User Identity Matching for Multisource Location Data

The widespread use of LBS services generates a lot of location data. To fuse location data from different sources can provide more complete user information for data analysts. Identity matching is the precondition of location data fusion. This paper made a comprehensive use of social relationship structure and behavioral pattern to realize the identity matching of users. Firstly, the information, such as user social relationships, personal important place and spatio-temporal co-occurrence area, and so on, were obtained from the location data. Then the active users were extracted to generate the matching user sets. Secondly, the matching algorithm was used to calculate the similarity of user social relationship structure and behavior pattern of users in the matching sets, and the matching relations was obtained according to the maximum similarity principle and the double selection criterion. Finally, based on the social relationship structure, the neighbor nodes of matched users were extended and used to refill the matching user sets. The process was iterated to complete identity matching. The method in this paper had the advantages of non-essential of prior information, good robustness and high reliability of matching results. The experiments on three different type of real datasets showed that: (1) with the selection of reasonable parameters, the active user extraction method could get the confidence user to meet the requirement of algorithm without prior information; (2) the proposed method was less affected by the type of dataset and the purity of the user set, which had good robustness; (3) the matching results of the proposed method had better accuracy and recall rate, and the reliability of the user identity matching results was higher.

I. Introduction

In recent years, with the development of mobile communication, Internet and GNSS (Global Navigation Satellite System) positioning technology, more and more mobile terminal equipment positioning sensors promotes the development of LBS (Location Based Services) [1], [2]. Nowadays, LBS has been widely used in many important areas of application: social network, intelligent navigation, counterterrorism intelligence support, traffic control, vehicle monitoring, intelligent logistics, digital city planning, digital battlefield and simulation, mobile e-commerce and so on, and more and more location data is generated. The location data implicit human behavior patterns and habits [3], which can be used for user sensitive information inference [2], group discovery [4], [5], location-based business recommendation [6], LBA (Location Based Advertising) [7] and other fields. At present, most of the research objects are limited to single source data. If the location data from multiple sources can be fused, it can provide more abundant user information for various research. The key of location data fusion is to find the identity matching relations of the users.