Modeling Weather-Aware Prediction of User Activities and Future Visits

Abstract

In recent years, Location-Based Social Networking (LBSN) sites such as Foursquare, Facebook Places, and Twitter have become extremely popular due to the extensive usage of location-enabled smart phone technologies. These LBSNs allow users to post their check-ins that provide important set of information about users' activities and preferences. Several existing research works predict users' activities from social media check-ins data considering various aspects such as time, venue, and occurrences. However, none of the earlier studies investigate the influence of weather on users' preferences of activities and mode of transportation preferences. Psychological studies show that weather has a strong influence on human activities. In this paper, we predict users' travel mode, day/night time activities and future visit from weather condition derived from social media check-ins. In particular, we develop several classification models to predict users' preferable mode of transportation, day/night activities, and future visits from the users' check-ins based on different weather conditions. We use two real datasets: Tokyo and New York city to validate our models. Our classifiers achieve substantial strength (on an average AUC of 72.77%) to predict users' mode of transportation, day/night activities and their future visit preferences for Tokyo dataset. We also compare performance of the classifiers developed by these two datasets.

Keywords

• Foursquare
• check-ins
• weather
• correlation
• classification