Open AccessAPSO-LSTM: An Improved LSTM Neural Network Model Based on APSO Algorithm
Author(s) -
Keqiao Chen
Publication year - 2020
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1651/1/012151
Subject(s) - artificial neural network , computer science , generalization , gradient descent , convergence (economics) , algorithm , mean squared error , artificial intelligence , series (stratigraphy) , optimization algorithm , mathematics , mathematical optimization , statistics , mathematical analysis , paleontology , economics , biology , economic growth
In the LSTM neural network model, the updating of the weights and threshold parameters depends on the gradient descent algorithm. When the number of hidden layers increases, the convergence rate decreases, and the adjustment of the weights may fall into local extremum, which affects the generalization ability and prediction performance of the LSTM model. Based on this, this paper proposes an improved LSTM neural network model based on APSO algorithm (APSO-LSTM). In this model, the root mean square error is designed as the fitness function, and APSO algorithm is used to build the optimization system. In the verification stage, the weight parameters of each neuron are globally optimized to improve the prediction performance of the model. The experimental results on time series datasets of UCI show that the prediction performance of APSO-LSTM model is significantly better than that of the standard LSTM model, which verifies the rationality of the model.