基于K-最近邻的交通事件持续时间预测模型
发布时间：2018-08-02 20:04
【摘要】：通过对高速公路交通事件的性质和特征进行分析,选择对持续时间影响较大的属性(事件类别、发生时间、地点、天气、伤亡程度、涉及车辆数、占用车道数)构成了描述交通事件的向量,对各属性进行了分类与量化.以交通事件的历史数据集合为基础构建N维搜索空间,计算了当前交通事件与历史交通事件之间的欧式距离,通过寻找距离最近的K个元素建立了最近邻预测模型.采用单因素方差分析法标定了变量权重,根据最小误差法确定了最佳K值.实例应用表明,K-最近邻预测模型对持续时间范围为30 min≤T90 min、90 min≤T180 min交通事件预测精度较高,适合高速公路有大量历史数据的情况下应用.
[Abstract]:By analyzing the nature and characteristics of expressway traffic events, we select the attributes (event category, time, place, weather, casualty degree, number of vehicles involved) that have a great impact on duration. The number of lanes occupied constitutes the vector describing traffic events, and classifies and quantifies each attribute. Based on the historical data set of traffic events, the N-dimensional search space is constructed, the Euclidean distance between current traffic events and historical traffic events is calculated, and the nearest neighbor prediction model is established by searching for the nearest K elements. The variable weight is calibrated by single factor variance analysis method and the optimum K value is determined according to the minimum error method. The practical application shows that the K- nearest neighbor prediction model has a high prediction accuracy for traffic events in the duration range of 30 min 鈮，
本文编号：2160574
[Abstract]:By analyzing the nature and characteristics of expressway traffic events, we select the attributes (event category, time, place, weather, casualty degree, number of vehicles involved) that have a great impact on duration. The number of lanes occupied constitutes the vector describing traffic events, and classifies and quantifies each attribute. Based on the historical data set of traffic events, the N-dimensional search space is constructed, the Euclidean distance between current traffic events and historical traffic events is calculated, and the nearest neighbor prediction model is established by searching for the nearest K elements. The variable weight is calibrated by single factor variance analysis method and the optimum K value is determined according to the minimum error method. The practical application shows that the K- nearest neighbor prediction model has a high prediction accuracy for traffic events in the duration range of 30 min 鈮，
本文编号：2160574
本文链接：https://www.wllwen.com/kejilunwen/jiaotonggongchenglunwen/2160574.html
