Mobility Prediction based on Markov Model with User Similarity using Location-Based Social Networks data
PSN, mobility prediction
The increasing availability of location-acquisition technology, e.g., embed GPS in smartphones, has created a new specificity of social networks, known as Location-Based Social Networks (LBSNs). It enables users to add a location dimension to existing online social networks in a variety of ways. In this context, LBSNs users stopped being only consumers to become data producers, offering various research opportunities such as mobility prediction and recommendation systems. In addition, LBSN data contains spatial, temporal, and social features of user activity, providing valuable information that is currently available on a large scale and low-cost form via traditional data collection methods. Several models have been proposed for mobility prediction based on LBSN, where most of them use historical records to identify user and group movements. In this sense, Markov Chain (MC) is one of the statistical models used in user mobility prediction, which aims to find the probability of an event happening given $n$ past events conforming to the order of the model. In this master thesis, we introduce the TEmporal Markov Model with User Similarity (TEMMUS) mobility prediction model. It considers an MC of variable order based on the day of the week (weekday or weekend) and the user similarity to predict the user's future location. The results highlight a higher performance of TEMMUS compared to other predictors.