Contextual Recommender System on a Location-Based Social Network for shopping places recommendation using association rules mining
Résumé
This article proposes a contextual recommender system of shopping places for users in mobility conditions and using location-based social network information. It is based on mining association rules for generating recommendations. The context is considered at multiple levels: first in selecting the interesting rules following to the current session of visits of the user; then into a relevance measure containing a geographic measure that considers her current geographic position. Besides the geographic measure, the relevance measure combines a social measure and an interest measure. The social measure takes into account the habits of the user's friends and the interest measure depends on the current intention of the user among predefined intention scenarios. The method we proposed for recommendation allows controlling the type of recommendation the user wishes by acting directly on this relevance measure. This proposition is tested on a real dataset and we show that the use of social and geographic features beyond usage information can improve contextual recommendations.