Time Series Prediction Based on Time Attention Mechanism and LSTM Neural Network
Jiayi Sun, Wenming Guo
Abstract
As a collection of time observations, time series has attracted extensive attention in artificial intelligence. Time series prediction is one of the important topics to obtain future trends. Therefore, based on the discussion of time series characteristics, temporal attention mechanism and deep learning time series prediction, this paper briefly discusses the open data set, experimental environment and parameter settings, and designs an improved time series PA-LSTM prediction model based on deep learning. Finally, through specific experimental analysis. The results show that the RMSLE and MAE values of the PA-LSTM prediction method designed in this paper are 0.012 and 0.010 respectively. The error is lower than other prediction methods. Therefore, the PA-LSTM prediction method has certain advantages.