💡 LSTM股价预测实战技巧

本章介绍了深度学习在量化中的应用，这里分享LSTM模型的实战技巧：

1. 数据预处理关键

# 使用对数收益率而非原始价格
df['log_returns'] = np.log(df['close'] / df['close'].shift(1))

# 标准化技巧：使用训练集统计量
from sklearn.preprocessing import StandardScaler

scaler = StandardScaler()
train_scaled = scaler.fit_transform(train_data)
test_scaled = scaler.transform(test_data)  # 注意：只用transform

2. 序列构建

def create_sequences(data, seq_len, pred_horizon):
    """创建输入序列和标签"""
    X, y = [], []
    for i in range(len(data) - seq_len - pred_horizon):
        X.append(data[i:i+seq_len])
        # 预测未来第N天的收益率
        y.append(data[i+seq_len+pred_horizon-1])
    return np.array(X), np.array(y)

# 建议参数
seq_len = 20  # 20天历史
pred_horizon = 5  # 预测5天后

3. 模型架构优化

class ImprovedLSTM(nn.Module):
    def __init__(self, input_size, hidden_size=64, num_layers=2):
        super().__init__()
        
        self.lstm = nn.LSTM(
            input_size=input_size,
            hidden_size=hidden_size,
            num_layers=num_layers,
            batch_first=True,
            dropout=0.2,
            bidirectional=True  # 双向LSTM
        )
        
        # 注意力机制
        self.attention = nn.Linear(hidden_size * 2, 1)
        
        self.fc = nn.Sequential(
            nn.Linear(hidden_size * 2, 32),
            nn.ReLU(),
            nn.Dropout(0.3),
            nn.Linear(32, 1)
        )
    
    def forward(self, x):
        lstm_out, _ = self.lstm(x)
        
        # 注意力权重
        attn_weights = torch.softmax(self.attention(lstm_out), dim=1)
        context = torch.sum(attn_weights * lstm_out, dim=1)
        
        return self.fc(context)

4. 防止过拟合

# 早停
from torch.utils.data import DataLoader
from copy import deepcopy

best_loss = float('inf')
patience = 10
counter = 0

for epoch in range(epochs):
    train_loss = train_one_epoch()
    val_loss = validate()
    
    if val_loss < best_loss:
        best_loss = val_loss
        best_model = deepcopy(model.state_dict())
        counter = 0
    else:
        counter += 1
    
    if counter >= patience:
        print(f"Early stopping at epoch {epoch}")
        break

model.load_state_dict(best_model)

5. 评估指标

def evaluate_predictions(predictions, actuals):
    """多维度评估"""
    # 价格预测误差
    mse = np.mean((predictions - actuals)**2)
    mae = np.mean(np.abs(predictions - actuals))
    
    # 方向准确率（更重要！）
    pred_direction = np.diff(predictions) > 0
    actual_direction = np.diff(actuals) > 0
    direction_accuracy = np.mean(pred_direction == actual_direction)
    
    print(f"MSE: {mse:.6f}")
    print(f"MAE: {mae:.6f}")
    print(f"方向准确率: {direction_accuracy:.2%}")
    
    return {'mse': mse, 'mae': mae, 'direction_acc': direction_accuracy}

核心建议：

• 价格预测很难，方向预测更实用
• 方向准确率>55%就有价值
• 结合其他指标（如RSI、MACD）效果更好