An Optimized LSTM Passenger Flow Prediction Model for Smart Cities

Abstract

Research on intelligent transportation systems is essential to the development of smart cities. Although it is frequently challenging for urban transportation planners to forecast urban traffic passenger flow due to the massive population flow and complicated flow patterns in cities, this often results in resource waste in urban building. In order to better use resources, this paper proposes the use of machine learning methods to predict the passenger flow of urban subway systems, to help relevant agencies make more rational urban transportation planning. Using a user passenger flow data set of a subway station system, To forecast the number of people using subway stations, LSTM (Long Short-Term Memory) neural networks, random forests, and gradient boosting tree models are created. Model performance testing is done using MAE and MSE to compare how well the three models perform on this task. The experimental findings demonstrate that the LSTM approach outperforms alternative models.