基于C-K-N-Cluster的居民出行时空特征分析  

Analysis of residents′ travel time-space characteristics based on C-K-N-Cluster

作　　者：戴兵 田博 高心雨 严李强 DAI Bing;TIAN Bo;GAO Xinyu;YAN Liqiang(School of Information Science and Technology,Tibet University,Lhasa 850000,China)

机构地区：[1]西藏大学信息科学技术学院,拉萨850000

出　　处：《智能计算机与应用》2022年第11期64-70,共7页Intelligent Computer and Applications

基　　金：2021年中央引导地方科技发展资金项目(XZ202101YD0014C);西藏大学研究生“高水平人才培养计划”项目(2020-GSP-S169)。

摘　　要：为解决传统聚类算法在大数据轨迹信息应用中的簇类数不确定、病态初始化等问题,文章提出了一种结合Canopy与K-Means++的小生境遗传智能聚类算法(C-K-N-Cluster),并应用于居民出行时空特征分析;以杭州市为例,对出租车轨迹数据进行降噪标准化等预处理,按照筛选原则提取载客点数据;提取出的数据投入智能聚类算法仿真识别城市上下载客热点地域,结合数据分析方法可视化研究城市居民出行特征。仿真结果表明:改进算法相比传统K-Means能够实现大数据应用场景下的簇类数与初始化自动最优化,分析了杭州市居民出行规律及出租车载客时空特征,为司乘服务和城市功能区优化提供参考。This paper proposes a niche genetic intelligent clustering algorithm(C-K-N-Cluster) combining Canopy and K-Means++ and applies it to the analysis of residents′ travel space-time characteristics in order to solve the problems of uncertain cluster number and ill initialization of traditional clustering algorithms in big data trajectory information application. In the research, Hangzhou is taken as an example, where taxi track data is preprocessed by noise reduction standardization and passenger point data is collected using the screening method. The collected data is fed into an intelligent clustering algorithm, which simulates and identifies the hot regions of urban inbound and outgoing passengers, and data analysis methods are used to show the travel characteristics of urban people. The simulation results demonstrate that, when compared to the classic K-Means approach, the upgraded algorithm can achieve automated optimization of cluster number and initialization in a big data application situation. The research examines Hangzhou inhabitants′ travel habits as well as the time-space features of taxi clients, serving as a guide for optimizing driver services and urban functional regions.

关 键 词：轨迹数据 C-K-N-Cluster算法 可视化分析 居民出行特征 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]