{"id":3903,"date":"2025-09-13T23:19:50","date_gmt":"2025-09-13T15:19:50","guid":{"rendered":"http:\/\/viplao.com\/?p=3903"},"modified":"2025-09-13T23:19:54","modified_gmt":"2025-09-13T15:19:54","slug":"%e3%80%90python%e5%ae%9e%e8%b7%b5%e6%a1%88%e4%be%8b%e3%80%91%e7%94%b5%e5%95%86%e5%b9%b3%e5%8f%b0%e6%95%b0%e6%8d%ae%e5%88%86%e6%9e%90%e5%92%8c%e6%8c%96%e6%8e%98-%e6%97%b6%e5%ba%8f%e9%a2%84%e6%b5%8b","status":"publish","type":"post","link":"http:\/\/viplao.com\/index.php\/2025\/09\/13\/%e3%80%90python%e5%ae%9e%e8%b7%b5%e6%a1%88%e4%be%8b%e3%80%91%e7%94%b5%e5%95%86%e5%b9%b3%e5%8f%b0%e6%95%b0%e6%8d%ae%e5%88%86%e6%9e%90%e5%92%8c%e6%8c%96%e6%8e%98-%e6%97%b6%e5%ba%8f%e9%a2%84%e6%b5%8b\/","title":{"rendered":"\u3010Python\u5b9e\u8df5\u6848\u4f8b\u3011\u7535\u5546\u5e73\u53f0\u6570\u636e\u5206\u6790\u548c\u6316\u6398 – \u65f6\u5e8f\u9884\u6d4b"},"content":{"rendered":"\n

\u6211\u4eec\u6765\u521b\u5efa\u4e00\u4e2a\u9488\u5bf9\u7535\u5546\u5e73\u53f0\u65e5\u5fd7\u6570\u636e\u65f6\u5e8f\u9884\u6d4b\u7a0b\u5e8f\u3002\u8fd9\u4e2a\u7a0b\u5e8f\u5c06\u6a21\u62df\u751f\u6210\u5305\u542b\u8bbf\u95ee\u91cf\uff08\u5982PV\u6216UV\uff09\u7684\u65f6\u5e8f\u65e5\u5fd7\u6570\u636e\uff0c\u5e76\u4f7f\u7528ARIMA\u548cProphet\u4e24\u79cd\u6a21\u578b\u5bf9\u672a\u6765\u6d41\u91cf\u8fdb\u884c\u9884\u6d4b\u3002<\/p>\n\n\n\n

\u91cd\u8981\u63d0\u793a<\/strong>\uff1a<\/p>\n\n\n\n

    \n
  1. Prophet\u5e93\u5b89\u88c5<\/strong>\uff1aprophet<\/code>\u5e93\u7684\u5b89\u88c5\u53ef\u80fd\u9700\u8981C++\u7f16\u8bd1\u5668\u548c\u4e00\u4e9b\u989d\u5916\u4f9d\u8d56\u3002\u5982\u679c\u5b89\u88c5\u9047\u5230\u95ee\u9898\uff0c\u8bf7\u53c2\u8003\u5176\u5b98\u65b9\u6587\u6863\u3002\u5982\u679c\u65e0\u6cd5\u5b89\u88c5\uff0c\u7a0b\u5e8f\u4f1a\u81ea\u52a8\u8df3\u8fc7Prophet\u90e8\u5206\u3002<\/li>\n\n\n\n
  2. pmdarima\u5e93\u5b89\u88c5<\/strong>\uff1a\u4e3a\u4e86\u5b9e\u73b0ARIMA\u6a21\u578b\u7684\u81ea\u52a8\u8c03\u53c2\uff0c\u6211\u4eec\u4f7f\u7528pmdarima<\/code>\u5e93\u3002\u5982\u679c\u672a\u5b89\u88c5\uff0c\u7a0b\u5e8f\u4f1a\u56de\u9000\u5230\u4f7f\u7528\u56fa\u5b9a\u53c2\u6570\u7684statsmodels ARIMA\u3002<\/li>\n<\/ol>\n\n\n\n

    <\/p>\n\n\n\n

    <\/p>\n\n\n\n

    import pandas as pd\nimport numpy as np\nfrom sklearn.metrics import mean_absolute_error, mean_squared_error\nimport matplotlib.pyplot as plt\nimport seaborn as sns\nfrom datetime import datetime, timedelta\nimport warnings\nwarnings.filterwarnings(\"ignore\")\n\n# --- \u5c1d\u8bd5\u5bfc\u5165\u53ef\u9009\u5e93 ---\nPROPHET_AVAILABLE = False\nPMDARIMA_AVAILABLE = False\n\ntry:\n    from prophet import Prophet\n    PROPHET_AVAILABLE = True\n    print(\"Info: Prophet\u5e93\u5df2\u5bfc\u5165\uff0c\u5c06\u5305\u542bProphet\u6a21\u578b\u9884\u6d4b\u3002\")\nexcept ImportError:\n    print(\"Warning: Prophet\u5e93\u672a\u5bfc\u5165\u3002\u5c06\u8df3\u8fc7Prophet\u6a21\u578b\u90e8\u5206\u3002\")\n\ntry:\n    from pmdarima import auto_arima\n    PMDARIMA_AVAILABLE = True\n    print(\"Info: pmdarima\u5e93\u5df2\u5bfc\u5165\uff0cARIMA\u6a21\u578b\u5c06\u81ea\u52a8\u9009\u62e9\u53c2\u6570\u3002\")\nexcept ImportError:\n    print(\"Warning: pmdarima\u5e93\u672a\u5bfc\u5165\u3002ARIMA\u6a21\u578b\u5c06\u4f7f\u7528\u56fa\u5b9a\u53c2\u6570(2,1,2)\u3002\")\n    from statsmodels.tsa.arima.model import ARIMA # \u57fa\u7840ARIMA\n\n# --- \u914d\u7f6e ---\n# \u6a21\u62df\u6570\u636e\u53c2\u6570\nNUM_DAYS_HISTORY = 730  # \u5386\u53f2\u6570\u636e\u5929\u6570 (\u4f8b\u59822\u5e74)\nFORECAST_HORIZON = 30   # \u9884\u6d4b\u672a\u6765\u5929\u6570\nREPORT_PREFIX = '\u7535\u5546\u65e5\u5fd7\u65f6\u5e8f\u9884\u6d4b\u62a5\u544a'\nRANDOM_SEED = 42\nnp.random.seed(RANDOM_SEED)\n\n# --- 1. \u6570\u636e\u751f\u6210 ---\n\ndef generate_sample_traffic_data(n_days):\n    \"\"\"\n    \u751f\u6210\u6a21\u62df\u7684\u7535\u5546\u5e73\u53f0\u65e5\u8bbf\u95ee\u91cf(\u5982PV)\u65f6\u5e8f\u6570\u636e\u3002\n    \u5305\u542b\u8d8b\u52bf\u3001\u591a\u91cd\u5b63\u8282\u6027\uff08\u5468\u3001\u5e74\uff09\u3001\u4e8b\u4ef6\u6548\u5e94\u548c\u566a\u58f0\u3002\n    \"\"\"\n    print(\"--- \u6b63\u5728\u751f\u6210\u6a21\u62df\u7535\u5546\u65e5\u5fd7\u6d41\u91cf\u6570\u636e ---\")\n    start_date = datetime.now() - timedelta(days=n_days-1)\n    dates = pd.date_range(start=start_date, periods=n_days, freq='D')\n    \n    # 1. \u57fa\u7840\u6c34\u5e73\u548c\u957f\u671f\u8d8b\u52bf\n    base_level = 50000\n    trend_slope = np.random.uniform(50, 150) # \u6bcf\u5929\u5e73\u5747\u589e\u957f50-150\u4e2aPV\n    trend = np.arange(n_days) * trend_slope\n    \n    # 2. \u5b63\u8282\u6027\n    # \u5468\u5ea6\u5b63\u8282\u6027 (\u5468\u672b\u6d41\u91cf\u9ad8)\n    day_of_week = pd.Series(dates).dt.dayofweek\n    weekly_effect = np.where(day_of_week.isin([5, 6]), 1.2, 1.0) # \u5468\u516d\u3001\u65e5\u6d41\u91cf\u589e\u52a020%\n    \n    # \u5e74\u5ea6\u5b63\u8282\u6027 (\u7b80\u5316\u4e3a\u6b63\u5f26\u6ce2\uff0c\u4f8b\u5982\u590f\u5b63\u6d41\u91cf\u53ef\u80fd\u66f4\u9ad8)\n    doy = pd.Series(dates).dt.dayofyear\n    annual_effect = 1 + 0.1 * np.sin(2 * np.pi * (doy - 80) \/ 365.25) # \u4ece3\u670821\u65e5(\u6625\u5206)\u5f00\u59cb\u8ba1\u7b97\uff0c\u5cf0\u503c\u5728\u590f\u81f3\n    \n    # 3. \u7279\u5b9a\u4e8b\u4ef6\/\u4fc3\u9500\u6548\u5e94 (\u6a21\u62df\u51e0\u4e2a\u5927\u4fc3\u8282\u65e5)\n    event_effect = np.ones(n_days)\n    # 618 (6\u670818\u65e5)\n    june_18_indices = [i for i, date in enumerate(dates) if date.month == 6 and date.day == 18]\n    for idx in june_18_indices:\n        event_effect[idx] *= np.random.uniform(3.0, 5.0) # \u6d41\u91cf\u7ffb3-5\u500d\n        if idx > 0: event_effect[idx-1] *= np.random.uniform(1.5, 2.5) # \u524d\u4e00\u5929\u9884\u70ed\n        if idx < n_days-1: event_effect[idx+1] *= np.random.uniform(1.2, 1.8) # \u540e\u4e00\u5929\u8fd4\u573a\n    \n    # \u53cc11 (11\u670811\u65e5)\n    nov_11_indices = [i for i, date in enumerate(dates) if date.month == 11 and date.day == 11]\n    for idx in nov_11_indices:\n        event_effect[idx] *= np.random.uniform(5.0, 8.0) # \u6d41\u91cf\u7ffb5-8\u500d\n        if idx > 0: event_effect[idx-1] *= np.random.uniform(2.0, 4.0)\n        if idx < n_days-1: event_effect[idx+1] *= np.random.uniform(1.5, 2.5)\n        \n    # 4. \u968f\u673a\u566a\u58f0 (\u4f7f\u7528\u6cca\u677e\u566a\u58f0\u66f4\u7b26\u5408\u8ba1\u6570\u6570\u636e\u7279\u6027\uff0c\u4f46\u8fd9\u91cc\u7528\u6b63\u6001\u7b80\u5316)\n    noise = np.random.normal(1.0, 0.05, n_days) # 5%\u7684\u968f\u673a\u6ce2\u52a8\n    \n    # 5. \u7ec4\u5408\u6240\u6709\u6210\u5206\n    traffic = base_level + trend\n    traffic = traffic * weekly_effect * annual_effect * event_effect * noise\n    traffic = np.maximum(traffic, 0) # \u6d41\u91cf\u4e0d\u80fd\u4e3a\u8d1f\n    \n    df = pd.DataFrame({\n        'date': dates,\n        'value': np.round(traffic, 0).astype(int) # \u6a21\u62dfPV\/UV\u7b49\u6574\u6570\u6307\u6807\n    })\n    \n    csv_filename = f'{REPORT_PREFIX}_\u6a21\u62df\u6d41\u91cf\u6570\u636e.csv'\n    df.to_csv(csv_filename, index=False, encoding='utf-8-sig')\n    print(f\"\u6a21\u62df\u6d41\u91cf\u6570\u636e\u5df2\u751f\u6210\u5e76\u4fdd\u5b58\u81f3: {csv_filename}\")\n    return df\n\n# --- 2. \u6570\u636e\u9884\u5904\u7406\u4e0eEDA ---\n\ndef preprocess_and_eda(df):\n    \"\"\"\u6570\u636e\u9884\u5904\u7406\u548c\u63a2\u7d22\u6027\u6570\u636e\u5206\u6790\"\"\"\n    print(\"\\n--- \u6b63\u5728\u8fdb\u884c\u6570\u636e\u9884\u5904\u7406\u4e0eEDA ---\")\n    df_processed = df.copy()\n    df_processed.set_index('date', inplace=True)\n    df_processed.sort_index(inplace=True)\n    \n    # \u68c0\u67e5\u7f3a\u5931\u503c\n    if df_processed.isnull().sum().sum() > 0:\n        print(\"\u8b66\u544a: \u53d1\u73b0\u7f3a\u5931\u503c\uff0c\u6b63\u5728\u8fdb\u884c\u524d\u5411\u586b\u5145...\")\n        df_processed['value'].fillna(method='ffill', inplace=True)\n    \n    # \u7ed8\u5236\u65f6\u95f4\u5e8f\u5217\u56fe\n    plt.figure(figsize=(15, 6))\n    plt.plot(df_processed.index, df_processed['value'], linewidth=1, label='\u5386\u53f2\u6d41\u91cf')\n    plt.title('\u7535\u5546\u5e73\u53f0\u65e5\u5fd7\u6d41\u91cf\u65f6\u95f4\u5e8f\u5217')\n    plt.xlabel('\u65e5\u671f')\n    plt.ylabel('\u6d41\u91cf\u6307\u6807 (\u6a21\u62dfPV)')\n    plt.legend()\n    plt.grid(True)\n    ts_plot_path = f'{REPORT_PREFIX}_\u6d41\u91cf\u65f6\u95f4\u5e8f\u5217\u56fe.png'\n    plt.savefig(ts_plot_path, bbox_inches='tight')\n    plt.close()\n    print(f\"\u6d41\u91cf\u65f6\u95f4\u5e8f\u5217\u56fe\u5df2\u4fdd\u5b58\u81f3: {ts_plot_path}\")\n    \n    return df_processed\n\n# --- 3. ARIMA \u6a21\u578b\u9884\u6d4b ---\n\ndef forecast_with_arima(df, forecast_periods):\n    \"\"\"\u4f7f\u7528ARIMA\u6a21\u578b\u8fdb\u884c\u9884\u6d4b\"\"\"\n    print(\"\\n--- \u6b63\u5728\u4f7f\u7528ARIMA\u6a21\u578b\u8fdb\u884c\u9884\u6d4b ---\")\n    \n    train_size = int(len(df) * 0.85)\n    train, test = df.iloc[:train_size], df.iloc[train_size:]\n    \n    # \u4f7f\u7528auto_arima\u81ea\u52a8\u9009\u62e9\u6700\u4f73\u53c2\u6570\uff0c\u6216\u4f7f\u7528\u56fa\u5b9a\u53c2\u6570\n    if PMDARIMA_AVAILABLE:\n        print(\"\u4f7f\u7528auto_arima\u81ea\u52a8\u9009\u62e9ARIMA\u53c2\u6570...\")\n        # seasonal=True \u5bf9\u4e8e\u5e74\u5ea6\u6570\u636e\u6548\u679c\u597d\uff0c\u4f46\u8ba1\u7b97\u6162\u3002\u8fd9\u91cc\u7b80\u5316\u5904\u7406\u3002\n        model_auto = auto_arima(train['value'], start_p=1, max_p=5, start_q=1, max_q=5,\n                                d=1, max_d=2, trace=False, error_action='ignore', \n                                suppress_warnings=True, stepwise=True, seasonal=False)\n        print(f\"auto_arima\u9009\u62e9\u7684\u6700\u4f73\u6a21\u578b: ARIMA{model_auto.order}\")\n        model = model_auto\n    else:\n        print(\"pmdarima\u4e0d\u53ef\u7528\uff0c\u4f7f\u7528\u56fa\u5b9a\u53c2\u6570 ARIMA(2,1,2)...\")\n        model = ARIMA(train['value'], order=(2,1,2))\n        \n    # \u62df\u5408\u6a21\u578b\n    model_fit = model.fit() if not PMDARIMA_AVAILABLE else model\n    \n    # \u9884\u6d4b\u6d4b\u8bd5\u96c6\u7528\u4e8e\u8bc4\u4f30\n    if len(test) > 0:\n        test_forecast = model_fit.forecast(steps=len(test))\n        test_forecast = np.maximum(test_forecast, 0) # \u786e\u4fdd\u9884\u6d4b\u503c\u975e\u8d1f\n    \n    # \u9884\u6d4b\u672a\u6765\n    future_forecast = model_fit.forecast(steps=forecast_periods)\n    future_forecast = np.maximum(future_forecast, 0)\n    future_dates = pd.date_range(start=df.index[-1] + timedelta(days=1), periods=forecast_periods, freq='D')\n    \n    # \u8bc4\u4f30 (\u4ec5\u5728\u6709\u6d4b\u8bd5\u96c6\u65f6)\n    metrics = {}\n    if len(test) > 0:\n        metrics['MAE'] = mean_absolute_error(test['value'], test_forecast)\n        metrics['RMSE'] = np.sqrt(mean_squared_error(test['value'], test_forecast))\n        print(f\"ARIMA\u6a21\u578b\u5728\u6d4b\u8bd5\u96c6\u4e0a\u7684\u8868\u73b0:\")\n        for k, v in metrics.items():\n            print(f\"  - {k}: {v:.2f}\")\n    \n    # \u7ed8\u5236\u9884\u6d4b\u7ed3\u679c\n    plt.figure(figsize=(15, 6))\n    plt.plot(df.index, df['value'], label='\u5386\u53f2\u6d41\u91cf', linewidth=1)\n    if len(test) > 0:\n        plt.plot(test.index, test_forecast, label='ARIMA\u6d4b\u8bd5\u96c6\u9884\u6d4b', linestyle='--', alpha=0.8)\n    plt.plot(future_dates, future_forecast, label=f'ARIMA\u672a\u6765{forecast_periods}\u5929\u9884\u6d4b', linestyle='-.', marker='o', markersize=4)\n    plt.axvline(x=df.index[train_size-1], color='black', linestyle=':', alpha=0.7, label='\u8bad\u7ec3\/\u6d4b\u8bd5\u5206\u5272\u70b9')\n    plt.title('ARIMA\u6a21\u578b\u6d41\u91cf\u9884\u6d4b')\n    plt.xlabel('\u65e5\u671f')\n    plt.ylabel('\u6d41\u91cf\u6307\u6807 (\u6a21\u62dfPV)')\n    plt.legend()\n    plt.grid(True)\n    arima_plot_path = f'{REPORT_PREFIX}_ARIMA\u9884\u6d4b\u56fe.png'\n    plt.savefig(arima_plot_path, bbox_inches='tight')\n    plt.close()\n    print(f\"ARIMA\u9884\u6d4b\u56fe\u5df2\u4fdd\u5b58\u81f3: {arima_plot_path}\")\n    \n    results = {\n        'model_name': 'ARIMA',\n        'predictions': future_forecast,\n        'future_dates': future_dates,\n        'metrics': metrics,\n        'plot_path': arima_plot_path\n    }\n    return results\n\n# --- 4. Prophet \u6a21\u578b\u9884\u6d4b ---\n\ndef forecast_with_prophet(df, forecast_periods):\n    \"\"\"\u4f7f\u7528Prophet\u6a21\u578b\u8fdb\u884c\u9884\u6d4b\"\"\"\n    if not PROPHET_AVAILABLE:\n        print(\"Prophet\u6a21\u578b\u8df3\u8fc7\uff0c\u56e0\u4e3a\u5e93\u672a\u5bfc\u5165\u3002\")\n        return None\n        \n    print(\"\\n--- \u6b63\u5728\u4f7f\u7528Prophet\u6a21\u578b\u8fdb\u884c\u9884\u6d4b ---\")\n    \n    # Prophet\u9700\u8981\u7279\u5b9a\u7684\u5217\u540d\n    df_prophet = df.reset_index()[['date', 'value']].rename(columns={'date': 'ds', 'value': 'y'})\n    \n    # \u521b\u5efa\u5e76\u62df\u5408\u6a21\u578b\n    model = Prophet(\n        daily_seasonality=False, # \u6570\u636e\u662f\u65e5\u7c92\u5ea6\uff0c\u5173\u95ed\u5185\u7f6e\u7684\u65e5\u5b63\u8282\u6027\n        yearly_seasonality='auto',\n        weekly_seasonality='auto',\n        seasonality_mode='multiplicative' # \u5bf9\u4e8e\u6709\u660e\u663e\u6bd4\u4f8b\u53d8\u5316\u7684\u5b63\u8282\u6027\u66f4\u5408\u9002\n    )\n    # \u53ef\u4ee5\u5728\u8fd9\u91cc\u6dfb\u52a0\u81ea\u5b9a\u4e49\u8282\u5047\u65e5\n    # model.add_country_holidays(country_name='CN')\n    \n    model.fit(df_prophet)\n    \n    # \u521b\u5efa\u672a\u6765\u65e5\u671f\u6846\u67b6\n    future = model.make_future_dataframe(periods=forecast_periods)\n    \n    # \u8fdb\u884c\u9884\u6d4b\n    forecast = model.predict(future)\n    \n    # \u7ed8\u5236\u9884\u6d4b\u7ed3\u679c\n    fig1 = model.plot(forecast, figsize=(15, 6))\n    plt.title('Prophet\u6a21\u578b\u6d41\u91cf\u9884\u6d4b')\n    prophet_plot_path = f'{REPORT_PREFIX}_Prophet\u9884\u6d4b\u56fe.png'\n    plt.savefig(prophet_plot_path, bbox_inches='tight')\n    plt.close(fig1)\n    \n    # \u7ed8\u5236\u6a21\u578b\u6210\u5206 (\u8d8b\u52bf\u3001\u5b63\u8282\u6027\u7b49)\n    fig2 = model.plot_components(forecast, figsize=(12, 10))\n    prophet_comp_plot_path = f'{REPORT_PREFIX}_Prophet\u6210\u5206\u56fe.png'\n    plt.savefig(prophet_comp_plot_path, bbox_inches='tight')\n    plt.close(fig2)\n    print(f\"Prophet\u9884\u6d4b\u56fe\u5df2\u4fdd\u5b58\u81f3: {prophet_plot_path}\")\n    print(f\"Prophet\u6210\u5206\u56fe\u5df2\u4fdd\u5b58\u81f3: {prophet_comp_plot_path}\")\n    \n    # \u63d0\u53d6\u672a\u6765\u9884\u6d4b\u503c\n    future_forecast_df = forecast[['ds', 'yhat', 'yhat_lower', 'yhat_upper']].tail(forecast_periods)\n    future_forecast_df['yhat'] = np.maximum(future_forecast_df['yhat'], 0)\n    future_forecast_df['yhat_lower'] = np.maximum(future_forecast_df['yhat_lower'], 0)\n    \n    results = {\n        'model_name': 'Prophet',\n        'forecast_df': future_forecast_df,\n        'metrics': {}, # Prophet\u6709\u5185\u7f6e\u4ea4\u53c9\u9a8c\u8bc1\uff0c\u8fd9\u91cc\u7b80\u5316\n        'plot_path': prophet_plot_path,\n        'comp_plot_path': prophet_comp_plot_path\n    }\n    return results\n\n# --- 5. \u62a5\u544a\u751f\u6210 ---\n\ndef generate_forecast_report(arima_res, prophet_res, forecast_horizon):\n    \"\"\"\u751f\u6210\u6700\u7ec8\u7684\u65f6\u5e8f\u9884\u6d4b\u5206\u6790\u62a5\u544a\"\"\"\n    print(\"\\n--- \u6b63\u5728\u751f\u6210\u7535\u5546\u65e5\u5fd7\u65f6\u5e8f\u9884\u6d4b\u62a5\u544a ---\")\n    from datetime import datetime\n    report_filename = f\"{REPORT_PREFIX}_{datetime.now().strftime('%Y%m%d_%H%M%S')}.txt\"\n    \n    with open(report_filename, 'w', encoding='utf-8') as f:\n        f.write(\"=\" * 60 + \"\\n\")\n        f.write(\"           \u7535\u5546\u5e73\u53f0\u65e5\u5fd7\u6570\u636e\u65f6\u5e8f\u9884\u6d4b\u5206\u6790\u62a5\u544a\\n\")\n        f.write(f\"              \u751f\u6210\u65f6\u95f4: {datetime.now().strftime('%Y-%m-%d %H:%M:%S')}\\n\")\n        f.write(\"=\" * 60 + \"\\n\\n\")\n\n        f.write(\"--- 1. \u9879\u76ee\u6982\u8ff0 ---\\n\")\n        f.write(\"\u672c\u62a5\u544a\u65e8\u5728\u901a\u8fc7\u5bf9\u7535\u5546\u5e73\u53f0\u5386\u53f2\u65e5\u5fd7\u6570\u636e\uff08\u6a21\u62df\u6d41\u91cf\u6307\u6807\uff09\u7684\u5206\u6790\uff0c\u9884\u6d4b\u672a\u6765\u4e00\u6bb5\u65f6\u95f4\u7684\u6d41\u91cf\u8d8b\u52bf\u3002\\n\")\n        f.write(\"\u6b64\u9884\u6d4b\u53ef\u4e3a\u670d\u52a1\u5668\u8d44\u6e90\u89c4\u5212\u3001\u5e26\u5bbd\u7ba1\u7406\u3001\u8425\u9500\u6d3b\u52a8\u5b89\u6392\u7b49\u63d0\u4f9b\u6570\u636e\u652f\u6301\u3002\\n\\n\")\n\n        f.write(\"--- 2. \u6570\u636e\u6982\u89c8 ---\\n\")\n        f.write(f\"\u6570\u636e\u6765\u6e90: \u6a21\u62df\u751f\u6210\u7684\u7535\u5546\u5e73\u53f0\u65e5\u8bbf\u95ee\u91cf(PV)\u6570\u636e\u3002\\n\")\n        f.write(f\"\u6570\u636e\u89c4\u6a21: {NUM_DAYS_HISTORY} \u5929\u7684\u5386\u53f2\u8bb0\u5f55\u3002\\n\")\n        f.write(\"\u6570\u636e\u7279\u5f81: \u5305\u542b\u957f\u671f\u589e\u957f\u8d8b\u52bf\u3001\u5468\u5ea6\u5b63\u8282\u6027\u3001\u5e74\u5ea6\u5b63\u8282\u6027\u4ee5\u53ca'618'\u3001'\u53cc11'\u7b49\u5927\u4fc3\u4e8b\u4ef6\u6548\u5e94\u3002\\n\")\n        f.write(\"\u539f\u59cb\u6570\u636e\u5df2\u4fdd\u5b58\u4e3a CSV \u6587\u4ef6\u3002\\n\\n\")\n\n        f.write(\"--- 3. \u63a2\u7d22\u6027\u6570\u636e\u5206\u6790 (EDA) ---\\n\")\n        f.write(\"\u5bf9\u65f6\u95f4\u5e8f\u5217\u6570\u636e\u8fdb\u884c\u4e86\u53ef\u89c6\u5316\u5206\u6790\uff0c\u4ee5\u8bc6\u522b\u5176\u6838\u5fc3\u6a21\u5f0f\uff1a\\n\")\n        f.write(\"- \u8d8b\u52bf: \u6570\u636e\u663e\u793a\u603b\u4f53\u5448\u7a33\u5b9a\u4e0a\u5347\u8d8b\u52bf\u3002\\n\")\n        f.write(\"- \u5b63\u8282\u6027: \u5b58\u5728\u663e\u8457\u7684\u5468\u5ea6\uff08\u5468\u672b\u9ad8\u5cf0\uff09\u548c\u5e74\u5ea6\u5468\u671f\u6027\u6ce2\u52a8\u3002\\n\")\n        f.write(\"- \u4e8b\u4ef6\u6548\u5e94: \u53ef\u4ee5\u6e05\u6670\u89c2\u5bdf\u5230'618'\u3001'\u53cc11'\u7b49\u5927\u4fc3\u8282\u65e5\u5e26\u6765\u7684\u6d41\u91cf\u5de8\u5927\u5cf0\u503c\u3002\\n\")\n        f.write(\"\u6d41\u91cf\u65f6\u95f4\u5e8f\u5217\u56fe\u5df2\u751f\u6210\u3002\\n\\n\")\n\n        f.write(\"--- 4. \u9884\u6d4b\u6a21\u578b ---\\n\")\n        f.write(\"\u91c7\u7528\u4e86\u4e24\u79cd\u4e3b\u6d41\u7684\u65f6\u95f4\u5e8f\u5217\u9884\u6d4b\u6a21\u578b\u8fdb\u884c\u5efa\u6a21\u4e0e\u5bf9\u6bd4\uff1a\\n\")\n        f.write(\"1. ARIMA (\u81ea\u56de\u5f52\u79ef\u5206\u6ed1\u52a8\u5e73\u5747\u6a21\u578b): \\n\")\n        f.write(\"   - \u4e00\u79cd\u57fa\u4e8e\u7edf\u8ba1\u5b66\u7684\u7ecf\u5178\u6a21\u578b\u3002\\n\")\n        f.write(\"   - \u9002\u7528\u4e8e\u5355\u53d8\u91cf\u3001\u5e73\u7a33\u6216\u53ef\u8f6c\u5316\u4e3a\u5e73\u7a33\u7684\u65f6\u95f4\u5e8f\u5217\u3002\\n\")\n        f.write(\"   - \u4f18\u70b9\u662f\u6a21\u578b\u53ef\u89e3\u91ca\u6027\u5f3a\u3002\\n\")\n        f.write(\"2. Prophet (\u7531Facebook\u5f00\u53d1): \\n\")\n        f.write(\"   - \u4e00\u79cd\u7075\u6d3b\u4e14\u5f3a\u5927\u7684\u6a21\u578b\uff0c\u7279\u522b\u64c5\u957f\u5904\u7406\u5177\u6709\u5f3a\u5b63\u8282\u6027\u3001\u5386\u53f2\u6570\u636e\u7f3a\u5931\u548c\u8d8b\u52bf\u7a81\u53d8\u7684\u5e8f\u5217\u3002\\n\")\n        f.write(\"   - \u5185\u7f6e\u5bf9\u8282\u5047\u65e5\u7684\u652f\u6301\u3002\\n\")\n        f.write(\"   - \u4f18\u70b9\u662f\u6613\u4e8e\u4f7f\u7528\u4e14\u9c81\u68d2\u6027\u597d\u3002\\n\\n\")\n        \n        f.write(\"--- 5. ARIMA\u6a21\u578b\u9884\u6d4b\u7ed3\u679c ---\\n\")\n        if arima_res:\n            f.write(f\"\u6a21\u578b: {arima_res['model_name']}\\n\")\n            if arima_res['metrics']:\n                f.write(\"\u6a21\u578b\u8bc4\u4f30 (\u5728\u6d4b\u8bd5\u96c6\u4e0a):\\n\")\n                for metric, value in arima_res['metrics'].items():\n                    f.write(f\"  - {metric}: {value:.2f}\\n\")\n            f.write(f\"\u672a\u6765{forecast_horizon}\u5929\u9884\u6d4b\u503c\u9884\u89c8:\\n\")\n            forecast_df_arima = pd.DataFrame({'date': arima_res['future_dates'], 'predicted_value': arima_res['predictions']})\n            f.write(forecast_df_arima.round(2).to_string(index=False))\n            f.write(f\"\\n\u9884\u6d4b\u56fe\u8868: {arima_res['plot_path']}\\n\\n\")\n        else:\n            f.write(\"ARIMA\u6a21\u578b\u6267\u884c\u5931\u8d25\u6216\u88ab\u8df3\u8fc7\u3002\\n\\n\")\n\n        f.write(\"--- 6. Prophet\u6a21\u578b\u9884\u6d4b\u7ed3\u679c ---\\n\")\n        if prophet_res:\n            f.write(f\"\u6a21\u578b: {prophet_res['model_name']}\\n\")\n            f.write(f\"\u672a\u6765{forecast_horizon}\u5929\u9884\u6d4b\u503c\u9884\u89c8 (\u5305\u542b\u4e0d\u786e\u5b9a\u6027\u533a\u95f4):\\n\")\n            forecast_summary = prophet_res['forecast_df'].rename(columns={\n                'ds': 'date', 'yhat': 'predicted_value', \n                'yhat_lower': 'lower_bound', 'yhat_upper': 'upper_bound'\n            })\n            f.write(forecast_summary.round(2).to_string(index=False))\n            f.write(f\"\\n\u9884\u6d4b\u56fe\u8868: {prophet_res['plot_path']}\\n\")\n            f.write(f\"\u8d8b\u52bf\u4e0e\u5b63\u8282\u6027\u6210\u5206\u56fe: {prophet_res['comp_plot_path']}\\n\\n\")\n        else:\n            f.write(\"Prophet\u6a21\u578b\u56e0\u5e93\u672a\u5bfc\u5165\u800c\u88ab\u8df3\u8fc7\u3002\\n\\n\")\n\n        f.write(\"--- 7. \u7ed3\u8bba\u4e0e\u5efa\u8bae ---\\n\")\n        f.write(\"1. \u8d8b\u52bf\u9884\u5224: \u4e24\u79cd\u6a21\u578b\u5747\u9884\u6d4b\u672a\u6765\u6d41\u91cf\u5c06\u5ef6\u7eed\u5f53\u524d\u7684\u589e\u957f\u8d8b\u52bf\u3002\\n\")\n        f.write(\"2. \u98ce\u9669\u7ba1\u7406: \u9884\u6d4b\u533a\u95f4\uff08\u5c24\u5176\u662fProphet\uff09\u63d0\u4f9b\u4e86\u4e0d\u786e\u5b9a\u6027\u4f30\u8ba1\uff0c\u6709\u52a9\u4e8e\u8bc4\u4f30\u98ce\u9669\u3002\\n\")\n        f.write(\"3. \u8d44\u6e90\u89c4\u5212: \\n\")\n        f.write(\"   - \u6839\u636e\u9884\u6d4b\u7684\u6d41\u91cf\u5cf0\u503c\uff08\u5982\u672a\u6765\u5927\u4fc3\u671f\uff09\u63d0\u524d\u6269\u5bb9\u670d\u52a1\u5668\u548c\u5e26\u5bbd\u3002\\n\")\n        f.write(\"   - \u5728\u9884\u6d4b\u6d41\u91cf\u8f83\u4f4e\u7684\u65f6\u671f\u8fdb\u884c\u7cfb\u7edf\u7ef4\u62a4\u548c\u66f4\u65b0\u3002\\n\")\n        f.write(\"4. \u8425\u9500\u7b56\u7565: \u5c06\u8425\u9500\u6d3b\u52a8\u4e0e\u9884\u6d4b\u7684\u6d41\u91cf\u9ad8\u5cf0\u76f8\u7ed3\u5408\uff0c\u4ee5\u6700\u5927\u5316\u6548\u679c\u3002\\n\")\n        f.write(\"5. \u6a21\u578b\u8fed\u4ee3: \u5b9a\u671f\u4f7f\u7528\u6700\u65b0\u7684\u65e5\u5fd7\u6570\u636e\u91cd\u65b0\u8bad\u7ec3\u6a21\u578b\uff0c\u4ee5\u4fdd\u6301\u9884\u6d4b\u7684\u65f6\u6548\u6027\u548c\u51c6\u786e\u6027\u3002\\n\\n\")\n\n        f.write(\"=\" * 60 + \"\\n\")\n        f.write(\"                         \u62a5\u544a\u7ed3\u675f\\n\")\n        f.write(\"=\" * 60 + \"\\n\")\n\n    print(f\"\u7535\u5546\u65e5\u5fd7\u65f6\u5e8f\u9884\u6d4b\u62a5\u544a\u5df2\u751f\u6210: {report_filename}\")\n\n# --- \u4e3b\u51fd\u6570 ---\n\ndef main():\n    \"\"\"\u4e3b\u51fd\u6570\"\"\"\n    # 1. \u751f\u6210\u6570\u636e\n    df_traffic = generate_sample_traffic_data(NUM_DAYS_HISTORY)\n    \n    # 2. \u6570\u636e\u9884\u5904\u7406\u4e0eEDA\n    df_processed = preprocess_and_eda(df_traffic)\n    \n    # 3. ARIMA\u9884\u6d4b\n    arima_results = forecast_with_arima(df_processed, FORECAST_HORIZON)\n    \n    # 4. Prophet\u9884\u6d4b\n    prophet_results = forecast_with_prophet(df_processed, FORECAST_HORIZON)\n    \n    # 5. \u751f\u6210\u62a5\u544a\n    generate_forecast_report(arima_results, prophet_results, FORECAST_HORIZON)\n    \n    print(\"\\n\u7535\u5546\u65e5\u5fd7\u6570\u636e\u65f6\u5e8f\u9884\u6d4b\u5206\u6790\u6d41\u7a0b\u5b8c\u6210\u3002\")\n\nif __name__ == \"__main__\":\n    main()\n<\/code><\/pre>\n","protected":false},"excerpt":{"rendered":"
    \u6211\u4eec\u6765\u521b\u5efa\u4e00\u4e2a\u9488\u5bf9\u7535\u5546\u5e73\u53f0\u65e5\u5fd7\u6570\u636e\u65f6\u5e8f\u9884\u6d4b\u7a0b\u5e8f\u3002\u8fd9\u4e2a\u7a0b\u5e8f\u5c06\u6a21\u62df\u751f\u6210\u5305\u542b\u8bbf\u95ee\u91cf\uff08\u5982PV\u6216UV\uff09\u7684\u65f6\u5e8f\u65e5… \u7ee7\u7eed\u9605\u8bfb \u3010Python\u5b9e\u8df5\u6848\u4f8b\u3011\u7535\u5546\u5e73\u53f0\u6570\u636e\u5206\u6790\u548c\u6316\u6398 – \u65f6\u5e8f\u9884\u6d4b<\/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":395,"_links":{"self":[{"href":"http:\/\/viplao.com\/index.php\/wp-json\/wp\/v2\/posts\/3903"}],"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=3903"}],"version-history":[{"count":2,"href":"http:\/\/viplao.com\/index.php\/wp-json\/wp\/v2\/posts\/3903\/revisions"}],"predecessor-version":[{"id":3917,"href":"http:\/\/viplao.com\/index.php\/wp-json\/wp\/v2\/posts\/3903\/revisions\/3917"}],"wp:attachment":[{"href":"http:\/\/viplao.com\/index.php\/wp-json\/wp\/v2\/media?parent=3903"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"http:\/\/viplao.com\/index.php\/wp-json\/wp\/v2\/categories?post=3903"},{"taxonomy":"post_tag","embeddable":true,"href":"http:\/\/viplao.com\/index.php\/wp-json\/wp\/v2\/tags?post=3903"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}