Traffic Flow Prediction Method Based on Improved Optimization Algorithm Mixed Kernel Extreme Learning Machine

The development of traffic flow time series prediction has been rapid. In this paper, a model combination approach is employed, optimizing and analyzing the applicability of time series prediction models using ETC gantry data. Two hybrid improved strategies, the SLWChoA-HKELM prediction model based on the Gorilla Optimization Algorithm and the MSWOA-HKELM prediction model based on the Whale Optimization Algorithm, are proposed. The results of the case study indicate that both MSWOA-HKELM and SLWChoA-HKELM models demonstrate satisfactory training and prediction capabilities. Among them, the SLWChoA-HKELM model achieves the highest prediction accuracy and requires less time for high-speed highway multi-scenario traffic flow time series prediction.