Understanding Job-Housing Relationship from Cell Phone Data Based on Hadoop

Abstract

Job-housing relationship has been a significant and classic topic in urban studies for many years. Compared with the studies based on data from census and public transportation surveys, cell phone data make it possible to understand job-housing relationship in finer time and spatial scales. Therefore, cell phone data contribute to the study of job-housing relationship at least in three aspects. First, cell phone data can make the indicators of job-housing relationship closer to their real meanings. Second, it makes analysis units more flexible according to analysis needs. Third, algorithms can be used to describe people’s movement traces more effectively and to classify various groups like commuters or travelers. In this chapter, a Hadoop-based cell phone data processing platform for job-housing relationship analysis is introduced as a solution to the storage and processing of large volumes of real-time cell phone data. The platform is innovative in its big data computation framework, cell phone data processing procedures and algorithms, as well as job-housing relationship and commuting analysis application system design. The big data computation framework is based on Hadoop for the collection and the processing of mass structured, semi-structured, and nonstructured data. The cell phone data processing platform is to realize the collection, processing, storage, export, and other processing of cell phone data and to save all the intermediate and final computational results for reuse. The job-housing relationship and commuting analysis application system provides functions such as the display, query, statistics, analysis, and comparison of indicators under multiple spatiotemporal scales. Chongqing is used as an example to demonstrate how the platform works by three main steps: original data cleansing and quality evaluation, trace series analysis, and user stop recognition and analysis. Job-housing relationship from cell phone data is observed and described by job-housing distributions, OD connections among analysis units, as well as three indicators, including external employment rate, foreign employment rate, and independence index.