麻烦提供基于pytorch的EEMD+STL+XGBoost+LightGBM+ConvLSTM的读取csv数据的多输入单输出的时序训练及预测代码

以下是基于PyTorch的EEMD-STL-XGBoost-LightGBM-ConvLSTM的多输入单输出时序训练和预测代码。这段代码实现了从CSV文件中读取多个时序输入的数据，然后使用EEMD和STL进行数据去噪和降维处理，然后使用XGBoost和LightGBM进行特征提取和预测，最后使用ConvLSTM模型进行时序预测。

import numpy as np
import pandas as pd
import torch
import torch.nn as nn
import torch.optim as optim
from torch.utils.data import Dataset, DataLoader
from sklearn.model_selection import train_test_split
from xgboost import XGBRegressor
from lightgbm import LGBMRegressor
from pytorch_forecasting import TimeSeriesDataSet
from pytorch_forecasting.models import LSTM, DeepAR
from pytorch_forecasting.metrics import SMAPE, RMSE
from statsmodels.tsa.seasonal import seasonal_decompose
from PyEMD import EMD

# Define constants
BATCH_SIZE = 64
LEARNING_RATE = 0.001
EPOCHS = 100
HORIZON = 24  # Number of time steps to predict
LOOKBACK = 168  # Number of time steps to look back
NUM_EMD_COMPONENTS = 5  # Number of EMD components to keep
NUM_STL_COMPONENTS = 3  # Number of STL components to keep

# Read data from CSV file
data = pd.read_csv('data.csv')
num_features = data.shape[1] - 1  # Number of input features

# Split data into training and validation sets
train_data, val_data = train_test_split(data, test_size=0.2)

# Define EEMD and STL transformers
class EEMDTransformer:
    def __init__(self, num_components):
        self.num_components = num_components
        self.emd = EMD()

    def fit_transform(self, X):
        X_transformed = np.zeros((X.shape[0], self.num_components, X.shape[1]))
        for i in range(X.shape[0]):
            imfs = self.emd(X[i])
            imfs = imfs[:self.num_components]
            for j in range(self.num_components):
                X_transformed[i, j, :] = imfs[j]
        return X_transformed

class STLTransformer:
    def __init__(self, num_components):
        self.num_components = num_components

    def fit_transform(self, X):
        X_transformed = np.zeros((X.shape[0], self.num_components, X.shape[1]))
        for i in range(X.shape[0]):
            result = seasonal_decompose(X[i], model='additive', period=24)
            for j in range(self.num_components):
                X_transformed[i, j, :] = result.seasonal[j*X.shape[1]:(j+1)*X.shape[1]]
        return X_transformed

# Define PyTorch dataset and data loader for training and validation sets
class TimeSeriesDataset(Dataset):
    def __init__(self, data, lookback, horizon, num_emd_components, num_stl_components):
        self.data = data.values[:, 1:]
        self.lookback = lookback
        self.horizon = horizon
        self.emd_transformer = EEMDTransformer(num_emd_components)
        self.stl_transformer = STLTransformer(num_stl_components)
        self.scaler = None

    def __len__(self):
        return