基于灰色关联的 LS-SVM道路交通事故预测

摘要

In order to improve the predicting accuracy and modeling speed of road traffic accidents,this paper proposed the LS-SVMroad traffic accidents forecast model with grey correlation analysis on the basis of analyzing the influencing factors of road traffic accidents.It screened the input variables of LS-SVMby grey correlation analysis to simplify the LS-SVMstructure, and applied the dynamic change inertia weight adaptive particle swarm optimization algorithm (DCW-APSO)to optimize selec-tion of model parameters.It also applied the model to forecast the integrated mortality rate of road traffic accidents from 1996 to 2000,and compared the results with other type of model.The comparison result shows that compared to other predictive models,the proposed model provides a better convergence rate and higher predicting accuracy.%为提高道路交通事故的预测精度以及建模速度，在分析道路交通事故影响因素基础上，提出了基于灰色关联分析的 LS-SVM道路交通事故预测模型。该模型采用灰色关联分析完成影响因素的相关性分析，结合关联度值，筛选最小二乘向量机模型的输入变量，简化 LS-SVM模型结构；然后运用动态改变惯性权重自适应粒子群算法（DCW-APSO）对模型参数进行优化选取；最后应用模型预测1996—2000年的综合道路交通事故死亡率，并将预测结果与其他模型进行对比分析。结果表明，相较其他预测模型，该模型具有较快的收敛速度，并能明显提高道路交通事故预测的精度。