S. Qiu, H.L. Jiao, F. Guo, Y.T. Zhou, C.X. Huang

Pages: 169-178

Abstract
The conventional intelligent prediction methods for urban road traffic congestion suffer from drawbacks including limited prediction accuracy and prolonged prediction times. To address these issues, this paper introduces a novel approach for predicting urban road traffic accident losses, leveraging a variable weight TOPSIS method in conjunction with a modified entropy weight scheme. By refining the entropy weight calculation to determine the significance of factors influencing urban road traffic accident losses, and integrating this with the variable weight TOPSIS method, we achieve effective screening of these factors. Subsequently, utilizing these identified factors and the XGBoost algorithm, we conduct predictions of urban road traffic accident losses. Experimental results demonstrate that our method attains a maximum precision of 98.67% in selecting influential factors, and a peak prediction accuracy of 98.26% for urban road traffic accident losses. Furthermore, the prediction time ranges efficiently between 0.29s and 0.71s, underlining the reliability and practicality of our proposed method.
Keywords: variable weight TOPSIS method; modified entropy weight; urban road; traffic accident; loss prediction; XGBoost