A Genetic Algorithm Approach to Context-Aware Recommendations Based on Spatio-temporal Aspects

Abstract

Context-aware recommender systems (CARS) have been extensively studied and effectively implemented over the past few years. Collaborative filtering (CF) has been established as a successful recommendation technique to provide web personalized services and products in an efficient way. In this chapter, we propose a spatio-temporal-based CF method for CARS to incorporate spatio-temporal relevance in the recommendation process. To deal with the new-user cold start problem, we exploit demographic features from the user’s rating profile and incorporate this into the recommendation process. Our spatio-temporal-based CF approach provides a combined model to utilize both a spatial and temporal context in ratings simultaneously, thereby providing effective and accurate predictions. Considering a user’s temporal preferences in visiting various venues to achieve better personalization, a genetic algorithm (GA) is used to learn temporal weights for each individual. Experimental results demonstrate that our proposed schemes using two benchmark real-world datasets outperform other traditional schemes.