基于知识的神经网络在出行方式选择中的应用研究

针对神经网络和决策树方法在算法上的本质联系和互补优势，将C4.5决策树提取规则的基于知识的神经网络（knowledgebased neural network，KBNN）用于出行方式预测。对居民通勤出行方式选择数据的分析表明，KBNN相比于决策树方法、普通前馈神经网络和多项Logit模型（MNL）有更高的预测精度，方法不仅提高了网络的可解释性，且易于构造、收敛速度更快，实用性较强，为出行方式选择预测提供了新的思路。

Research of travel mode choice with knowledge based neural network

Based on the similarity between neural network and decision tree,the method of knowledge-based neural network(KBNN) combined the rule induction of decision tree and the accurate approximation of neural network.This research showed how to construct a neural network based on rules from a decision tree generated by C4.5 method.A network built by this method and models based on decision tree,neural network and multinomial Logit(MNL) were specified,estimated and comparatively evaluated.The prediction results show that decision tree and neural network models offer slightly better performance than MNL model and the KBNN model demonstrates highest performance.The analysis of actual investigation data shows that the model has fast convergence and high precision,which is of great importance for travel mode choice prediction.