Analyzing User Behaviors: A Study of Tips in Foursquare

Foursquare is a popular Location Based Social Network (LBSN). It has become a major platform that enables users to share their opinions on locations they have visited through check-ins and writing tips. The massive amount of data generated by Foursquare provides unexpected opportunities to analyze and obtain interesting insights into people and places. Most of the previous research addressed the interesting findings regarding user behavior through check-ins, but not the characteristics of the most visited venues, which we address in our paper. We also analyze sentiment of Arabic text in LBSNs, focusing on Saudi Arabia. We collected data of more than 1000 venues, 50,000 check-ins and 12,000 tips to investigate the different aspects of those venues with low rating and positive comments by our proposed algorithm using sentiment analysis on Arabic tips. More interestingly, we discovered different communities in Saudi Arabia by applying the Latent Dirichlet Allocation (LDA) model as one of the of topic model approaches. We concluded that some venues with low ratings have more visitors due to the range of services available in the region. In addition, the high number of positive tips proves that certain people influence the others’ opinions regardless of the restaurant’s rating. The LDA model produces latent collections of people with similar interests as communities which indicates their behavior and patterns.