论文部分内容阅读
GPS定位系统对车辆的运行调控以及拥堵性分析具有重要意义。但定时采样的GPS数据难免存在坏点的情况,而坏点的存在对分析结果容易产生较大错误,从而影响交通管理决策。本文通过高斯混合模型、K-均值聚类分析以及SOM自组织神经网络三种方法完成对原始数据时间段划分、字段提取以及坏值清理的操作。这三种方法主要用于对数据进行聚类分析,根据分析结果识别孤立点从而进行清理。结果显示,高斯聚类与K-均值聚类算法的坏点识别精度小于SOM自组织神经网络,但前两种算法的运行效率较后者高。
GPS positioning system for vehicle operation control and congestion analysis is of great significance. However, it is inevitable that there is a dead pixel in regularly sampled GPS data. However, the existence of a dead pixel is likely to cause a large error in the analysis result, thereby affecting traffic management decision-making. In this paper, we use the Gaussian mixture model, K-means clustering analysis and SOM self-organizing neural network to complete the original data time segment, field extraction and bad value cleanup operation. These three methods are mainly used for cluster analysis of data, based on the results of the analysis to identify isolated points in order to clean up. The results show that the accuracy of Gaussian clustering and K-means clustering algorithm is less than the SOM self-organizing neural network, but the former two algorithms are more efficient than the latter.