ANOMALOUS ACTIVITY DETECTION FROM DAILY SOCIAL MEDIA USER MOBILITY DATA

Anomalous activities are the activities that do not fit into normal and routine behavior of people or objects. Anomalous activity, account, or sharing detection from social networks play an important role for preventing social media users from harmful and annoying contents. However, detecting anomalous activities is challenging due to the difficulty of separating anomalous activities from real ones, limitations of current algorithms and interest measures, the challenge of analyzing social media big data, and hardness of handling spatial and temporal dimensions. In this study, anomalous activities are detected using daily social media user mobility data. In particular, two features are extracted from daily social media user mobility, namely, daily total number of visited locations and daily total distance, and these features are used for detecting anomalous activities. An algorithm, that employs DBSCAN clustering algorithm, is proposed for detecting such activities. The results show that proposed algorithm could learn normal daily activities of social media users and detect anomalous activities