L.Y. Yu, S.P. Huang

Pages: 125-136

Abstract
Accurate short-term prediction of campus road traffic flow can help improve traffic efficiency, alleviate traffic pressure, and provide travel guidance. This paper proposes a campus road traffic flow short-term prediction method based on an improved extreme learning machine. Firstly, determine the traffic flow parameters of campus roads, including flow rate, occupancy rate, and speed, and use the average of historical traffic flow as the predicted value of future traffic flow. Secondly, the ARIMA model is used to transform non-stationary short-term traffic flow sequences into stationary sequences. Finally, the improved extreme learning machine is used to extract the periodic features of the target road traffic flow data, and the extracted features are input into a random forest for training to obtain short-term traffic flow prediction results. The experimental results show that the proposed method has higher accuracy, shorter prediction time, and higher prediction recall rate.
Keywords: ARIMA model; random forest; time series data; historical traffic flow; improving extreme learning machines