Explainable Collaborative Filtering Recommendations Enriched with Contextual Information

Today, the most important requirement of intelligent systems is to be able to explain their decisions to the end-user. Fulfilling this requirement is the goal of explainable AI (XAI) that proposes to produce explainable models that enable end-users to understand and trust the models. This research addresses the explainability of the recommendations. This paper presents an explainable recommender system for point of interest recommendations taking into account the context of the user. In our experiments we have used the STS (South Tyrol Suggests) dataset. The following major steps are part of our methodology: i) presenting the dataset, ii) using Restricted Boltzmann Machine based collaborative filtering recommendations, iii) using contextual information, and iv) extracting and presenting explanations for recommendations based on contextual information. The novelty that we propose in explainable recommender systems is a new explainable recommendation technique, which is quantitative and qualitative, providing both the list of top-n recommendations and the explanations of the recommendations based on context. This paper also provides an overview of research on this topic.