Solutions for improving transportation in South Africa: traffic demand forecasting of public bicycle station based on BP neural network

Abstract

To attain sustainable development, developing countries must focus on the expansion of public transportation systems, this is especially the case in South Africa. The forecasting of the traffic demand for public bicycles is of great significance in optimizing the deployment of vehicles and improving the efficiency of public resources utilization. This paper establishes a traffic demand prediction model using the BP neural network, namely the error back-propagation neural network. The data and network structures of the model were adjusted, and the basic parameters of the model were determined. The international data set was applied to validate the model, and the test results indicate that the BP neural network traffic demand forecasting model outweighs the traditional linear regression prediction method in Mean Squared Error (MSE) and Mean Absolute Percentage Error (MAPE). Finally, the paper offers recommendations for local authorities in South Africa on how to utilize data to effectively improve public transportation system.