Recommendation in location based social networks : by mining user check-in behaviour based on social factor, individual preference and POI popularity

Abstract

Recommendation of urban Points-Of-Interest (POI), such as restaurants, based on social information has attracted a lot of attention in recent years. Most of the recommendation methods nowadays focus based on only the individual or friends’ check-in behaviours. Thus the recommended POI results are often constrained by users’ or friends’ living area. Moreover with the ever-changing information in urban areas, extracting appropriate features from heterogeneous data is a critical and challenging task. In this report, the author elaborates an Urban POI-Mine (UPOI-Mine) approach [8] that incorporates location-based social networks for recommending users urban POIs based on the user preferences and location properties concurrently. In order to support the prediction of POI related to individual user’s preference, the main idea of UPOI-Mine is to build a regression-tree based predictor in the normalized check-in space. Based on the LSBN data from Foursquare, the author extracts the features of places based on (1) Social Factor, (2) Individual Preference, and (3) POI Popularity for building the model. In this report, the author also describes the detailed UPOI-Mine algorithm and the implementation of the algorithm’s two feature extraction phases. Finally, the implementation results and discussion on the results is elaborated.