中文摘要：

分析了智能交通系统（ITS）的特征使得获取交通流信息的质量和准确性难以保证，且ITS的分析和预测与数据的存在时间远近关联的紧密性。如果不考虑时间关联，将这些含有噪声、冗余、错误或不一致源信息应用到以参数驱动的预测模型，就无法得到全面的分析和精确的预测。研究了引入时间关联因子的曲线拟合对交通流源数据进行预处理和异常检测并解决时间关联性问题。基于ITS已有的流量一时间比例曲线模型，运用最小二乘法曲线拟合原理，提出了一种加入时间关联因子曲线拟合的交通流异常挖掘方法，并运用分箱思想设定交通系统动态正常数据范围，从而在曲线拟合的基础上剔除异常数据，最后用实例验证了拟合效果及该方法对异常交通流数据的有效识别。

英文摘要：

It is analyzed that the features of ITS make the quality and accuracy of traffic flow information difficult to guarantee, and the relationship of the analysis and prediction of ITS with data on the time presence of distance is very tight. If we do not consider the time correlation, these information of contain noise, redundancy, error or inconsistency application to parameter driven prediction model, is unable to obtain the comprehensive analysis and accurate prediction. Studying preprocess and anomaly detect traffic flow source data and resolve time correlation problem adding time correlation factor. Based on the existing ITS flow- ratio curve model, using the method of least squares curve fitting, a traffic flow anomaly mining method of curve fitting of adding time correlation factor is put forword, and dividing box thought is used to set range of traffic system dynamic normal data, the abnormal data based on curve fitting is eliminated. Finally, the effects of fitting and the methods of effective identification of the abnormal traffic data are verified by examples.