基于LS-LSVM的居民出行方式选择预测模型及影响因素研究

Research of Travel Mode Predication Model and Its Influencing Factors Based on LS-SVM

采用最小二乘支持向量机(LS-SVM)构建居民出行方式选择预测模型,并利用遗传算法对支持向量机的参数进行寻优.通过对南京市居民出行调查数据的拟合,得出不论在全方式分类准确率上,还是在单个方式分类准确率上,最小二乘支持向量机相比于多项logistic模型均具有明显的优势;同时基于经济评价的敏感性分析方法也利用在支持向量机模型中,弥补了支持向量机无法显化变量对模型性能的影响及影响程度的不足.

This paper proposes a travel mode predication model based on LS-SVM, adopting Genetic Algorithm to specify parameters. With using the data of resident＇s trip survey in Nanjing, the result shows that LS-SVM model has an evident advantage compared to MNL model on both accuracy rates of all modes classification predication and one mode classification predication. Meanwhile, sensitivity analysis is estimated to investigate what factors affect the performance of model prediction.