论文部分内容阅读
为了提高交通需求预测精度,综合考虑居民出行行为在时间维度上的分布,采用支持向量机、径向基神经网络和多项logit三种方法,基于居民活动目的,建立了出行链模式识别模型,并利用敏感性分析方法研究了解释因素对出行链模式选择的影响和对模型性能的贡献程度.结果显示:支持向量机模型在总体准确度和分类准确度上均优于其他2种方法,体现了支持向量机在小样本下的识别性能优势;证明了支持向量机能够较准确地反映多分类因素对于出行链模式选择行为的影响程度;因素对于不同出行链模式识别精度的贡献度差异表明了细化出行链模式及探索各个模式特有影响因素的重要性.支持向量机技术在交通需求预测建模及影响因素分析方面均具有实践意义.
In order to improve the prediction accuracy of traffic demand, considering the distribution of residents’ travel behavior in the time dimension, three models of support vector machines, RBF neural network and multiple logit were used to establish a traffic pattern recognition model based on the purpose of residential activities. And use the sensitivity analysis method to study the influence of explanatory factors on the choice of travel mode and the contribution to the performance of the model.The results show that SVM model is better than the other two methods in overall accuracy and classification accuracy The recognition performance advantage of support vector machine under small sample is proved. It is proved that support vector machine (SVM) can more accurately reflect the influence degree of multi-classification factors on travel mode selection behavior. The difference of contributions to the identification accuracy of different travel modes shows that Refinement of the mode of travel chain and exploring the importance of each influencing factor of the mode.VSP technology has practical significance in traffic demand forecasting modeling and influential factor analysis.