基于transformer模型预测流感数据 结合平均高温和平均低温去预测流感 代码示例 进行预测值和真实值的对比图

以下是一个基于transformer模型预测流感数据的代码示例，其中结合平均高温和平均低温去预测流感，并输出预测值和真实值的对比图：

import torch
import torch.nn as nn
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
from sklearn.preprocessing import MinMaxScaler

# 加载数据
data = pd.read_csv('flu_data.csv')

# 数据预处理
scaler = MinMaxScaler()
data['flu'] = scaler.fit_transform(data['flu'].values.reshape(-1, 1))
data['avg_high_temp'] = scaler.fit_transform(data['avg_high_temp'].values.reshape(-1, 1))
data['avg_low_temp'] = scaler.fit_transform(data['avg_low_temp'].values.reshape(-1, 1))

# 划分训练集和测试集
train_size = int(len(data) * 0.8)
train_data = data[:train_size]
test_data = data[train_size:]

# 定义模型
class TransformerModel(nn.Module):
    def __init__(self, input_size, output_size, nhead, num_layers, hidden_size, dropout):
        super().__init__()
        self.transformer_encoder = nn.TransformerEncoder(nn.TransformerEncoderLayer(input_size, nhead, hidden_size, dropout), num_layers)
        self.fc = nn.Linear(input_size, output_size)

    def forward(self, src):
        src = self.transformer_encoder(src)
        out = self.fc(src[-1, :, :])
        return out

# 训练模型
def train_model(model, train_data, optimizer, criterion, num_epochs):
    for epoch in range(num_epochs):
        train_loss = 0.0
        for i in range(len(train_data)-1):
            optimizer.zero_grad()
            input_data = torch.tensor(train_data[['avg_high_temp', 'avg_low_temp', 'flu']].iloc[i:i+1].values, dtype=torch.float32).transpose(0, 1)
            target_data = torch.tensor(train_data['flu'].iloc[i+1:i+2].values, dtype=torch.float32)
            output_data = model(input_data)
            loss = criterion(output_data, target_data)
            loss.backward()
            optimizer.step()
            train_loss += loss.item()
        print('Epoch [{}/{}], Loss: {:.4f}'.format(epoch+1, num_epochs, train_loss/len(train_data)))

# 测试模型
def test_model(model, test_data):
    model.eval()
    predictions = []
    with torch.no_grad():
        for i in range(len(test_data)-1):
            input_data = torch.tensor(test_data[['avg_high_temp', 'avg_low_temp', 'flu']].iloc[i:i+1].values, dtype=torch.float32).transpose(0, 1)
            output_data = model(input_data)
            predictions.append(output_data.item())
    predictions = scaler.inverse_transform(np.array(predictions).reshape(-1, 1))
    actuals = scaler.inverse_transform(test_data['flu'].iloc[1:].values.reshape(-1, 1))
    return predictions, actuals

# 设置参数和训练模型
input_size = 3
output_size = 1
nhead = 2
num_layers = 2
hidden_size = 128
dropout = 0.2
num_epochs = 100
learning_rate = 0.001

model = TransformerModel(input_size, output_size, nhead, num_layers, hidden_size, dropout)
criterion = nn.MSELoss()
optimizer = torch.optim.Adam(model.parameters(), lr=learning_rate)

train_model(model, train_data, optimizer, criterion, num_epochs)

# 测试模型并绘制对比图
predictions, actuals = test_model(model, test_data)
plt.plot(predictions, label='Predictions')
plt.plot(actuals, label='Actuals')
plt.legend()
plt.show()

在运行上述代码后，将得到一个预测值和真实值的对比图，如下所示：