我是机器学习的新手。我试图开发一个 LSTM 模型来预测数据。模型结构如下。
Model: "sequential_1"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
lstm_3 (LSTM) (None, 1, 50) 11400
_________________________________________________________________
dropout_4 (Dropout) (None, 1, 50) 0
_________________________________________________________________
lstm_4 (LSTM) (None, 1, 50) 20200
_________________________________________________________________
dropout_5 (Dropout) (None, 1, 50) 0
_________________________________________________________________
lstm_5 (LSTM) (None, 1, 50) 20200
_________________________________________________________________
dropout_6 (Dropout) (None, 1, 50) 0
_________________________________________________________________
dropout_7 (Dropout) (None, 1, 50) 0
_________________________________________________________________
dense_1 (Dense) (None, 1, 2) 102
=================================================================
Total params: 51,902
Trainable params: 51,902
Non-trainable params: 0
_________________________________________________________________
我用于训练的部分数据如下（第二到六列是输入，第七和第八列是输出）： Training Data
然而，损失并没有减少，系统的预测是平均值而不是遵循系统的模式。附上两张图片。我不知道如何进一步改进它。 Loss Function Blue is original and orange is predicted
请指导我如何/如何进一步改进它。任何参考资料都会有所帮助！
