A contextualized and personalized model to predict user interest using location-based social networks. (July 2016)

• Record Type: Journal Article
• Authors: Li, Ming; Sagl, Günther; Mburu, Lucy; Fan, Hongchao
• Is Part Of: Computers, environment and urban systems. Volume 58(2016)
• Page Start: 97
• Page End: 106
• Publication Date: 2016-07
• Subjects: User interest -- Context-awareness -- Personalization -- Prediction -- Location-based social networks; City planning -- Data processing -- Periodicals; Regional planning -- Data processing -- Periodicals; 303.4834
• DOI: 10.1016/j.compenvurbsys.2016.03.006
• Languages: English
• ISSNs: 0198-9715

Abstract

Abstract: The accurate determination of user interest in terms of geographic information is essential to numerous mobile applications, such as recommender systems and mobile advertising. User interest is greatly influenced by the usage context and varies across individuals; therefore, a user interest model should incorporate these individual needs and propensities. In this paper, we present an approach to model user interest in a contextualized and personalized manner based on location-based social networks. Multinomial logistic regression is employed to quantify the relationship between user interest and usage context at both the aggregate and individual levels. The proposed approach is tested in a real-world application using Foursquare check-ins issued between February and June 2014 in the three major cities of Chicago, Los Angeles and New York. Results demonstrate the capability of the contextualization process for capturing contextual influences on user interest, and that such influences can be observed at a fine-grained scale at the individual level through the personalization process. The proposed approach therefore enables contextualized and personalized estimation of user interest, thereby contributing useful information to follow-up mobile applications. Highlights: We propose an approach to model and predict user interest for follow-up mobile applications. We introduce a new idea to utilize contextual information in mobile environment. We prove that contextualization can capture contextual influences on user interest at the aggregate level. We prove that personalization can capture contextual influences on user interest at the individual level. We find that personalization does not always improve prediction when the personal training data is less than 12.