中文摘要：

运用信息挖掘技术最大限度地榨取实测数据所携带的驾驶行为个体有用信息，通过数据修正和精简形成样本数据库，采用K最近邻概率密度函数法对样本数据库进行数据过滤和冗余信息清洗，利用预处理后的数据和非参数回归法构建了驾驶行为非参数仿真模型。模拟得到的后车多元信息与其实际值有很好的拟舍性，且实际值以模拟平均值为轴小幅度摆动。仿真结果表明，合适的光滑参数能提高模型精度，使模型避免大样本标定数据的限制，很好地反映和预测跟驰过程中的驾驶员行为。

英文摘要：

Information mining technology was used to extract useful individual driving information from field measure data, driving sample database was established by amending and condensing field measure data, field measure data and redundant information were filtered by K nearest probability density function estimation method, a new simulation model of driving behavior was established based on nonparametric regression method. It is pointed that the real multi-resource information values of following vehicle are same as the simulated average values, and the field data swing around the simulated ones. Simulation result shows that appropriate smoothing parameter can improve the model accuracy, avoid the calibration limitation of giant sample data, and driving behavior can be reflected and forecasted in car-following process. 1 tab, 5 figs, 10 refs.