Predicting and Understanding Human Mobility Based on Social Media Check-in Data.

Exploring the regular pattern of people's movement activities, predicting people “where to go” and “what to do” is a critical task in many applications. In this paper, we attempt to explore human mobility through social media check-in data. This study mainly includes two aspects: one is based on the considering of a multi-scale spatial characteristics of the spatiotemporal prediction method to predict the crowd volume in a certain region; one is to improve the classic TFIDF method, and propose a method to calculate the significance of POI type in the region in each period, so as to analyze and explain people's mobility tendency. We use the check-in data set of Manhattan in New York to carry out the experiment and the results show that the methods used in this paper can better predict the volume of people's activities, and realize the exploratory analysis of people’s mobility tendency.