### 使用PyTorch中的CNN模型处理CSV数据进行回归预测 为了使用卷积神经网络 (CNN) 对 CSV 数据执行回归预测，可以遵循以下方法： #### 准备环境和加载数据 确保安装必要的库并导入所需模块。假设已经有一个名为 `data.csv` 的文件。 ```python import pandas as pd from sklearn.model_selection import train_test_split from sklearn.preprocessing import MinMaxScaler import torch import torch.nn as nn import torch.optim as optim from torch.utils.data import DataLoader, TensorDataset ``` 读取 CSV 文件并将目标列与其他特征分开。这里以温度 (`temp`) 作为要预测的目标变量为例[^1]。 ```python df = pd.read_csv('data.csv') features = df.drop(columns=['temp']) target = df[['temp']] ``` 对输入特征和目标值应用最小最大缩放器(min-max scaling)，以便它们处于相似的比例范围内，这有助于加速训练过程并提高性能。 ```python scaler_features = MinMaxScaler() scaled_features = scaler_features.fit_transform(features) scaler_target = MinMaxScaler() scaled_targets = scaler_target.fit_transform(target) ``` 创建 PyTorch 张量，并划分训练集与测试集。 ```python X_train, X_test, y_train, y_test = train_test_split( scaled_features, scaled_targets, test_size=0.2, random_state=42 ) tensor_X_train = torch.tensor(X_train).float().unsqueeze(1) # 添加通道维度 tensor_y_train = torch.tensor(y_train).float() tensor_X_test = torch.tensor(X_test).float().unsqueeze(1) tensor_y_test = torch.tensor(y_test).float() batch_size = 64 train_loader = DataLoader(TensorDataset(tensor_X_train, tensor_y_train), batch_size=batch_size, shuffle=True) test_loader = DataLoader(TensorDataset(tensor_X_test, tensor_y_test), batch_size=batch_size * 2) ``` 定义 CNN 架构用于回归任务。此架构可以根据具体需求调整层数、滤波器数量和其他超参数。 ```python class CNNet(nn.Module): def __init__(self): super(CNNet, self).__init__() self.conv1d_1 = nn.Conv1d(in_channels=1, out_channels=32, kernel_size=3, padding='same') # 卷积层 self.fc1 = nn.Linear(32 * features.shape[1], 64) # 全连接层 self.fc2 = nn.Linear(64, 1) # 输出层 def forward(self, x): x = torch.relu(self.conv1d_1(x)) x = x.view(x.size(0), -1) # 展平张量 x = torch.relu(self.fc1(x)) output = self.fc2(x) return output ``` 初始化模型实例、损失函数和优化算法。 ```python model = CNNet() criterion = nn.MSELoss() # 均方误差损失 optimizer = optim.Adam(model.parameters(), lr=0.001) # Adam优化器 ``` 编写循环来迭代通过批次更新权重直到满足停止条件为止。 ```python num_epochs = 50 for epoch in range(num_epochs): model.train() running_loss = 0. for inputs, labels in train_loader: optimizer.zero_grad() outputs = model(inputs) loss = criterion(outputs.squeeze(), labels.squeeze()) loss.backward() optimizer.step() running_loss += loss.item() avg_loss = running_loss / len(train_loader) print(f'Epoch [{epoch+1}/{num_epochs}], Loss: {avg_loss:.4f}') ``` 评估模型的表现，在验证集上计算均方根误差(RMSE)或其他合适的度量标准。 ```python model.eval() with torch.no_grad(): predictions = [] actuals = [] for inputs, labels in test_loader: preds = model(inputs) predictions.extend(preds.numpy()) actuals.extend(labels.numpy()) rmse = np.sqrt(mean_squared_error(actuals, predictions)) print(f'Test RMSE: {rmse:.4f}') ``` 最后一步是反向转换预测结果到原始尺度，从而获得最终的预测值。 ```python predicted_values = scaler_target.inverse_transform(np.array(predictions).reshape(-1, 1)) actual_values = scaler_target.inverse_transform(np.array(actuals).reshape(-1, 1)) plt.plot(predicted_values[:100], label="Predictions") plt.plot(actual_values[:100], label="Actual Values") plt.legend() plt.show() ```