基于手机信令数据的数据清洗挖掘与居民职住空间分析

发布时间：2018-03-23 06:33

本文选题：手机信令数据　切入点：数据清洗　出处：《西南交通大学》2017年硕士论文　论文类型：学位论文

【摘要】：作为所辖区域的政治、经济、文化中心,城市的发展能有力地带动周边区域的发展。城市中合理的居民职住空间规划能有效提高居民的生活质量与工作效率,是城市建设中不可或缺的一环。因此如何获取城市居民的人口密度以及职住空间分布是城市规划建设中的大问题,传统获取职住数据的方式主要为上门走访、发放调查问卷等人工方式,这些方式既消耗了大量的人力物力,又不能保证结果的准确性。随着时代的发展,一些较为先进的手段已经应用在了城市信息获取中,如电话访问调查,视频监控记录等,但这些方法不可避免地存在成本高、覆盖面小等缺点。目前,利用手机信令数据获取城市居民的职住空间分布是大数据时代下的研究热点,本文通过对手机信令数据进行清洗、挖掘、提取这一系列步骤,获取居民的职住空间分布信息。本文的工作主要体现在以下几个方面:在数据清洗方面,提出了对手机信令数据进行层次化清洗的原则:首先将信令数据视为常规数据,清洗其中包含的无效数据;随后针对手机信令数据的特点,清洗信令数据所特有的漂移数据;最后基于后续停留点提取的需求,提出一种经K-means聚类算法改良的LOF离群点检测算法,清洗影响停留点提取的轨迹点。经实验验证,该改良算法能够在不影响检测精度的前提下提高算法的效率。在数据挖掘方面,针对DBSCAN聚类算法不适合处理时空数据的缺陷,从聚类算法的基本概念入手,对其进行时间维度的语义扩充,使其适合手机信令数据的停留点提取;在停留点提取算法的参数确定方面,创新性地结合停留点的定义与数据源的特征,动态地确定算法的参数,简化了参数确定过程;最后,与经典的停留点提取算法对比,证明本文提出算法的优越性。在职住空间提取方面,首先对实验获取的停留点数据进行分析,得到其基本特征;随后结合资料设定了职住停留点划分的时间阈值,并设计了相应算法从停留点集合中提取职住停留点;最后,根据获取的停留点集合绘制居民职住分布热力图,并结合实际情况对热力图进行分析。
[Abstract]:As the political, economic and cultural center of the region under its jurisdiction, the development of the city can powerfully drive the development of the surrounding area. The reasonable planning of the occupation and housing space of the residents in the city can effectively improve the quality of life and work efficiency of the residents. It is an indispensable link in urban construction. Therefore, how to obtain the population density of urban residents and the distribution of occupation and housing space is a major problem in urban planning and construction. These methods not only consume a lot of manpower and material resources, but also can not guarantee the accuracy of the results. With the development of the times, some more advanced methods have been used in obtaining urban information. Such as telephone interview investigation, video surveillance record and so on, but these methods inevitably have some disadvantages, such as high cost and low coverage. At present, the use of mobile phone signaling data to obtain the spatial distribution of urban residents' occupation and residence is a hot research topic under big data's time. In this paper, the mobile phone signaling data cleaning, mining, extraction of this series of steps, to obtain residents of the occupation and residential space distribution information. The work of this paper is mainly reflected in the following aspects: in data cleaning, The principle of hierarchical cleaning for signaling data of mobile phone is put forward: firstly, the signaling data is regarded as regular data, and the invalid data is cleaned, then the drift data of signaling data is cleaned according to the characteristics of signaling data of mobile phone. Finally, based on the requirement of subsequent stay point extraction, an improved LOF outlier detection algorithm based on K-means clustering algorithm is proposed to clean the locus points that affect the stay point extraction. The improved algorithm can improve the efficiency of the algorithm without affecting the detection accuracy. In the aspect of data mining, the DBSCAN clustering algorithm is not suitable to deal with the defects of space-time data. The semantic extension of time dimension makes it suitable for the stopover point extraction of mobile phone signaling data, and innovatively combines the definition of stopover point with the characteristics of data source in parameter determination of stopover point extraction algorithm. The parameters of the algorithm are determined dynamically, which simplifies the process of parameter determination. Finally, compared with the classical stop-point extraction algorithm, the superiority of this algorithm is proved. Firstly, the data obtained from the experiment are analyzed, and the basic characteristics of the data are obtained. Then, the threshold value of the occupation and residence points is set according to the data, and the corresponding algorithm is designed to extract the occupation and residence points from the set of residence points. According to the set of residence points obtained, the thermal diagram of resident occupation and residence distribution is drawn, and the thermodynamic diagram is analyzed in combination with the actual situation.
【学位授予单位】：西南交通大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP311.13

【相似文献】