Value extraction and collaborative mining methods for location big data

Uncountable geographical location information, vehicle trajectories and users' application location records have been recorded from different location-based service (LBS) applications. These records are forming to a location big data resource which facilitates mining human migrating patterns, analyzing geographic conditions and building smart cities. Comparing with traditional data mining, location big data has its own characteristics, including the variety of resources, the complexity of data and the sparsity in its data space. To restore and recreate data analysis network model from location big data, this study applies data value extraction and cooperative mining on location big data to create trajectories behavior pattern and local geographical feature. In this paper, three major aspects of analysis methods on location big data are systematically explained follows: (1) For the variety of resources, how to extract potential contents, generate behavior patterns and discover transferring features of moving objects in a partial region; (2) For complexity of data, how to conduct dimension reduction analysis on complex location networks in temporal and spatial scale, and thus to construct learning and inferential methods for mobility behavior of individuals in communities; (3) For sparsity, how to construct the global model of location big data by using collaborative filtering and probabilistic graphical model. Finally, an integral framework is provided to analyze location big data using software engineering approach. Under this framework, location data is used not only for analyzing traffic problems, but also for promoting cognition on a much wider-range of human social economic activities and mastering a better knowledge of nature. This study incarnates the practical value of location big data. © Copyright 2014, Institute of Software, the Chinese Academy of Sciences. All rights reserved.