Research on the Application of Improved LSTM Model in Time Series Problems
Jian Dai, Maojun Liao, Xuanyu Guo · 2023
With the rapid development of deep learning technology, Recurrent neural network has made remarkable achievements in the tasks of sequence modeling and time series prediction. LSTM, as a special RNN structure, solves the long term dependence problem effectively by introducing gating mechanism, and has become one of the widely used models. In order to further improve the performance of LSTM model, this paper have improved some existing problems. The traditional LSTM model only considers the information of a time step, and ignores the context of the future moment. To solve this problem, this paper introduce a multi-step LSTM model. The multi-step LSTM model can better predict the future time series data by selecting the information of multiple time steps. The multi-step LSTM model proposed in this paper has potential application value in the field of time series prediction, and can provide reference for related research.