基于贝叶斯网络的交通事件持续时间预测

随着数据采集手段的不断提高和相关研究技术的发展,基于数据挖掘的模型逐渐成为交通事件持续时间研究的主要方向。根据荷兰交通部门提供的交通事件采集数据,进行分类和预处理,观察事件持续时间的频数图,并根据相关的研究按照事件典型的类别把采集的数据进行分类。使用主成分分析和逐步回归提取出显著性的影响因子,利用数据挖掘软件WEKA建立贝叶斯网络模型,用数据集中80%的数据进行学习建模,20%的数据作为测试集来检测模型的预测效果,并做出性能评价。实验结果表明,与同类数据集的其他预测方法相比,贝叶斯网络模型对于变数众多,随机性特别大的交通事件,预测精度较高,证明贝叶斯网络模型的算法是具有一定优越性和实用价值。

Traffic Incident Duration Prediction Based on Bayesian Network

With the continuous improvement of data collection instruments and related research and technologicaldevelopment,establishing models based on data mining has become the main direction for studying the traffic inci-dent duration.Based on traffic incident data from the Dutch transport sector,this paper conducts classification andpre-processing,analyzes the event duration frequency chart,and classifies the collected data according to the typi-cal event category.By using principal component analysis and stepwise regression to extract significant impact fac-tor,it establishes Bayesian network model through data mining software WEKA.Then with 80% of the data in thedataset to learn modeling,20% of the data as a test set to test the predicted effects,this study makes performanceevaluation.Experimental results show that compared to other prediction methods,Bayesian network model algo-rithm has higher prediction accuracy and high randomicity for a number of large traffic events with many variables.