Transformers 实现时间序列预测_transformer时间序列预测

使用Transformers实现时间序列预测通常涉及使用预训练的Transformer模型（如BERT、GPT等）来处理时间序列数据。下面是一个简单的示例，演示如何使用Transformers库中的模型来进行时间序列预测。

import torch
import torch.nn as nn
from transformers import BertModel, BertConfig
import numpy as np
import pandas as pd
from sklearn.preprocessing import StandardScaler
from sklearn.model_selection import train_test_split
 
# 创建一个简单的时间序列数据集
# 这里假设时间序列是一个简单的sin函数
np.random.seed(42)
n_points = 1000
X = np.linspace(0, 100, n_points)
y = np.sin(X) + np.random.normal(0, 0.1, n_points)
 
# 划分训练集和测试集
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)
 
# 数据标准化
scaler = StandardScaler()
X_train_scaled = scaler.fit_transform(X_train.reshape(-1, 1))
X_test_scaled = scaler.transform(X_test.reshape(-1, 1))
 
# 转换为PyTorch张量
X_train_tensor = torch.tensor(X_train_scaled, dtype=torch.float32)
X_test_tensor = torch.tensor(X_test_scaled, dtype=torch.float32)
y_train_tensor = torch.tensor(y_train, dtype=torch.float32).unsqueeze(1)  # 添加一个维度以适应模型输入
y_test_tensor = torch.tensor(y_test, dtype=torch.float32).unsqueeze(1)
 
# 定义一个简单的Transformer模型作为时间序列预测器
class TransformerTimeSeriesPredictor(nn.Module):
    def __init__(self, input_dim, output_dim, num_layers=6, hidden_dim=64, n_heads=8):
        super(TransformerTimeSeriesPredictor, self).__init__()
        config = BertConfig(
            hidden_size=hidden_dim,
            num_hidden_layers=num_layers,
            num_attention_heads=n_heads,
            intermediate_size=hidden_dim * 4,
            hidden_dropout_prob=0.1,
            attention_probs_dropout_prob=0.1
        )
        self.encoder = BertModel(config)
        self.fc = nn.Linear(hidden_dim, output_dim)
 
    def forward(self, x):
        _, pooled_output = self.encoder(x)
        output = self.fc(pooled_output)
        return output
 
# 初始化模型并定义优化器和损失函数
model = TransformerTimeSeriesPredictor(input_dim=1, output_dim=1)
optimizer = torch.optim.Adam(model.parameters(), lr=1e-3)
criterion = nn.MSELoss()
 
# 训练模型
num_epochs = 100
for epoch in range(num_epochs):
    model.train()
    optimizer.zero_grad()
    outputs = model(X_train_tensor)
    loss = criterion(outputs, y_train_tensor)
    loss.backward()
    optimizer.step()
    print(f'Epoch [{epoch+1}/{num_epochs}], Loss: {loss.item()}')
 
# 在测试集上进行预测
model.eval()
with torch.no_grad():
    predicted = model(X_test_tensor)
    test_loss = criterion(predicted, y_test_tensor)
    print(f'Test Loss: {test_loss.item()}')
 
# 可视化结果
import matplotlib.pyplot as plt
 
plt.figure(figsize=(10, 5))
plt.plot(X_test, y_test, label='True')
plt.plot(X_test, predicted.numpy(), label='Predicted')
plt.legend()
plt.show()