GeoSocial : um modelo de análise e agrupamento de população de pessoas baseado em hábitos de frequência e semântica de locais

Abstract

Information about user navigation behavior on the web has been widely used to draw user behavioral profiles in order to offer advertisements segmented by categories. In this same line, behavior habits based on places that an individual attends in their daily life can also be analyzed. This paper proposes a clustering model of individuals for further analysis of their habits of frequency in places (GeoSocial). Patterns of the formed groups represent characteristics of population’s behavior and can help to identify market opportunities or to help decision makers linked to government to propose improvements/changes in the infrastructure of a city. Users information about their frequented interest places are captured by GPS coordinates by a mobile app developed. App tracks and storages places that are frequent individuals. It allows visualize their time permanency on places and connect they to a social network formed from the similarities between their habits and the others. The proposed model includes: i. a user clustering module based on Affinity Propagation technique; ii. an interactive visualization module to analyze individual data correlation of groups based on Parallel Coordinates technique. GeoSocial is evaluated by different scenarios, making use of artificial data generated. Evaluation indicates the possibility of the model to a multitude of objectives.