Short-Term Traffic Flow Prediction Method Based on Dung Beetle Optimizer-Optimized LSTM Network

Abstract

Short-term traffic flow is affected by various factors and usually presents strong nonlinear and stochastic characteristics, which makes accurate prediction difficult. Meanwhile, the selection of LSTM network parameters is often based on manual experience, and an inappropriate parameter setting may affect the prediction performance of the model. Therefore, this study proposes a Dung Beetle Optimizer-based LSTM (DBO-LSTM) model for short-term traffic flow prediction. The LSTM network is used to learn the temporal characteristics of traffic flow, and the DBO algorithm is applied to optimize important parameters of the LSTM network, including the learning rate and the number of hidden-layer neurons. In this way, the parameter selection process can be improved and the dependence on manual parameter adjustment can be reduced. Traffic flow data collected from a real road network are selected for model validation. The proposed DBO-LSTM model is compared with the conventional LSTM, PSO-LSTM, and WOA-LSTM models. The results show that the DBO algorithm can obtain a better parameter combination for the LSTM network. Compared with the three comparison models, the DBO-LSTM model obtains lower RMSE and MAPE values and has better overall prediction performance. The results verify the effectiveness of the proposed model for short-term traffic flow prediction and provide a reference for traffic congestion warning and traffic management in intelligent transportation systems.