A Novel Matrix for Location Categorization In Human Mobility Mining

Human mobility mining is feasible due to easy availability of the users' GPS traces. User's trajectory derived from GPS based track points is the raw data for mining various trends in user mobility. Identification of staypoints and region of interest (RoI) is one of the most important aspects while processing trajectory data for the various location-based application. Staypoints and RoIs are the locations where the user spends some measurable time and performs some activity. These are the important locations of the user's daily routine and probably interesting locations from the user's perspective. Each location possesses substantial importance in user's routine. One of the basic methods to find such locations is to analyze user's mobility traces to find staypoints from it and then to perform clustering on these staypoints. Existing algorithms gives equal importance to all such identified locations while designing mobility-based applications. However, realistically, all locations cannot have the same importance from the user's perspective. Also, there are chances to find few locations like home or work almost in each mobility trace even after keeping less time granularity. So, we have proposed a novel location categorization matrix LoCMat to classify location as per their importance in the user's routine based on frequency and duration of visits to each location. Identifying, retaining and assigning appropriate weightage to all such locations will enhance results of mobility mining. We have tested our proposed matrix with real dataset "Mobi-India". Our method identifies meaningful locations, assigns weightage to each one of them which can be used further for mobility mining.