{"id":419,"date":"2026-03-06T13:32:19","date_gmt":"2026-03-06T05:32:19","guid":{"rendered":"https:\/\/www.liaoxinghui.com\/?p=419"},"modified":"2026-03-06T13:32:19","modified_gmt":"2026-03-06T05:32:19","slug":"lstm-time-series-prediction-server-traffic","status":"publish","type":"post","link":"https:\/\/www.liaoxinghui.com\/?p=419","title":{"rendered":"\u4f7f\u7528LSTM\u795e\u7ecf\u7f51\u7edc\u9884\u6d4b\u670d\u52a1\u5668\u672a\u676524\u5c0f\u65f6\u6d41\u91cf\uff1a\u4ece\u6570\u636e\u5230\u90e8\u7f72\u7684\u5b8c\u6574\u5b9e\u6218"},"content":{"rendered":"

\u4f7f\u7528LSTM\u795e\u7ecf\u7f51\u7edc\u9884\u6d4b\u670d\u52a1\u5668\u672a\u676524\u5c0f\u65f6\u6d41\u91cf\uff1a\u4ece\u6570\u636e\u5230\u90e8\u7f72\u7684\u5b8c\u6574\u5b9e\u6218<\/h1>\n

1. \u4e1a\u52a1\u573a\u666f\u4e0e\u76ee\u6807<\/h2>\n

\u4e92\u8054\u7f51\u516c\u53f8\u8fd0\u7ef4\u56e2\u961f\u9700\u8981\u63d0\u524d\u9884\u6d4b\u670d\u52a1\u5668\u672a\u676524\u5c0f\u65f6\u7684\u6d41\u91cf\u53d8\u5316\uff0c\u4ee5\u4fbf\u8fdb\u884c\u5bb9\u91cf\u89c4\u5212\u548c\u8d44\u6e90\u8c03\u914d\u3002\u57fa\u4e8e\u8fc7\u53bb7\u5929\u7684\u5386\u53f2\u6d41\u91cf\u6570\u636e\uff08\u6bcf\u5c0f\u65f6\u4e00\u4e2a\u70b9\uff09\uff0c\u8bad\u7ec3\u4e00\u4e2aLSTM\u795e\u7ecf\u7f51\u7edc\u6a21\u578b\uff0c\u8f93\u51fa\u672a\u676524\u5c0f\u65f6\u7684\u6d41\u91cf\u9884\u6d4b\u503c\u3002\u76ee\u6807\u662f\u51cf\u5c11\u56e0\u6d41\u91cf\u7a81\u589e\u5bfc\u81f4\u7684\u670d\u52a1\u5668\u8fc7\u8f7d\u98ce\u9669\u3002<\/p>\n

2. \u73af\u5883\u51c6\u5907\uff08uv + \u4f9d\u8d56\uff09<\/h2>\n

\u4f7f\u7528uv\u521b\u5efa\u865a\u62df\u73af\u5883\u5e76\u5b89\u88c5\u4f9d\u8d56\uff1a<\/p>\n

uv venv lstm-env\nsource lstm-env\/bin\/activate  # Windows: lstm-env\\Scripts\\activate\nuv pip install torch==2.3.0 numpy==1.26.4 pandas==2.2.2 matplotlib==3.8.4 scikit-learn==1.5.0<\/code><\/pre>\n
3. \u6570\u636e\u8bf4\u660e\uff08\u771f\u5b9e\u6570\u636e\u53e3\u5f84\u6216\u6a21\u62df\u6570\u636e\u751f\u6210\u903b\u8f91\uff09<\/h2>\n
\u771f\u5b9e\u573a\u666f\u4e2d\uff0c\u6570\u636e\u6765\u81ea\u670d\u52a1\u5668\u76d1\u63a7\u7cfb\u7edf\uff08\u5982Prometheus\uff09\uff0c\u5305\u542b\u65f6\u95f4\u6233\u548c\u6d41\u91cf\u503c\uff08\u5355\u4f4d\uff1aMB\/s\uff09\u3002\u4e3a\u65b9\u4fbf\u6f14\u793a\uff0c\u6211\u4eec\u6a21\u62df\u751f\u62107\u5929\uff08168\u5c0f\u65f6\uff09\u7684\u6d41\u91cf\u6570\u636e\uff0c\u5305\u542b\u65e5\u5468\u671f\u6027\u548c\u968f\u673a\u6ce2\u52a8\uff1a<\/p>\n
import numpy as np\nimport pandas as pd\nfrom datetime import datetime, timedelta\n\n# \u751f\u6210\u6a21\u62df\u6570\u636e\nnp.random.seed(42)\nbase_traffic = 100  # \u57fa\u7840\u6d41\u91cf\nhours = 168  # 7\u5929*24\u5c0f\u65f6\n\n# \u751f\u6210\u65f6\u95f4\u5e8f\u5217\uff1a\u65e5\u5468\u671f\u6027 + \u968f\u673a\u566a\u58f0\ntimestamps = [datetime(2024, 1, 1) + timedelta(hours=i) for i in range(hours)]\ntraffic = []\nfor i in range(hours):\n    hour_of_day = i % 24\n    # \u65e5\u5468\u671f\uff1a\u767d\u5929\u6d41\u91cf\u9ad8\uff0c\u591c\u95f4\u4f4e\n    daily_pattern = 20 * np.sin(2 * np.pi * hour_of_day \/ 24)\n    # \u968f\u673a\u6ce2\u52a8\n    noise = np.random.normal(0, 5)\n    traffic.append(base_traffic + daily_pattern + noise)\n\ndf = pd.DataFrame({'timestamp': timestamps, 'traffic': traffic})\nprint(f"\u6570\u636e\u5f62\u72b6: {df.shape}")\nprint(df.head())<\/code><\/pre>\n
4. \u8bad\u7ec3\/\u5b9e\u73b0\u6b65\u9aa4\uff08\u5b8c\u6574\u4ee3\u7801\uff09<\/h2>\n
\u4efb\u52a1\u7c7b\u578b<\/strong>\uff1a\u65f6\u95f4\u5e8f\u5217\u56de\u5f52\u9884\u6d4b\uff08\u9884\u6d4b\u672a\u676524\u5c0f\u65f6\u6d41\u91cf\u503c\uff09\u3002<\/p>\n
\u5b8c\u6574\u8bad\u7ec3\u4ee3\u7801\uff1a<\/p>\n
import torch\nimport torch.nn as nn\nimport torch.optim as optim\nfrom sklearn.preprocessing import MinMaxScaler\nfrom sklearn.model_selection import train_test_split\nimport matplotlib.pyplot as plt\n\n# \u6570\u636e\u9884\u5904\u7406\nscaler = MinMaxScaler()\ndf['traffic_scaled'] = scaler.fit_transform(df[['traffic']])\n\n# \u521b\u5efa\u5e8f\u5217\u6570\u636e\uff1a\u7528\u8fc7\u53bb24\u5c0f\u65f6\u9884\u6d4b\u672a\u676524\u5c0f\u65f6\ndef create_sequences(data, seq_length, pred_length):\n    X, y = [], []\n    for i in range(len(data) - seq_length - pred_length + 1):\n        X.append(data[i:i+seq_length])\n        y.append(data[i+seq_length:i+seq_length+pred_length])\n    return np.array(X), np.array(y)\n\nseq_length = 24  # \u8f93\u5165\u5e8f\u5217\u957f\u5ea6\npred_length = 24  # \u9884\u6d4b\u957f\u5ea6\nX, y = create_sequences(df['traffic_scaled'].values, seq_length, pred_length)\n\n# \u5212\u5206\u8bad\u7ec3\u96c6\u548c\u9a8c\u8bc1\u96c6\uff0880%\u8bad\u7ec3\uff0c20%\u9a8c\u8bc1\uff09\nX_train, X_val, y_train, y_val = train_test_split(X, y, test_size=0.2, shuffle=False)\n\n# \u8f6c\u6362\u4e3aPyTorch\u5f20\u91cf\nX_train = torch.FloatTensor(X_train).unsqueeze(-1)  # \u5f62\u72b6: (\u6837\u672c\u6570, seq_length, 1)\ny_train = torch.FloatTensor(y_train)\nX_val = torch.FloatTensor(X_val).unsqueeze(-1)\ny_val = torch.FloatTensor(y_val)\n\n# \u5b9a\u4e49LSTM\u6a21\u578b\nclass LSTMModel(nn.Module):\n    def __init__(self, input_size=1, hidden_size=50, num_layers=2, output_size=24):\n        super().__init__()\n        self.lstm = nn.LSTM(input_size, hidden_size, num_layers, batch_first=True)\n        self.fc = nn.Linear(hidden_size, output_size)\n\n    def forward(self, x):\n        lstm_out, _ = self.lstm(x)  # lstm_out\u5f62\u72b6: (batch, seq_len, hidden_size)\n        # \u53d6\u6700\u540e\u4e00\u4e2a\u65f6\u95f4\u6b65\u7684\u8f93\u51fa\n        last_time_step = lstm_out[:, -1, :]\n        output = self.fc(last_time_step)\n        return output\n\n# \u521d\u59cb\u5316\u6a21\u578b\u3001\u635f\u5931\u51fd\u6570\u548c\u4f18\u5316\u5668\nmodel = LSTMModel()\ncriterion = nn.MSELoss()\noptimizer = optim.Adam(model.parameters(), lr=0.001)\n\n# \u8bad\u7ec3\u6a21\u578b\nepochs = 100\ntrain_losses = []\nval_losses = []\n\nfor epoch in range(epochs):\n    model.train()\n    optimizer.zero_grad()\n    outputs = model(X_train)\n    loss = criterion(outputs, y_train)\n    loss.backward()\n    optimizer.step()\n    train_losses.append(loss.item())\n\n    # \u9a8c\u8bc1\n    model.eval()\n    with torch.no_grad():\n        val_outputs = model(X_val)\n        val_loss = criterion(val_outputs, y_val)\n        val_losses.append(val_loss.item())\n\n    if (epoch+1) % 20 == 0:\n        print(f'Epoch [{epoch+1}\/{epochs}], Train Loss: {loss.item():.4f}, Val Loss: {val_loss.item():.4f}')\n\n# \u7ed8\u5236\u8bad\u7ec3\u548c\u9a8c\u8bc1\u635f\u5931\u66f2\u7ebf\nplt.figure(figsize=(10, 5))\nplt.plot(train_losses, label='Train Loss')\nplt.plot(val_losses, label='Validation Loss')\nplt.xlabel('Epoch')\nplt.ylabel('Loss (MSE)')\nplt.title('Training and Validation Loss')\nplt.legend()\nplt.grid(True)\nplt.show()\n\n# \u4fdd\u5b58\u6a21\u578b\ntorch.save(model.state_dict(), 'lstm_traffic_model.pth')\nprint("\u6a21\u578b\u5df2\u4fdd\u5b58\u4e3a 'lstm_traffic_model.pth'")<\/code><\/pre>\n
5. \u8c03\u7528\u65b9\u5f0f<\/h2>\n
\u79bb\u7ebf\u6279\u91cf\u9884\u6d4b<\/strong>\uff08\u5bf9\u6574\u4e2a\u9a8c\u8bc1\u96c6\u8fdb\u884c\u9884\u6d4b\u5e76\u53ef\u89c6\u5316\uff09\uff1a<\/p>\n
# \u52a0\u8f7d\u6a21\u578b\nmodel = LSTMModel()\nmodel.load_state_dict(torch.load('lstm_traffic_model.pth'))\nmodel.eval()\n\n# \u5bf9\u9a8c\u8bc1\u96c6\u8fdb\u884c\u9884\u6d4b\nwith torch.no_grad():\n    predictions = model(X_val)\n\n# \u53cd\u6807\u51c6\u5316\npredictions_np = predictions.numpy()\ny_val_np = y_val.numpy()\n\n# \u53cd\u6807\u51c6\u5316\u51fd\u6570\ndef inverse_transform(scaled_data, scaler, original_shape):\n    return scaler.inverse_transform(scaled_data.reshape(-1, 1)).reshape(original_shape)\n\n# \u53cd\u6807\u51c6\u5316\u9884\u6d4b\u503c\u548c\u771f\u5b9e\u503c\npredictions_original = inverse_transform(predictions_np, scaler, predictions_np.shape)\ny_val_original = inverse_transform(y_val_np, scaler, y_val_np.shape)\n\n# \u53ef\u89c6\u5316\u7b2c\u4e00\u4e2a\u6837\u672c\u7684\u9884\u6d4b\u7ed3\u679c\nsample_idx = 0\nplt.figure(figsize=(12, 6))\nplt.plot(range(24), y_val_original[sample_idx], label='Actual Traffic', marker='o')\nplt.plot(range(24), predictions_original[sample_idx], label='Predicted Traffic', marker='x')\nplt.xlabel('Hour Ahead')\nplt.ylabel('Traffic (MB\/s)')\nplt.title('24-Hour Traffic Prediction vs Actual')\nplt.legend()\nplt.grid(True)\nplt.show()<\/code><\/pre>\n
\u5355\u6761\u5b9e\u65f6\u9884\u6d4b\u793a\u4f8b<\/strong>\uff08\u57fa\u4e8e\u6700\u65b024\u5c0f\u65f6\u6570\u636e\u9884\u6d4b\u672a\u676524\u5c0f\u65f6\uff09\uff1a<\/p>\n
def predict_next_24h(recent_24h_data):\n    """\n    recent_24h_data: \u5217\u8868\u6216\u6570\u7ec4\uff0c\u6700\u8fd124\u5c0f\u65f6\u7684\u6d41\u91cf\u6570\u636e\uff08\u539f\u59cb\u503c\uff0c\u672a\u6807\u51c6\u5316\uff09\n    \u8fd4\u56de\uff1a\u672a\u676524\u5c0f\u65f6\u7684\u9884\u6d4b\u6d41\u91cf\u503c\uff08\u539f\u59cb\u503c\uff09\n    """\n    # \u6807\u51c6\u5316\u8f93\u5165\u6570\u636e\n    recent_scaled = scaler.transform(np.array(recent_24h_data).reshape(-1, 1)).flatten()\n\n    # \u8f6c\u6362\u4e3a\u6a21\u578b\u8f93\u5165\u683c\u5f0f\n    input_tensor = torch.FloatTensor(recent_scaled).unsqueeze(0).unsqueeze(-1)  # \u5f62\u72b6: (1, 24, 1)\n\n    # \u9884\u6d4b\n    model.eval()\n    with torch.no_grad():\n        prediction_scaled = model(input_tensor).numpy().flatten()\n\n    # \u53cd\u6807\u51c6\u5316\n    prediction = scaler.inverse_transform(prediction_scaled.reshape(-1, 1)).flatten()\n    return prediction\n\n# \u793a\u4f8b\uff1a\u4f7f\u7528\u6700\u540e24\u5c0f\u65f6\u6570\u636e\u9884\u6d4b\nlast_24h = df['traffic'].values[-24:]\npredicted_traffic = predict_next_24h(last_24h)\nprint("\u672a\u676524\u5c0f\u65f6\u9884\u6d4b\u6d41\u91cf:", predicted_traffic)<\/code><\/pre>\n
6. \u6307\u6807\u8bf4\u660e<\/h2>\n
    \n
  • LSTM\uff08\u957f\u77ed\u671f\u8bb0\u5fc6\u7f51\u7edc\uff09<\/strong>\uff1a\u4e00\u79cd\u7279\u6b8a\u7684\u5faa\u73af\u795e\u7ecf\u7f51\u7edc\uff0c\u80fd\u8bb0\u4f4f\u957f\u671f\u4f9d\u8d56\u5173\u7cfb\uff0c\u9002\u5408\u5904\u7406\u65f6\u95f4\u5e8f\u5217\u6570\u636e\u3002<\/li>\n
  • \u65f6\u95f4\u5e8f\u5217<\/strong>\uff1a\u6309\u65f6\u95f4\u987a\u5e8f\u6392\u5217\u7684\u6570\u636e\u70b9\u5e8f\u5217\uff0c\u5982\u6bcf\u5c0f\u65f6\u6d41\u91cf\u8bb0\u5f55\u3002<\/li>\n
  • \u5e8f\u5217\u957f\u5ea6\uff08seq_length\uff09<\/strong>\uff1a\u6a21\u578b\u8f93\u5165\u7684\u5386\u53f2\u6570\u636e\u70b9\u6570\uff0c\u8fd9\u91cc\u752824\u5c0f\u65f6\u9884\u6d4b\u672a\u676524\u5c0f\u65f6\u3002<\/li>\n
  • MSE\uff08\u5747\u65b9\u8bef\u5dee\uff09<\/strong>\uff1a\u9884\u6d4b\u503c\u4e0e\u771f\u5b9e\u503c\u5dee\u503c\u7684\u5e73\u65b9\u7684\u5e73\u5747\u503c\uff0c\u8d8a\u5c0f\u8868\u793a\u9884\u6d4b\u8d8a\u51c6\u3002<\/li>\n
  • \u6807\u51c6\u5316<\/strong>\uff1a\u5c06\u6570\u636e\u7f29\u653e\u52300-1\u8303\u56f4\uff0c\u907f\u514d\u5927\u6570\u503c\u5f71\u54cd\u6a21\u578b\u8bad\u7ec3\u3002<\/li>\n
  • \u8bad\u7ec3\u96c6\/\u9a8c\u8bc1\u96c6<\/strong>\uff1a\u8bad\u7ec3\u96c6\u7528\u4e8e\u8bad\u7ec3\u6a21\u578b\u53c2\u6570\uff0c\u9a8c\u8bc1\u96c6\u7528\u4e8e\u8bc4\u4f30\u6a21\u578b\u6027\u80fd\uff0c\u9632\u6b62\u8fc7\u62df\u5408\u3002<\/li>\n
  • \u8fc7\u62df\u5408<\/strong>\uff1a\u6a21\u578b\u5728\u8bad\u7ec3\u96c6\u4e0a\u8868\u73b0\u597d\u4f46\u5728\u65b0\u6570\u636e\u4e0a\u5dee\uff0c\u901a\u5e38\u56e0\u6a21\u578b\u592a\u590d\u6742\u6216\u8bad\u7ec3\u6570\u636e\u4e0d\u8db3\u5bfc\u81f4\u3002<\/li>\n
  • MAE\uff08\u5e73\u5747\u7edd\u5bf9\u8bef\u5dee\uff09<\/strong>\uff1a\u9884\u6d4b\u503c\u4e0e\u771f\u5b9e\u503c\u7edd\u5bf9\u5dee\u503c\u7684\u5e73\u5747\u503c\uff0c\u5355\u4f4d\u4e0e\u6d41\u91cf\u76f8\u540c\uff08MB\/s\uff09\uff0c\u76f4\u89c2\u53cd\u6620\u8bef\u5dee\u5927\u5c0f\u3002<\/li>\n
  • RMSE\uff08\u5747\u65b9\u6839\u8bef\u5dee\uff09<\/strong>\uff1aMSE\u7684\u5e73\u65b9\u6839\uff0c\u5355\u4f4d\u4e0e\u6d41\u91cf\u76f8\u540c\uff0c\u5bf9\u8f83\u5927\u8bef\u5dee\u66f4\u654f\u611f\u3002<\/li>\n<\/ul>\n
    7. \u4e0a\u7ebf\u540e\u8bc4\u4f30<\/h2>\n
    \u6a21\u578b\u4e0a\u7ebf\u540e\uff0c\u9700\u6301\u7eed\u76d1\u63a7\u4ee5\u4e0b\u6307\u6807\uff1a<\/p>\n
      \n
    • \u9884\u6d4b\u51c6\u786e\u7387<\/strong>\uff1a\u8ba1\u7b97MAE\u548cRMSE\uff0c\u4f8b\u5982MAE<10 MB\/s\u53ef\u89c6\u4e3a\u5408\u683c\u3002<\/li>\n
    • \u63a8\u7406\u5ef6\u8fdf<\/strong>\uff1a\u5355\u6b21\u9884\u6d4b\u8017\u65f6\u5e94<100ms\uff0c\u4ee5\u6ee1\u8db3\u5b9e\u65f6\u6027\u8981\u6c42\u3002<\/li>\n
    • \u8d44\u6e90\u4f7f\u7528<\/strong>\uff1a\u76d1\u63a7CPU\/\u5185\u5b58\u5360\u7528\uff0c\u786e\u4fdd\u4e0d\u5f71\u54cd\u670d\u52a1\u5668\u6027\u80fd\u3002\n\u8bc4\u4f30\u65b9\u6cd5\uff1a\u6bcf\u65e5\u5bf9\u6bd4\u9884\u6d4b\u6d41\u91cf\u4e0e\u5b9e\u9645\u6d41\u91cf\uff0c\u8ba1\u7b97\u8bef\u5dee\u6307\u6807\uff1b\u8bbe\u7f6e\u62a5\u8b66\u9608\u503c\uff08\u5982MAE\u8fde\u7eed3\u5929>15 MB\/s\u89e6\u53d1\u544a\u8b66\uff09\u3002<\/li>\n<\/ul>\n
      8. \u5e38\u89c1\u5751\u4e0e\u6392\u67e5<\/h2>\n
        \n
      1. \u6570\u636e\u5b63\u8282\u6027\u5904\u7406\u4e0d\u5f53<\/strong>\uff1a\u5982\u679c\u6d41\u91cf\u6709\u5468\u5468\u671f\uff08\u5982\u5468\u672b\u6d41\u91cf\u4f4e\uff09\uff0c\u9700\u589e\u52a0\u8f93\u5165\u5e8f\u5217\u957f\u5ea6\uff08\u5982168\u5c0f\u65f6\u8986\u76d6\u4e00\u5468\uff09\u3002<\/li>\n
      2. LSTM\u8bad\u7ec3\u65f6\u95f4\u8fc7\u957f<\/strong>\uff1a\u51cf\u5c11\u9690\u85cf\u5c42\u5927\u5c0f\uff08\u5982\u4ece50\u964d\u523030\uff09\u6216\u5c42\u6570\uff08\u5982\u4ece2\u5c42\u964d\u52301\u5c42\uff09\uff0c\u6216\u4f7f\u7528GPU\u52a0\u901f\u3002<\/li>\n
      3. \u8fc7\u62df\u5408\u5bfc\u81f4\u9884\u6d4b\u4e0d\u51c6<\/strong>\uff1a\u589e\u52a0Dropout\u5c42\u3001\u65e9\u505c\uff08\u5f53\u9a8c\u8bc1\u635f\u5931\u4e0d\u518d\u4e0b\u964d\u65f6\u505c\u6b62\u8bad\u7ec3\uff09\u6216\u6536\u96c6\u66f4\u591a\u8bad\u7ec3\u6570\u636e\u3002<\/li>\n
      4. \u751f\u4ea7\u73af\u5883\u90e8\u7f72\u590d\u6742<\/strong>\uff1a\u4f7f\u7528TorchScript\u5c06\u6a21\u578b\u8f6c\u6362\u4e3a\u811a\u672c\u6a21\u5f0f\uff0c\u6216\u901a\u8fc7ONNX\u683c\u5f0f\u90e8\u7f72\u5230\u4e0d\u540c\u5e73\u53f0\u3002\n\u6392\u67e5\u6b65\u9aa4\uff1a\u5148\u68c0\u67e5\u8f93\u5165\u6570\u636e\u683c\u5f0f\u662f\u5426\u6b63\u786e\uff0c\u518d\u9a8c\u8bc1\u6a21\u578b\u8f93\u51fa\u662f\u5426\u5408\u7406\uff0c\u6700\u540e\u76d1\u63a7\u8bef\u5dee\u6307\u6807\u662f\u5426\u5f02\u5e38\u3002<\/li>\n<\/ol>","protected":false},"excerpt":{"rendered":"
        \u672c\u6587\u624b\u628a\u624b\u6559\u4f60\u7528PyTorch\u6784\u5efaLSTM\u6a21\u578b\uff0c\u57fa\u4e8e\u5386\u53f2\u670d\u52a1\u5668\u6d41\u91cf\u6570\u636e\u9884\u6d4b\u672a\u676524\u5c0f\u65f6\u8d8b\u52bf\u3002\u5305\u542b\u6570\u636e\u6a21\u62df\u3001\u6a21\u578b\u8bad\u7ec3\u3001\u8bc4\u4f30\u6307\u6807\u89e3\u91ca\u53ca\u751f\u4ea7\u73af\u5883\u5e38\u89c1\u95ee\u9898\u6392\u67e5\u3002<\/p>","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[7],"tags":[139,140],"class_list":["post-419","post","type-post","status-publish","format-standard","hentry","category-linux","tag-lstm","tag-140"],"views":140,"_links":{"self":[{"href":"https:\/\/www.liaoxinghui.com\/index.php?rest_route=\/wp\/v2\/posts\/419","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.liaoxinghui.com\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.liaoxinghui.com\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.liaoxinghui.com\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.liaoxinghui.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=419"}],"version-history":[{"count":1,"href":"https:\/\/www.liaoxinghui.com\/index.php?rest_route=\/wp\/v2\/posts\/419\/revisions"}],"predecessor-version":[{"id":449,"href":"https:\/\/www.liaoxinghui.com\/index.php?rest_route=\/wp\/v2\/posts\/419\/revisions\/449"}],"wp:attachment":[{"href":"https:\/\/www.liaoxinghui.com\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=419"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.liaoxinghui.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=419"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.liaoxinghui.com\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=419"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}