GRAPH-BASED HYBRID RECOMMENDER SYSTEMS

Recommender systems provide recommendations about various products and services to their users by using other users? data. These systems depend on personal user preferences on items via ratings and recommend items based on choices of similar users. Their success is imperative for both users and the e-commerce vendors utilizing such systems. Since inaccurate and unreliable product recommendations make users search alternative sites for shopping. Hence, recommender systems are a challenging research field with many unresolved problems and many different hybrid recommendation algorithms have been proposed to overcome these problems. Hybrid models that use different information sources (text, images, ratings, etc.) for recommendation are getting more attention in recent years. In this dissertation, a graph-based hybrid recommender system is proposed that is incorporating numerical ratings and product images to learn items and the corresponding user?s representations. Moreover, another graph-based recommender system, that utilizes only user-item ratings, is proposed. In the current literature, recommendation generation in a graph based model is a link prediction problem and link prediction approaches are used to distinguish between fundamental relational dualities of like or dislike and similar or dissimilar. However, similar and dissimilar relationships between users (or items) are mostly disregarded. Hence, a link prediction method is proposed that utilizes user-user and item-item similar/dissimilar relationships with like/dislike dualities in order to improve the accuracy of the system. Similarly, triangle closing model is expanded with similarity relationships, and then the number systems are investigated to represent each similarity entity as a number. The usage of link prediction algorithms is examined for the quaternion and the complex number systems. On the standard Amazon and MovieLens datasets, the proposed similarity-inclusive link prediction method performed empirically well compared to other methods operating in the quaternion and complex domain. The experimental results show that the proposed recommender system can be a plausible alternative to overcome the deficiencies in recommender systems.