550"}],"version-history":[{"count":2,"href":"http:\/\/viplao.com\/index.php\/wp-json\/wp\/v2\/posts\/3550\/revisions"}],"predecessor-version":[{"id":3566,"href":"http:\/\/viplao.com\/index.php\/wp-json\/wp\/v2\/posts\/3550\/revisions\/3566"}],"wp:attachment":[{"href":"http:\/\/viplao.com\/index.php\/wp-json\/wp\/v2\/media?parent=3550"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"http:\/\/viplao.com\/index.php\/wp-json\/wp\/v2\/categories?post=3550"},{"taxonomy":"post_tag","embeddable":true,"href":"http:\/\/viplao.com\/index.php\/wp-json\/wp\/v2\/tags?post=3550"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}