Robustness Analysis of Multi-Criteria Top-n Collaborative Recommender System

Recommendation systems have popular methods that help users generate predictions for a product and create product lists based on users' feedback. The output of such systems can increase customer satisfaction by producing accurate recommendations. However, the success of predictions depends on the level of personalization of user preference data. It is possible to increase the personalization level by collecting more detailed customer data. Multi-criteria recommender systems using the user-item matrix which users evaluate in terms of more than one criterion, are introduced to achieve such a goal. Besides the accuracy of predictions, vulnerability against shilling attacks is an important challenge for recommender systems. However, there is no study in the literature that deals with the robustness of multi-criteria top-n recommendation methods. Hence, in this study, robustness analysis of three different item-based multi-criteria top-n recommendation methods is performed. We use new and existing evaluation criteria for the effectiveness of attacks, an attack model adapted to multi-criteria systems, and an experimental design that included target and power item selection. According to the experimental outcomes based on a real-world dataset, it can be said that multi-criteria top-n algorithms are vulnerable to manipulations.