基于数据挖掘的高速公路行程时间预测  

Freeway travel time prediction based on clustering method with data mining

作　　者：邢雪[1,2] 于德新[2,3,4] 田秀娟[2] 程泽阳 

机构地区：[1]吉林化工学院信息与控制工程学院,吉林吉林132022 [2]吉林大学交通学院,吉林长春130022 [3]吉林大学汽车仿真与控制国家重点实验室,吉林长春130022 [4]吉林大学吉林省道路交通重点实验室,吉林长春130022

出　　处：《华中科技大学学报（自然科学版）》2016年第8期36-40,共5页Journal of Huazhong University of Science and Technology(Natural Science Edition)

基　　金：国家自然科学基金资助项目(51308249);山东省省管企业科技创新项目(20122150251-1)

摘　　要：以高速公路上路段行程时间历史数据集作为研究对象,以交通数据实际特征作为行程时间历史数据集分类的依据,建立基于数据挖掘的高速公路行程时间预测模型.采用山东高速收费站记录实际数据,以实例数据验证模型并计算平均绝对误差.为对比分析预测强度聚类预测模型的有效性,采用多种算法对测试集行程时间预测,对预测结果误差进行了对比.结果表明:预测强度修正的k-means法可以提高高速公路路段行程时间预测的准确性,模型在降低数据采集成本的同时可为高速信息服务提供可靠的预测行程时间,为出行者提供有力的决策依据.Model chooses freeway history data was set as the research object to predict freeway travel time.Historical data of traffic travel time was categorized into different sample types using cluster analysis.Traffic data of travel time was classified and identified by actual characteristics of historical data.Freeway travel time prediction model was structured using data mining.The instance data was the real data recorded from toll stations of Shandong province.Analysis used the instance data forecast model and calculated the mean absolute percentage error of algorithm.To illustrate the validity of the mode,actual test set was applied to a variety of algorithms of travel time prediction.The comparison between the prediction errors of the algorithms was given.The results show that the modified kmeans algorithm mentioned in the paper improves the accuracy of prediction.Model decreases the cost of data acquisition,and provides reliable prediction of travel time for the information service.The model provides powerful decision basis for travelers.

关 键 词：高速公路 行程时间预测 预测强度 数据挖掘 k-means法 联网收费数据 

分 类 号：U491.1[交通运输工程—交通运输规划与管理]