CNN-Transformer时间序列预测

部分代码：

# CNN-Transformer
class CNNTransformerEncoder(nn.Module):
    def __init__(self, input_features, transformer_encoder_heads,
                 embedding_features, cnn_kernel_size, dim_feedforward_enc, n_encoder_layer):
        super(CNNTransformerEncoder, self).__init__()
        # input: [batch_size, input_features, input_seq_len]
        # output: [batch_size, embedding_features, output_len(related to kernel_size, padding and stride)]
        self.cnn_embedding = nn.Conv1d(input_features, embedding_features, cnn_kernel_size)  # CNN部分
        self.position_embedding = PositionalEncoder(d_model=embedding_features)  # 位置编码
        transformer_encoder_layer = nn.TransformerEncoderLayer(d_model=embedding_features,
                                                               nhead=transformer_encoder_heads,
                                                               dim_feedforward=dim_feedforward_enc,
                                                               activation='gelu')  # transformer编码器
        self.transformer_encoder = nn.TransformerEncoder(transformer_encoder_layer, num_layers=n_encoder_layer)

    def forward(self, input_seq):
        cnn_embedding_results = self.cnn_embedding(input_seq)  # 输入经过CNN
        embediing_with_position = self.position_embedding(cnn_embedding_results.permute((0, 2, 1)))  # 进行位置编码
        encoder_res = self.transformer_encoder(embediing_with_position.permute((1, 0, 2)))  # 通过transformer encoder
        return encoder_res

项目截图： 

数据：

 

测试集预测对比： 

#完整代码
https://mbd.pub/o/works/592982