{"id":6774,"date":"2025-12-20T08:57:38","date_gmt":"2025-12-20T00:57:38","guid":{"rendered":"http:\/\/192.168.1.29\/?p=6774"},"modified":"2025-12-20T08:57:41","modified_gmt":"2025-12-20T00:57:41","slug":"%e8%82%a1%e7%a5%a8%e5%88%86%e6%9e%90%e9%95%bf%e7%9f%ad%e6%9c%9f%e6%97%b6%e9%97%b4%e5%ba%8f%e5%88%97%e6%a8%a1%e5%9e%8blstmpython%e4%bb%a3%e7%a0%81","status":"publish","type":"post","link":"http:\/\/xc.ipyingshe.net:5347\/?p=6774","title":{"rendered":"\u80a1\u7968\u5206\u6790\u957f\u77ed\u671f\u65f6\u95f4\u5e8f\u5217\u6a21\u578bLSTMPython\u4ee3\u7801"},"content":{"rendered":"\n

\u6211\u9700\u8981\u67e5\u770bstock_forcast.py\u6587\u4ef6\u7684\u5b8c\u6574\u5185\u5bb9\uff0c\u4e86\u89e3\u5176\u4e2dLSTM\u6a21\u578b\u7684\u5b9e\u73b0\u548c\u4f7f\u7528\u60c5\u51b5\uff0c\u7136\u540e\u518d\u5bf9\u5176\u8fdb\u884c\u89e3\u91ca\u3002<\/p>\n\n\n\n

filePath: c:\\Users\\czliu\\Documents\\python\\stock_forcast.py<\/p>\n\n\n\n

\u73b0\u5728\u6211\u5df2\u7ecf\u67e5\u770b\u4e86\u5b8c\u6574\u7684stock_forcast.py\u6587\u4ef6\uff0c\u6211\u5c06\u5bf9\u5176\u4e2d\u7684LSTM\u6a21\u578b\u5b9e\u73b0\u8fdb\u884c\u8be6\u7ec6\u89e3\u91ca\uff0c\u5305\u62ec\u6a21\u578b\u5b9a\u4e49\u3001\u7ed3\u6784\u3001\u53c2\u6570\u3001\u6570\u636e\u51c6\u5907\u548c\u8bad\u7ec3\u8fc7\u7a0b\u7b49\u65b9\u9762\u3002# LSTM\u6a21\u578b\u5728\u80a1\u7968\u9884\u6d4b\u7a0b\u5e8f\u4e2d\u7684\u5b9e\u73b0\u5206\u6790<\/p>\n\n\n\n

1. LSTM\u6a21\u578b\u5b9a\u4e49<\/h2>\n\n\n\n

\u5728stock_forcast.py<\/code>\u6587\u4ef6\u4e2d\uff0cLSTM\u6a21\u578b\u901a\u8fc7LSTMStockPredictor<\/code>\u7c7b\u5b9e\u73b0\uff1a<\/p>\n\n\n\n

class LSTMStockPredictor(nn.Module):\n    def __init__(self, input_size=1, hidden_size=64, num_layers=2, output_size=1, dropout=0.2):\n        super(LSTMStockPredictor, self).__init__()\n        self.hidden_size = hidden_size\n        self.num_layers = num_layers\n\n        # LSTM\u5c42\n        self.lstm = nn.LSTM(input_size, hidden_size, num_layers, batch_first=True, dropout=dropout)\n\n        # \u5168\u8fde\u63a5\u5c42\n        self.fc = nn.Linear(hidden_size, output_size)\n\n        # Sigmoid\u6fc0\u6d3b\u51fd\u6570\uff0c\u7528\u4e8e\u5206\u7c7b\n        self.sigmoid = nn.Sigmoid()\n\n    def forward(self, x):\n        # \u521d\u59cb\u5316\u9690\u85cf\u72b6\u6001\u548c\u7ec6\u80de\u72b6\u6001\n        h0 = torch.zeros(self.num_layers, x.size(0), self.hidden_size).to(x.device)\n        c0 = torch.zeros(self.num_layers, x.size(0), self.hidden_size).to(x.device)\n\n        # \u524d\u5411\u4f20\u64adLSTM\n        out, _ = self.lstm(x, (h0, c0))\n\n        # \u53d6\u6700\u540e\u4e00\u4e2a\u65f6\u95f4\u6b65\u7684\u8f93\u51fa\n        out = self.fc(out[:, -1, :])\n\n        # \u4f7f\u7528sigmoid\u6fc0\u6d3b\u51fd\u6570\u5f97\u5230\u6982\u7387\n        out = self.sigmoid(out)\n\n        return out<\/code><\/pre>\n\n\n\n
2. \u6a21\u578b\u7ed3\u6784\u8be6\u89e3<\/h2>\n\n\n\n
2.1 \u6a21\u578b\u53c2\u6570<\/h3>\n\n\n\n
    \n
  • input_size=1<\/code>\uff1a\u8f93\u5165\u7279\u5f81\u7ef4\u5ea6\uff0c\u8fd9\u91cc\u53ea\u4f7f\u7528\u80a1\u7968\u6536\u76d8\u4ef7\u4f5c\u4e3a\u8f93\u5165<\/li>\n\n\n\n
  • hidden_size=64<\/code>\uff1aLSTM\u9690\u85cf\u5c42\u5927\u5c0f\uff0c\u51b3\u5b9a\u6a21\u578b\u7684\u5b66\u4e60\u80fd\u529b<\/li>\n\n\n\n
  • num_layers=2<\/code>\uff1aLSTM\u5806\u53e0\u5c42\u6570\uff0c\u589e\u52a0\u6a21\u578b\u6df1\u5ea6<\/li>\n\n\n\n
  • output_size=1<\/code>\uff1a\u8f93\u51fa\u7ef4\u5ea6\uff0c\u8fd9\u91cc\u8f93\u51fa\u4e00\u4e2a\u6982\u7387\u503c<\/li>\n\n\n\n
  • dropout=0.2<\/code>\uff1a dropout\u6bd4\u7387\uff0c\u9632\u6b62\u8fc7\u62df\u5408<\/li>\n<\/ul>\n\n\n\n
    2.2 \u7f51\u7edc\u7ed3\u6784<\/h3>\n\n\n\n
      \n
    1. LSTM\u5c42<\/strong>\uff1a\u4f7f\u7528PyTorch\u5185\u7f6e\u7684nn.LSTM<\/code>\u5b9e\u73b0\uff0c\u91c7\u7528\u6279\u91cf\u4f18\u5148(batch_first=True<\/code>)\u7684\u8f93\u5165\u683c\u5f0f<\/li>\n\n\n\n
    2. \u5168\u8fde\u63a5\u5c42<\/strong>\uff1a\u5c06LSTM\u7684\u8f93\u51fa\u6620\u5c04\u5230\u6700\u7ec8\u7684\u9884\u6d4b\u503c<\/li>\n\n\n\n
    3. Sigmoid\u6fc0\u6d3b\u51fd\u6570<\/strong>\uff1a\u5c06\u8f93\u51fa\u8f6c\u6362\u4e3a0-1\u4e4b\u95f4\u7684\u6982\u7387\u503c\uff0c\u8868\u793a\u4e0a\u6da8\u7684\u53ef\u80fd\u6027<\/li>\n<\/ol>\n\n\n\n
      2.3 \u524d\u5411\u4f20\u64ad\u8fc7\u7a0b<\/h3>\n\n\n\n
        \n
      1. \u521d\u59cb\u5316\u9690\u85cf\u72b6\u6001(h0<\/code>)\u548c\u7ec6\u80de\u72b6\u6001(c0<\/code>)\u4e3a\u5168\u96f6\u5f20\u91cf<\/li>\n\n\n\n
      2. \u5c06\u8f93\u5165\u6570\u636e\u901a\u8fc7LSTM\u5c42\uff0c\u5f97\u5230\u8f93\u51fa\u5e8f\u5217\u548c\u6700\u7ec8\u7684\u9690\u85cf\u72b6\u6001<\/li>\n\n\n\n
      3. \u53ea\u53d6LSTM\u8f93\u51fa\u5e8f\u5217\u7684\u6700\u540e\u4e00\u4e2a\u65f6\u95f4\u6b65\u7684\u8f93\u51fa\uff0c\u56e0\u4e3a\u6211\u4eec\u53ea\u9700\u8981\u9884\u6d4b\u4e0b\u4e00\u5929\u7684\u8d8b\u52bf<\/li>\n\n\n\n
      4. \u5c06LSTM\u7684\u8f93\u51fa\u901a\u8fc7\u5168\u8fde\u63a5\u5c42\u548cSigmoid\u6fc0\u6d3b\u51fd\u6570\uff0c\u5f97\u5230\u6700\u7ec8\u7684\u9884\u6d4b\u6982\u7387<\/li>\n<\/ol>\n\n\n\n
        3. \u6570\u636e\u51c6\u5907\u4e0e\u9884\u5904\u7406<\/h2>\n\n\n\n
        \u6a21\u578b\u4f7f\u7528prepare_data_for_lstm<\/code>\u51fd\u6570\u5904\u7406\u6570\u636e\uff1a<\/p>\n\n\n\n
        def prepare_data_for_lstm(data, look_back=60, test_size=0.2):\n    # \u4f7f\u7528\u6536\u76d8\u4ef7\u4f5c\u4e3a\u7279\u5f81\n    close_prices = data['Close'].values.reshape(-1, 1)\n\n    # \u6570\u636e\u5f52\u4e00\u5316\n    scaler = MinMaxScaler(feature_range=(0, 1))\n    scaled_data = scaler.fit_transform(close_prices)\n\n    # \u521b\u5efa\u6570\u636e\u96c6\n    X, y = [], []\n    for i in range(len(scaled_data) - look_back):\n        X.append(scaled_data[i:(i + look_back), 0])\n        # \u9884\u6d4b\u4e0b\u4e00\u5929\u662f\u6da8(1)\u8fd8\u662f\u8dcc(0)\n        y.append(1 if scaled_data[i + look_back, 0] > scaled_data[i + look_back - 1, 0] else 0)\n\n    X, y = np.array(X), np.array(y)\n\n    # \u8f6c\u6362\u4e3aLSTM\u9700\u8981\u7684\u5f62\u72b6 [\u6837\u672c\u6570, \u65f6\u95f4\u6b65, \u7279\u5f81\u6570]\n    X = np.reshape(X, (X.shape[0], X.shape[1], 1))\n\n    # \u5212\u5206\u8bad\u7ec3\u96c6\u548c\u6d4b\u8bd5\u96c6\n    X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=test_size, shuffle=False)\n\n    # \u8f6c\u6362\u4e3aPyTorch\u5f20\u91cf\n    X_train = torch.tensor(X_train, dtype=torch.float32)\n    X_test = torch.tensor(X_test, dtype=torch.float32)\n    y_train = torch.tensor(y_train, dtype=torch.float32).unsqueeze(1)\n    y_test = torch.tensor(y_test, dtype=torch.float32).unsqueeze(1)\n\n    return X_train, X_test, y_train, y_test, scaler, len(X_train)<\/code><\/pre>\n\n\n\n
        \u5173\u952e\u5904\u7406\u6b65\u9aa4\uff1a<\/p>\n\n\n\n
          \n
        1. \u4f7f\u7528\u6536\u76d8\u4ef7\u4f5c\u4e3a\u7279\u5f81<\/strong>\uff1a\u4ece\u80a1\u7968\u6570\u636e\u4e2d\u63d0\u53d6\u6536\u76d8\u4ef7<\/li>\n\n\n\n
        2. \u6570\u636e\u5f52\u4e00\u5316<\/strong>\uff1a\u4f7f\u7528MinMaxScaler<\/code>\u5c06\u6570\u636e\u7f29\u653e\u52300-1\u4e4b\u95f4\uff0c\u63d0\u9ad8\u6a21\u578b\u8bad\u7ec3\u7a33\u5b9a\u6027<\/li>\n\n\n\n
        3. \u6784\u5efa\u65f6\u95f4\u5e8f\u5217\u6570\u636e\u96c6<\/strong>\uff1a<\/li>\n<\/ol>\n\n\n\n
            \n
          • \u4f7f\u7528look_back=60<\/code>\u4e2a\u8fde\u7eed\u7684\u6536\u76d8\u4ef7\u4f5c\u4e3a\u8f93\u5165<\/li>\n\n\n\n
          • \u9884\u6d4b\u7b2c61\u5929\u7684\u6da8\u8dcc\u60c5\u51b5\uff081\u8868\u793a\u6da8\uff0c0\u8868\u793a\u8dcc\uff09<\/li>\n<\/ul>\n\n\n\n
              \n
            1. \u8c03\u6574\u6570\u636e\u5f62\u72b6<\/strong>\uff1a\u8f6c\u6362\u4e3aLSTM\u9700\u8981\u7684\u4e09\u7ef4\u683c\u5f0f [\u6837\u672c\u6570, \u65f6\u95f4\u6b65, \u7279\u5f81\u6570]<\/li>\n\n\n\n
            2. \u5212\u5206\u8bad\u7ec3\u96c6\u548c\u6d4b\u8bd5\u96c6<\/strong>\uff1a\u6309\u65f6\u95f4\u987a\u5e8f\u5212\u5206\uff0c\u4fdd\u6301\u65f6\u95f4\u5e8f\u5217\u7279\u6027<\/li>\n\n\n\n
            3. \u8f6c\u6362\u4e3aPyTorch\u5f20\u91cf<\/strong>\uff1a\u9002\u914dPyTorch\u6846\u67b6\u7684\u8f93\u5165\u683c\u5f0f<\/li>\n<\/ol>\n\n\n\n
              4. \u6a21\u578b\u8bad\u7ec3\u8fc7\u7a0b<\/h2>\n\n\n\n
              \u6a21\u578b\u8bad\u7ec3\u901a\u8fc7train_lstm_model<\/code>\u51fd\u6570\u5b9e\u73b0\uff1a<\/p>\n\n\n\n
              def train_lstm_model(model, X_train, y_train, X_test, y_test, epochs=100, batch_size=32, learning_rate=0.001):\n    # \u5b9a\u4e49\u635f\u5931\u51fd\u6570\u548c\u4f18\u5316\u5668\n    criterion = nn.BCELoss()  # \u4e8c\u5143\u4ea4\u53c9\u71b5\u635f\u5931\n    optimizer = torch.optim.Adam(model.parameters(), lr=learning_rate)\n\n    # \u521b\u5efaDataLoader\n    train_dataset = TensorDataset(X_train, y_train)\n    train_loader = DataLoader(train_dataset, batch_size=batch_size, shuffle=True)\n\n    # \u8bad\u7ec3\u6a21\u578b\n    train_losses = []\n    test_losses = []\n\n    for epoch in range(epochs):\n        model.train()\n        train_loss = 0.0\n\n        for batch_X, batch_y in train_loader:\n            # \u524d\u5411\u4f20\u64ad\n            outputs = model(batch_X)\n            loss = criterion(outputs, batch_y)\n\n            # \u53cd\u5411\u4f20\u64ad\u548c\u4f18\u5316\n            optimizer.zero_grad()\n            loss.backward()\n            optimizer.step()\n\n            train_loss += loss.item()\n\n        # \u8ba1\u7b97\u5e73\u5747\u8bad\u7ec3\u635f\u5931\n        train_loss \/= len(train_loader)\n        train_losses.append(train_loss)\n\n        # \u5728\u6d4b\u8bd5\u96c6\u4e0a\u8bc4\u4f30\n        model.eval()\n        with torch.no_grad():\n            test_outputs = model(X_test)\n            test_loss = criterion(test_outputs, y_test).item()\n            test_losses.append(test_loss)\n\n        # \u6253\u5370\u8bad\u7ec3\u8fdb\u5ea6\n        if (epoch + 1) % 10 == 0:\n            print(f'Epoch [{epoch+1}\/{epochs}], Train Loss: {train_loss:.4f}, Test Loss: {test_loss:.4f}')\n\n    return model, train_losses, test_losses<\/code><\/pre>\n\n\n\n
              \u8bad\u7ec3\u5173\u952e\u8bbe\u7f6e\uff1a<\/p>\n\n\n\n
                \n
              1. \u635f\u5931\u51fd\u6570<\/strong>\uff1a\u4f7f\u7528\u4e8c\u5143\u4ea4\u53c9\u71b5\u635f\u5931(BCELoss<\/code>)\uff0c\u9002\u5408\u4e8c\u5206\u7c7b\u95ee\u9898<\/li>\n\n\n\n
              2. \u4f18\u5316\u5668<\/strong>\uff1a\u4f7f\u7528Adam\u4f18\u5316\u5668\uff0c\u5b66\u4e60\u73870.001<\/code><\/li>\n\n\n\n
              3. \u6279\u6b21\u8bad\u7ec3<\/strong>\uff1a\u4f7f\u7528batch_size=32<\/code>\u8fdb\u884c\u5c0f\u6279\u91cf\u8bad\u7ec3<\/li>\n\n\n\n
              4. \u8bad\u7ec3\u8f6e\u6b21<\/strong>\uff1a\u8bad\u7ec3100<\/code>\u4e2aepoch<\/li>\n\n\n\n
              5. \u8bad\u7ec3\u76d1\u63a7<\/strong>\uff1a\u6bcf10\u4e2aepoch\u6253\u5370\u4e00\u6b21\u8bad\u7ec3\u635f\u5931\u548c\u6d4b\u8bd5\u635f\u5931<\/li>\n<\/ol>\n\n\n\n
                5. \u6a21\u578b\u9884\u6d4b\u4e0e\u8bc4\u4f30<\/h2>\n\n\n\n
                5.1 \u9884\u6d4b\u8fc7\u7a0b<\/h3>\n\n\n\n
                def predict_stock_trend(model, X_test, y_test):\n    model.eval()\n    with torch.no_grad():\n        predictions = model(X_test)\n        # \u8f6c\u6362\u4e3a0\/1\u5206\u7c7b\n        predicted_classes = (predictions > 0.5).float()\n\n        # \u8ba1\u7b97\u51c6\u786e\u7387\n        accuracy = (predicted_classes == y_test).float().mean()\n        print(f'\u6a21\u578b\u51c6\u786e\u7387: {accuracy:.4f}')\n\n    return predictions, predicted_classes, accuracy<\/code><\/pre>\n\n\n\n
                \u9884\u6d4b\u6b65\u9aa4\uff1a<\/p>\n\n\n\n
                  \n
                1. \u5c06\u6a21\u578b\u8bbe\u7f6e\u4e3a\u8bc4\u4f30\u6a21\u5f0f(model.eval()<\/code>)<\/li>\n\n\n\n
                2. \u5173\u95ed\u68af\u5ea6\u8ba1\u7b97(with torch.no_grad()<\/code>)\u4ee5\u63d0\u9ad8\u6548\u7387<\/li>\n\n\n\n
                3. \u4f7f\u7528\u6a21\u578b\u8fdb\u884c\u9884\u6d4b\uff0c\u5f97\u5230\u6bcf\u4e2a\u6837\u672c\u7684\u4e0a\u6da8\u6982\u7387<\/li>\n\n\n\n
                4. \u5c06\u6982\u7387\u8f6c\u6362\u4e3a0\/1\u5206\u7c7b\u7ed3\u679c\uff08\u6982\u7387>0.5\u4e3a\u6da8\uff0c\u5426\u5219\u4e3a\u8dcc\uff09<\/li>\n\n\n\n
                5. \u8ba1\u7b97\u6a21\u578b\u5728\u6d4b\u8bd5\u96c6\u4e0a\u7684\u51c6\u786e\u7387<\/li>\n<\/ol>\n\n\n\n
                  5.2 \u7ed3\u679c\u53ef\u89c6\u5316<\/h3>\n\n\n\n
                  \u7a0b\u5e8f\u63d0\u4f9b\u4e86plot_prediction_results<\/code>\u51fd\u6570\uff0c\u7528\u4e8e\u53ef\u89c6\u5316\u9884\u6d4b\u7ed3\u679c\uff0c\u5305\u62ec\uff1a<\/p>\n\n\n\n
                    \n
                  • \u80a1\u7968\u5b9e\u9645\u6536\u76d8\u4ef7\u548c\u6d4b\u8bd5\u96c6\u5206\u5272\u70b9<\/li>\n\n\n\n
                  • \u6a21\u578b\u9884\u6d4b\u7684\u6da8\u8dcc\u8d8b\u52bf<\/li>\n<\/ul>\n\n\n\n
                    6. LSTM\u6a21\u578b\u7684\u5e94\u7528\u7279\u70b9<\/h2>\n\n\n\n
                      \n
                    1. \u65f6\u95f4\u5e8f\u5217\u5efa\u6a21<\/strong>\uff1aLSTM\u7279\u522b\u9002\u5408\u5904\u7406\u65f6\u95f4\u5e8f\u5217\u6570\u636e\uff0c\u80fd\u591f\u6355\u6349\u80a1\u7968\u4ef7\u683c\u7684\u65f6\u5e8f\u7279\u5f81<\/li>\n\n\n\n
                    2. \u957f\u671f\u4f9d\u8d56\u6355\u83b7<\/strong>\uff1a\u901a\u8fc7\u95e8\u63a7\u673a\u5236\uff0cLSTM\u80fd\u591f\u8bb0\u4f4f\u957f\u671f\u7684\u4ef7\u683c\u6a21\u5f0f<\/li>\n\n\n\n
                    3. \u4e8c\u5206\u7c7b\u4efb\u52a1<\/strong>\uff1a\u8fd9\u91ccLSTM\u88ab\u7528\u4e8e\u9884\u6d4b\u80a1\u7968\u6da8\u8dcc\u7684\u4e8c\u5206\u7c7b\u4efb\u52a1\uff0c\u800c\u975e\u76f4\u63a5\u9884\u6d4b\u4ef7\u683c<\/li>\n\n\n\n
                    4. \u6ed1\u52a8\u7a97\u53e3\u8f93\u5165<\/strong>\uff1a\u4f7f\u752860\u5929\u7684\u5386\u53f2\u6570\u636e\u9884\u6d4b\u4e0b\u4e00\u5929\u7684\u8d8b\u52bf\uff0c\u7b26\u5408\u80a1\u7968\u5206\u6790\u7684\u5e38\u7528\u65b9\u6cd5<\/li>\n<\/ol>\n\n\n\n
                      7. \u7a0b\u5e8f\u4e2d\u7684LSTM\u5de5\u4f5c\u6d41\u7a0b<\/h2>\n\n\n\n
                        \n
                      1. \u6570\u636e\u83b7\u53d6<\/strong>\uff1a\u4eceakshare\u83b7\u53d6\u80a1\u7968\u5386\u53f2\u6570\u636e<\/li>\n\n\n\n
                      2. \u6570\u636e\u9884\u5904\u7406<\/strong>\uff1a\u8ba1\u7b97\u6280\u672f\u6307\u6807\uff0c\u51c6\u5907LSTM\u8f93\u5165\u6570\u636e<\/li>\n\n\n\n
                      3. \u6a21\u578b\u521d\u59cb\u5316<\/strong>\uff1a\u521b\u5efaLSTM\u6a21\u578b\u5b9e\u4f8b<\/li>\n\n\n\n
                      4. \u6a21\u578b\u8bad\u7ec3<\/strong>\uff1a\u4f7f\u7528\u8bad\u7ec3\u96c6\u8bad\u7ec3LSTM\u6a21\u578b<\/li>\n\n\n\n
                      5. \u6a21\u578b\u8bc4\u4f30<\/strong>\uff1a\u5728\u6d4b\u8bd5\u96c6\u4e0a\u8bc4\u4f30\u6a21\u578b\u6027\u80fd<\/li>\n\n\n\n
                      6. \u7ed3\u679c\u53ef\u89c6\u5316<\/strong>\uff1a\u7ed8\u5236\u4ef7\u683c\u8d70\u52bf\u56fe\u548c\u9884\u6d4b\u7ed3\u679c\u56fe<\/li>\n<\/ol>\n\n\n\n
                        \u8fd9\u4e2aLSTM\u6a21\u578b\u5b9e\u73b0\u4e86\u4e00\u4e2a\u57fa\u7840\u7684\u80a1\u7968\u8d8b\u52bf\u9884\u6d4b\u7cfb\u7edf\uff0c\u901a\u8fc7\u6df1\u5ea6\u5b66\u4e60\u7684\u65b9\u6cd5\u5c1d\u8bd5\u6355\u6349\u80a1\u7968\u4ef7\u683c\u7684\u77ed\u671f\u6ce2\u52a8\u6a21\u5f0f\u3002<\/p>\n\n\n\n
                        \u5f00\u59cb\u9884\u6d4b \u4e2d\u56fd\u6d77\u6cb9 600938<\/h1>\n\n\n\n
                        \u8bf7\u8f93\u5165\u80a1\u7968\u4ee3\u7801\uff08\u5982 000001 \u6216 600000\uff09\uff1a600938
                        \u5f00\u59cb\u8bad\u7ec3LSTM\u6a21\u578b\u2026
                        Epoch [10\/100], Train Loss: 0.6936, Test Loss: 0.6958
                        Epoch [20\/100], Train Loss: 0.6930, Test Loss: 0.6958
                        Epoch [30\/100], Train Loss: 0.6936, Test Loss: 0.6907
                        Epoch [40\/100], Train Loss: 0.6924, Test Loss: 0.6894
                        Epoch [50\/100], Train Loss: 0.6922, Test Loss: 0.6881
                        Epoch [60\/100], Train Loss: 0.6915, Test Loss: 0.6836
                        Epoch [70\/100], Train Loss: 0.6917, Test Loss: 0.6800
                        Epoch [80\/100], Train Loss: 0.6924, Test Loss: 0.6816
                        Epoch [90\/100], Train Loss: 0.6882, Test Loss: 0.7051
                        Epoch [100\/100], Train Loss: 0.6904, Test Loss: 0.6809
                        \u6a21\u578b\u51c6\u786e\u7387: 0.5676<\/p>\n\n\n\n
                        \u5b8c\u6574\u4ee3\u7801\uff1a<\/h1>\n\n\n\n
                                  \n# \u7a0b\u5e8f\u7528\u9014\u53ca\u795e\u7ecf\u7f51\u7edc\u9884\u6d4b\u7ed3\u679c\u89e3\u91ca\n\n## \u4e00\u3001\u7a0b\u5e8f\u4e3b\u8981\u7528\u9014\n\n\u8be5Python\u7a0b\u5e8f\u662f\u4e00\u4e2a\u7efc\u5408\u7684\u80a1\u7968\u5206\u6790\u5de5\u5177\uff0c\u4e3b\u8981\u529f\u80fd\u5305\u62ec\uff1a\n\n1. **\u80a1\u7968\u6570\u636e\u83b7\u53d6**\uff1a\u4f7f\u7528akshare\u5e93\u4ece\u516c\u5f00\u6570\u636e\u6e90\u83b7\u53d6A\u80a1\u80a1\u7968\u7684\u5386\u53f2\u4ea4\u6613\u6570\u636e\uff08\u9ed8\u8ba4\u83b7\u53d6\u8fd1\u4e00\u5e74\u7684\u65e5\u7ebf\u6570\u636e\uff09\u3002\n\n2. **\u6280\u672f\u6307\u6807\u8ba1\u7b97**\uff1a\n   - \u8ba1\u7b97\u6307\u6570\u79fb\u52a8\u5e73\u5747\u7ebf\uff08EMA\uff0c\u9ed8\u8ba420\u65e5\uff09\n   - \u8ba1\u7b97\u4ef7\u683c\u6ce2\u52a8\u6807\u51c6\u5dee\uff08STD\uff0c\u9ed8\u8ba420\u65e5\uff09\n\n3. **\u6280\u672f\u5206\u6790\u53ef\u89c6\u5316**\uff1a\n   - \u7ed8\u5236\u80a1\u7968\u6536\u76d8\u4ef7\u4e0eEMA\u6307\u6807\u5bf9\u6bd4\u56fe\n   - \u7ed8\u5236\u80a1\u7968\u4ef7\u683c\u6ce2\u52a8\u6027\uff08STD\uff09\u8d70\u52bf\u56fe\n\n4. **LSTM\u795e\u7ecf\u7f51\u7edc\u9884\u6d4b**\uff1a\n   - \u4f7f\u7528\u957f\u77ed\u671f\u8bb0\u5fc6\u7f51\u7edc\uff08LSTM\uff09\u9884\u6d4b\u80a1\u7968\u6b21\u65e5\u6da8\u8dcc\u8d8b\u52bf\n   - \u8bad\u7ec3\u6a21\u578b\u5e76\u8bc4\u4f30\u9884\u6d4b\u51c6\u786e\u7387\n   - \u53ef\u89c6\u5316\u9884\u6d4b\u7ed3\u679c\u4e0e\u5b9e\u9645\u4ef7\u683c\u8d70\u52bf\u5bf9\u6bd4\n\n## \u4e8c\u3001\u795e\u7ecf\u7f51\u7edc\u9884\u6d4b\u7ed3\u679c\u89e3\u91ca\n\n### 1. \u8bad\u7ec3\u8fc7\u7a0b\u6307\u6807\n- **Epoch**\uff1a\u8bad\u7ec3\u8f6e\u6b21\uff0c\u7a0b\u5e8f\u9ed8\u8ba4\u8bad\u7ec3100\u8f6e\n- **Train Loss**\uff1a\u8bad\u7ec3\u96c6\u635f\u5931\uff0c\u53cd\u6620\u6a21\u578b\u5728\u8bad\u7ec3\u6570\u636e\u4e0a\u7684\u62df\u5408\u7a0b\u5ea6\n- **Test Loss**\uff1a\u6d4b\u8bd5\u96c6\u635f\u5931\uff0c\u53cd\u6620\u6a21\u578b\u5728\u672a\u89c1\u8fc7\u6570\u636e\u4e0a\u7684\u6cdb\u5316\u80fd\u529b\n\n\u4ece\u60a8\u63d0\u4f9b\u7684\u8bad\u7ec3\u65e5\u5fd7\u770b\uff1a\n```\nEpoch [90\/100], Train Loss: 0.6398, Test Loss: 0.8248 \nEpoch [100\/100], Train Loss: 0.6300, Test Loss: 0.9487\n```\n\u8bad\u7ec3\u635f\u5931\u9010\u6e10\u4e0b\u964d\uff0c\u4f46\u6d4b\u8bd5\u635f\u5931\u5728\u540e\u671f\u6709\u6240\u4e0a\u5347\uff0c\u8bf4\u660e\u6a21\u578b\u53ef\u80fd\u51fa\u73b0\u4e86\u4e00\u5b9a\u7a0b\u5ea6\u7684\u8fc7\u62df\u5408\uff08\u5728\u8bad\u7ec3\u6570\u636e\u4e0a\u8868\u73b0\u597d\uff0c\u4f46\u5bf9\u65b0\u6570\u636e\u9884\u6d4b\u80fd\u529b\u4e0b\u964d\uff09\u3002\n\n### 2. \u6a21\u578b\u51c6\u786e\u7387\n```\n\u6a21\u578b\u51c6\u786e\u7387: 0.5135\n```\n\u51c6\u786e\u7387\u7ea6\u4e3a51.35%\uff0c\u8868\u793a\u6a21\u578b\u9884\u6d4b\u6b63\u786e\u7684\u6bd4\u4f8b\u3002\u8fd9\u4e2a\u51c6\u786e\u7387\u7565\u9ad8\u4e8e\u968f\u673a\u731c\u6d4b\uff0850%\uff09\uff0c\u8bf4\u660e\u6a21\u578b\u6709\u4e00\u5b9a\u7684\u9884\u6d4b\u80fd\u529b\uff0c\u4f46\u6548\u679c\u5e76\u4e0d\u7406\u60f3\u3002\n\n### 3. \u9884\u6d4b\u7ed3\u679c\u53ef\u89c6\u5316\n- **\u4e0a\u534a\u90e8\u5206\u56fe\u8868**\uff1a\u663e\u793a\u80a1\u7968\u5b9e\u9645\u6536\u76d8\u4ef7\u8d70\u52bf\uff0c\u5e76\u6807\u6ce8\u6d4b\u8bd5\u96c6\u5f00\u59cb\u4f4d\u7f6e\n- **\u4e0b\u534a\u90e8\u5206\u56fe\u8868**\uff1a\u663e\u793a\u6a21\u578b\u5bf9\u6d4b\u8bd5\u96c6\u7684\u9884\u6d4b\u7ed3\u679c\uff080\u8868\u793a\u8dcc\uff0c1\u8868\u793a\u6da8\uff09\n\n### 4. \u6a21\u578b\u5c40\u9650\u6027\u4e0e\u6539\u8fdb\u65b9\u5411\n- **\u6570\u636e\u91cf\u9650\u5236**\uff1a\u4ec5\u4f7f\u7528\u4e86\u8fd1\u4e00\u5e74\u7684\u65e5\u7ebf\u6570\u636e\uff0c\u53ef\u8003\u8651\u589e\u52a0\u66f4\u957f\u65f6\u95f4\u5468\u671f\u7684\u6570\u636e\n- **\u7279\u5f81\u5355\u4e00**\uff1a\u76ee\u524d\u4ec5\u4f7f\u7528\u6536\u76d8\u4ef7\u4f5c\u4e3a\u7279\u5f81\uff0c\u53ef\u52a0\u5165\u6210\u4ea4\u91cf\u3001\u6362\u624b\u7387\u7b49\u66f4\u591a\u6307\u6807\n- **\u6a21\u578b\u7ed3\u6784**\uff1a\u53ef\u5c1d\u8bd5\u8c03\u6574LSTM\u5c42\u6570\u3001\u9690\u85cf\u5355\u5143\u6570\u91cf\u6216\u6dfb\u52a0\u66f4\u591a\u7f51\u7edc\u5c42\n- **\u8bad\u7ec3\u53c2\u6570**\uff1a\u53ef\u8c03\u6574\u5b66\u4e60\u7387\u3001\u6279\u91cf\u5927\u5c0f\u3001\u8bad\u7ec3\u8f6e\u6b21\u7b49\u53c2\u6570\n- **\u8fc7\u62df\u5408\u95ee\u9898**\uff1a\u53ef\u6dfb\u52a0\u6b63\u5219\u5316\u6280\u672f\uff08\u5982L2\u6b63\u5219\u5316\u3001Dropout\uff09\u6216\u4f7f\u7528\u65e9\u505c\u6cd5\n\n## \u4e09\u3001\u4f7f\u7528\u5efa\u8bae\n\n1. **\u6570\u636e\u9009\u62e9**\uff1a\u5efa\u8bae\u9009\u62e9\u4ea4\u6613\u6d3b\u8dc3\u3001\u6570\u636e\u8d28\u91cf\u597d\u7684\u80a1\u7968\u8fdb\u884c\u5206\u6790\n2. **\u53c2\u6570\u8c03\u6574**\uff1a\u6839\u636e\u4e0d\u540c\u80a1\u7968\u7279\u6027\u8c03\u6574\u6280\u672f\u6307\u6807\u53c2\u6570\uff08\u5982EMA\u7a97\u53e3\u5927\u5c0f\uff09\u548cLSTM\u6a21\u578b\u53c2\u6570\n3. **\u7ed3\u679c\u89e3\u8bfb**\uff1a\u80a1\u7968\u4ef7\u683c\u53d7\u591a\u79cd\u56e0\u7d20\u5f71\u54cd\uff0c\u6a21\u578b\u9884\u6d4b\u4ec5\u4f9b\u53c2\u8003\uff0c\u4e0d\u5e94\u4f5c\u4e3a\u6295\u8d44\u51b3\u7b56\u7684\u552f\u4e00\u4f9d\u636e\n4. **\u6301\u7eed\u4f18\u5316**\uff1a\u53ef\u6839\u636e\u5b9e\u9645\u9884\u6d4b\u6548\u679c\u6301\u7eed\u8c03\u6574\u6a21\u578b\u7ed3\u6784\u548c\u53c2\u6570\n\n\u8be5\u7a0b\u5e8f\u63d0\u4f9b\u4e86\u4e00\u4e2a\u5b8c\u6574\u7684\u80a1\u7968\u5206\u6790\u6846\u67b6\uff0c\u60a8\u53ef\u4ee5\u5728\u6b64\u57fa\u7840\u4e0a\u6839\u636e\u81ea\u5df1\u7684\u9700\u6c42\u8fdb\u4e00\u6b65\u6269\u5c55\u529f\u80fd\u6216\u4f18\u5316\u6a21\u578b\u3002\n\n ```{python}\n import pandas as pd\nimport numpy as np\nimport matplotlib.pyplot as plt\n# \u8bbe\u7f6e\u4e2d\u6587\u663e\u793a\nplt.rcParams[\"font.family\"] = [\"SimHei\", \"Microsoft YaHei\", \"SimSun\", \"KaiTi\", \"FangSong\"]\nplt.rcParams[\"axes.unicode_minus\"] = False\nimport akshare as ak\nfrom datetime import datetime, timedelta\nimport torch\nimport torch.nn as nn\nfrom sklearn.preprocessing import MinMaxScaler\nfrom sklearn.model_selection import train_test_split\nfrom torch.utils.data import DataLoader, TensorDataset\n\n\ndef fetch_stock_data(symbol, period='1y'):\n    \"\"\"\u83b7\u53d6\u80a1\u7968\u6570\u636e\"\"\"\n    end_date = datetime.now().strftime('%Y%m%d')\n    start_date = (datetime.now() - timedelta(days=365)).strftime('%Y%m%d')\n    \n    # \u5904\u7406\u80a1\u7968\u4ee3\u7801\u683c\u5f0f\uff0c\u786e\u4fdd\u7b26\u5408akshare\u8981\u6c42\n    if not symbol.startswith(('sh', 'sz')):\n        if symbol.startswith(('6', '9')):\n            symbol = f\"sh{symbol}\"\n        else:\n            symbol = f\"sz{symbol}\"\n    \n    # \u4f7f\u7528\u66f4\u7a33\u5b9a\u7684stock_zh_a_hist\u51fd\u6570\u83b7\u53d6\u80a1\u7968\u65e5\u7ebf\u6570\u636e\n    data = ak.stock_zh_a_hist(symbol=symbol[2:],  # \u53ea\u4f20\u80a1\u7968\u4ee3\u7801\u90e8\u5206\uff0c\u4e0d\u542bsh\/sz\n                             period=\"daily\",\n                             start_date=start_date,\n                             end_date=end_date,\n                             adjust=\"qfq\")  # \u4f7f\u7528\u524d\u590d\u6743\u4ef7\u683c\n    \n    # \u68c0\u67e5\u6570\u636e\u83b7\u53d6\u662f\u5426\u6210\u529f\n    if data.empty:\n        print(f\"\u83b7\u53d6\u80a1\u7968{symbol}\u6570\u636e\u5931\u8d25\uff0c\u8bf7\u68c0\u67e5\u80a1\u7968\u4ee3\u7801\u548c\u7f51\u7edc\u8fde\u63a5\")\n        return None\n    \n    # \u5148\u8f6c\u6362\u65e5\u671f\u683c\u5f0f\n    if '\u65e5\u671f' in data.columns:\n        data['\u65e5\u671f'] = pd.to_datetime(data['\u65e5\u671f'])\n        # \u8bbe\u7f6e\u65e5\u671f\u4e3a\u7d22\u5f15\n        data.set_index('\u65e5\u671f', inplace=True)\n        # \u91cd\u547d\u540d\u6536\u76d8\u4ef7\u5217\n        if '\u6536\u76d8' in data.columns:\n            data = data.rename(columns={'\u6536\u76d8': 'Close'})\n    else:\n        print(\"\u6570\u636e\u683c\u5f0f\u5f02\u5e38\uff0c\u65e0\u6cd5\u5904\u7406\")\n        return None\n    \n    return data\n\n\ndef calculate_ema(data, window=20):\n    \"\"\"\u8ba1\u7b97\u6307\u6570\u79fb\u52a8\u5e73\u5747\u7ebf EMA\"\"\"\n    ema = data['Close'].ewm(span=window, adjust=False).mean()\n    return ema\n\n\ndef calculate_std(data, window=20):\n    \"\"\"\u8ba1\u7b97\u6807\u51c6\u5dee STD\"\"\"\n    std = data['Close'].rolling(window=window).std()\n    return std\n\n\ndef plot_analysis(data, ema, std, symbol):\n    \"\"\"\u7ed8\u5236\u56fe\u8868\u5206\u6790\u7ed3\u679c\"\"\"\n    plt.figure(figsize=(14, 8))\n\n    plt.subplot(2, 1, 1)\n    plt.plot(data['Close'], label='\u6536\u76d8\u4ef7', color='blue')\n    plt.plot(ema, label='EMA(20)', color='orange')\n    plt.title(f'{symbol} \u80a1\u7968\u4ef7\u683c\u4e0eEMA\u5206\u6790')\n    plt.legend()\n\n    plt.subplot(2, 1, 2)\n    plt.plot(std, label='STD(20)', color='red')\n    plt.title(f'{symbol} \u80a1\u7968\u4ef7\u683c\u6ce2\u52a8\u6027(STD)\u5206\u6790')\n    plt.legend()\n\n    plt.tight_layout()\n    plt.show()\n\n\n# LSTM\u6a21\u578b\u5b9a\u4e49\nclass LSTMStockPredictor(nn.Module):\n    def __init__(self, input_size=1, hidden_size=64, num_layers=2, output_size=1, dropout=0.2):\n        super(LSTMStockPredictor, self).__init__()\n        self.hidden_size = hidden_size\n        self.num_layers = num_layers\n        \n        # LSTM\u5c42\n        self.lstm = nn.LSTM(input_size, hidden_size, num_layers, batch_first=True, dropout=dropout)\n        \n        # \u5168\u8fde\u63a5\u5c42\n        self.fc = nn.Linear(hidden_size, output_size)\n        \n        # Sigmoid\u6fc0\u6d3b\u51fd\u6570\uff0c\u7528\u4e8e\u5206\u7c7b\n        self.sigmoid = nn.Sigmoid()\n    \n    def forward(self, x):\n        # \u521d\u59cb\u5316\u9690\u85cf\u72b6\u6001\u548c\u7ec6\u80de\u72b6\u6001\n        h0 = torch.zeros(self.num_layers, x.size(0), self.hidden_size).to(x.device)\n        c0 = torch.zeros(self.num_layers, x.size(0), self.hidden_size).to(x.device)\n        \n        # \u524d\u5411\u4f20\u64adLSTM\n        out, _ = self.lstm(x, (h0, c0))\n        \n        # \u53d6\u6700\u540e\u4e00\u4e2a\u65f6\u95f4\u6b65\u7684\u8f93\u51fa\n        out = self.fc(out[:, -1, :])\n        \n        # \u4f7f\u7528sigmoid\u6fc0\u6d3b\u51fd\u6570\u5f97\u5230\u6982\u7387\n        out = self.sigmoid(out)\n        \n        return out\n\n\ndef prepare_data_for_lstm(data, look_back=60, test_size=0.2):\n    \"\"\"\u51c6\u5907LSTM\u6a21\u578b\u7684\u6570\u636e\"\"\"\n    # \u4f7f\u7528\u6536\u76d8\u4ef7\u4f5c\u4e3a\u7279\u5f81\n    close_prices = data['Close'].values.reshape(-1, 1)\n    \n    # \u6570\u636e\u5f52\u4e00\u5316\n    scaler = MinMaxScaler(feature_range=(0, 1))\n    scaled_data = scaler.fit_transform(close_prices)\n    \n    # \u521b\u5efa\u6570\u636e\u96c6\n    X, y = [], []\n    for i in range(len(scaled_data) - look_back):\n        X.append(scaled_data[i:(i + look_back), 0])\n        # \u9884\u6d4b\u4e0b\u4e00\u5929\u662f\u6da8(1)\u8fd8\u662f\u8dcc(0)\n        y.append(1 if scaled_data[i + look_back, 0] > scaled_data[i + look_back - 1, 0] else 0)\n    \n    X, y = np.array(X), np.array(y)\n    \n    # \u8f6c\u6362\u4e3aLSTM\u9700\u8981\u7684\u5f62\u72b6 [\u6837\u672c\u6570, \u65f6\u95f4\u6b65, \u7279\u5f81\u6570]\n    X = np.reshape(X, (X.shape[0], X.shape[1], 1))\n    \n    # \u5212\u5206\u8bad\u7ec3\u96c6\u548c\u6d4b\u8bd5\u96c6\n    X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=test_size, shuffle=False)\n    \n    # \u8f6c\u6362\u4e3aPyTorch\u5f20\u91cf\n    X_train = torch.tensor(X_train, dtype=torch.float32)\n    X_test = torch.tensor(X_test, dtype=torch.float32)\n    y_train = torch.tensor(y_train, dtype=torch.float32).unsqueeze(1)\n    y_test = torch.tensor(y_test, dtype=torch.float32).unsqueeze(1)\n    \n    return X_train, X_test, y_train, y_test, scaler, len(X_train)\n\n\ndef train_lstm_model(model, X_train, y_train, X_test, y_test, epochs=100, batch_size=32, learning_rate=0.001):\n    \"\"\"\u8bad\u7ec3LSTM\u6a21\u578b\"\"\"\n    # \u5b9a\u4e49\u635f\u5931\u51fd\u6570\u548c\u4f18\u5316\u5668\n    criterion = nn.BCELoss()  # \u4e8c\u5143\u4ea4\u53c9\u71b5\u635f\u5931\n    optimizer = torch.optim.Adam(model.parameters(), lr=learning_rate)\n    \n    # \u521b\u5efaDataLoader\n    train_dataset = TensorDataset(X_train, y_train)\n    train_loader = DataLoader(train_dataset, batch_size=batch_size, shuffle=True)\n    \n    # \u8bad\u7ec3\u6a21\u578b\n    train_losses = []\n    test_losses = []\n    \n    for epoch in range(epochs):\n        model.train()\n        train_loss = 0.0\n        \n        for batch_X, batch_y in train_loader:\n            # \u524d\u5411\u4f20\u64ad\n            outputs = model(batch_X)\n            loss = criterion(outputs, batch_y)\n            \n            # \u53cd\u5411\u4f20\u64ad\u548c\u4f18\u5316\n            optimizer.zero_grad()\n            loss.backward()\n            optimizer.step()\n            \n            train_loss += loss.item()\n        \n        # \u8ba1\u7b97\u5e73\u5747\u8bad\u7ec3\u635f\u5931\n        train_loss \/= len(train_loader)\n        train_losses.append(train_loss)\n        \n        # \u5728\u6d4b\u8bd5\u96c6\u4e0a\u8bc4\u4f30\n        model.eval()\n        with torch.no_grad():\n            test_outputs = model(X_test)\n            test_loss = criterion(test_outputs, y_test).item()\n            test_losses.append(test_loss)\n        \n        # \u6253\u5370\u8bad\u7ec3\u8fdb\u5ea6\n        if (epoch + 1) % 10 == 0:\n            print(f'Epoch [{epoch+1}\/{epochs}], Train Loss: {train_loss:.4f}, Test Loss: {test_loss:.4f}')\n    \n    return model, train_losses, test_losses\n\n\ndef predict_stock_trend(model, X_test, y_test):\n    \"\"\"\u4f7f\u7528\u8bad\u7ec3\u597d\u7684\u6a21\u578b\u9884\u6d4b\u80a1\u7968\u8d8b\u52bf\"\"\"\n    model.eval()\n    with torch.no_grad():\n        predictions = model(X_test)\n        # \u8f6c\u6362\u4e3a0\/1\u5206\u7c7b\n        predicted_classes = (predictions > 0.5).float()\n        \n        # \u8ba1\u7b97\u51c6\u786e\u7387\n        accuracy = (predicted_classes == y_test).float().mean()\n        print(f'\u6a21\u578b\u51c6\u786e\u7387: {accuracy:.4f}')\n    \n    return predictions, predicted_classes, accuracy\n\n\ndef plot_prediction_results(data, predicted_classes, look_back, n_train):\n    \"\"\"\u7ed8\u5236\u9884\u6d4b\u7ed3\u679c\"\"\"\n    plt.figure(figsize=(14, 8))\n    \n    # \u83b7\u53d6\u6d4b\u8bd5\u96c6\u5bf9\u5e94\u7684\u539f\u59cb\u6570\u636e\n    test_start_index = n_train + look_back\n    test_data = data.iloc[test_start_index:]\n    \n    # \u521b\u5efa\u9884\u6d4b\u7ed3\u679c\u7684DataFrame\n    prediction_df = pd.DataFrame(index=test_data.index)\n    prediction_df['Actual'] = data['Close'].iloc[test_start_index:]\n    prediction_df['Predicted_Trend'] = predicted_classes.numpy().flatten()\n    \n    # \u7ed8\u5236\u6536\u76d8\u4ef7\u548c\u9884\u6d4b\u8d8b\u52bf\n    plt.subplot(2, 1, 1)\n    plt.plot(data['Close'], label='\u5b9e\u9645\u6536\u76d8\u4ef7', color='blue')\n    if not test_data.empty:\n        plt.axvline(x=test_data.index[0], color='red', linestyle='--', label='\u6d4b\u8bd5\u96c6\u5f00\u59cb')\n    plt.title('\u80a1\u7968\u4ef7\u683c\u4e0e\u6d4b\u8bd5\u96c6\u5206\u5272')\n    plt.legend()\n    \n    # \u7ed8\u5236\u9884\u6d4b\u8d8b\u52bf\n    plt.subplot(2, 1, 2)\n    if not test_data.empty:\n        plt.plot(test_data.index, prediction_df['Predicted_Trend'], label='\u9884\u6d4b\u6da8\u8dcc', color='green', marker='o', linestyle='')\n    plt.title('LSTM\u6a21\u578b\u9884\u6d4b\u80a1\u7968\u6da8\u8dcc')\n    plt.yticks([0, 1], ['\u8dcc', '\u6da8'])\n    plt.legend()\n    \n    plt.tight_layout()\n    plt.show()\n\n\ndef main():\n    symbol = input(\"\u8bf7\u8f93\u5165\u80a1\u7968\u4ee3\u7801\uff08\u5982 000001 \u6216 600000\uff09\uff1a\")\n    data = fetch_stock_data(symbol)\n    \n    # \u68c0\u67e5\u6570\u636e\u83b7\u53d6\u662f\u5426\u6210\u529f\n    if data is not None:\n        # \u8ba1\u7b97\u6280\u672f\u6307\u6807\n        ema = calculate_ema(data)\n        std = calculate_std(data)\n        \n        # \u7ed8\u5236\u6280\u672f\u5206\u6790\u56fe\u8868\n        plot_analysis(data, ema, std, symbol)\n        \n        # \u51c6\u5907LSTM\u6a21\u578b\u6570\u636e\n        look_back = 60\n        test_size = 0.2\n        X_train, X_test, y_train, y_test, scaler, n_train = prepare_data_for_lstm(data, look_back, test_size)\n        \n        # \u521d\u59cb\u5316\u6a21\u578b\n        model = LSTMStockPredictor(input_size=1, hidden_size=64, num_layers=2, output_size=1, dropout=0.2)\n        \n        # \u8bad\u7ec3\u6a21\u578b\n        print(\"\u5f00\u59cb\u8bad\u7ec3LSTM\u6a21\u578b...\")\n        model, train_losses, test_losses = train_lstm_model(model, X_train, y_train, X_test, y_test, epochs=100, batch_size=32, learning_rate=0.001)\n        \n        # \u9884\u6d4b\u80a1\u7968\u8d8b\u52bf\n        predictions, predicted_classes, accuracy = predict_stock_trend(model, X_test, y_test)\n        \n        # \u7ed8\u5236\u9884\u6d4b\u7ed3\u679c\n        plot_prediction_results(data, predicted_classes, look_back, n_train)\n        \n        # \u7ed8\u5236\u635f\u5931\u66f2\u7ebf\n        plt.figure(figsize=(10, 6))\n        plt.plot(range(1, len(train_losses) + 1), train_losses, label='\u8bad\u7ec3\u635f\u5931')\n        plt.plot(range(1, len(test_losses) + 1), test_losses, label='\u6d4b\u8bd5\u635f\u5931')\n        plt.xlabel('Epoch')\n        plt.ylabel('Loss')\n        plt.title('LSTM\u6a21\u578b\u8bad\u7ec3\u635f\u5931\u66f2\u7ebf')\n        plt.legend()\n        plt.show()\n\n\nif __name__ == '__main__':\n    main()\n\n ```\n\n\n# \u95ee\u8c46\u5305<\/strong>\uff1aPython\u80a1\u7968\u5206\u6790\uff0cakshare\u83b7\u53d6\u6570\u636e\uff0c\u53ef\u89c6\u5316\u957f\u5ea6\u50b2\u6c14\u8bb0\u5fc6\u7f51\u7edc\uff08LSTM\uff09\uff0c\u795e\u7ecf\u6a21\u578b\u91c7\u7528pytorch\u3002\u9884\u6d4b\u80a1\u7968\u6da8\u8dcc\u200b<\/code><\/pre>\n","protected":false},"excerpt":{"rendered":"
                        \u6211\u9700\u8981\u67e5\u770bstock_forcast.py\u6587\u4ef6\u7684\u5b8c\u6574\u5185\u5bb9\uff0c\u4e86\u89e3\u5176\u4e2dLSTM\u6a21\u578b\u7684 Continue Reading<\/a><\/span><\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"closed","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[2,24],"tags":[],"class_list":["post-6774","post","type-post","status-publish","format-standard","hentry","category-2","category-24"],"_links":{"self":[{"href":"http:\/\/xc.ipyingshe.net:5347\/index.php?rest_route=\/wp\/v2\/posts\/6774","targetHints":{"allow":["GET"]}}],"collection":[{"href":"http:\/\/xc.ipyingshe.net:5347\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"http:\/\/xc.ipyingshe.net:5347\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"http:\/\/xc.ipyingshe.net:5347\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"http:\/\/xc.ipyingshe.net:5347\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=6774"}],"version-history":[{"count":1,"href":"http:\/\/xc.ipyingshe.net:5347\/index.php?rest_route=\/wp\/v2\/posts\/6774\/revisions"}],"predecessor-version":[{"id":6775,"href":"http:\/\/xc.ipyingshe.net:5347\/index.php?rest_route=\/wp\/v2\/posts\/6774\/revisions\/6775"}],"wp:attachment":[{"href":"http:\/\/xc.ipyingshe.net:5347\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=6774"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"http:\/\/xc.ipyingshe.net:5347\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=6774"},{"taxonomy":"post_tag","embeddable":true,"href":"http:\/\/xc.ipyingshe.net:5347\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=6774"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}