Using Linguistic Incomplete Preference Relations to Cold start Recommendations

Abstract

Purpose – Analyzing current recommender systems, it is observed that the cold start problem is still too far away to be satisfactorily solved. This paper aims to present a hybrid recommender system which uses a knowledge-based recommendation model to provide good cold start recommendations. Design/methodology/approach – Hybridizing a collaborative system and a knowledge-based system, which uses incomplete preference relations means that the cold start problem is solved. The management of customers’ preferences, necessities and perceptions implies uncertainty. To manage such an uncertainty, this information has been modeled by means of the fuzzy linguistic approach. Findings – The use of linguistic information provides flexibility, usability and facilitates the management of uncertainty in the computation of recommendations, and the use of incomplete preference relations in knowledge-based recommender systems improves the performance in those situations when collaborative models do not work properly. Research limitations/implications – Collaborative recommender systems have been successfully applied in many situations, but when the information is scarce such systems do not provide good recommendations. Practical implications – A linguistic hybrid recommendation model to solve the cold start problem and provide good recommendations in any situation is presented and then applied to a recommender system for restaurants. Originality/value – Current recommender systems have limitations in providing successful recommendations mainly related to information scarcity, such as the cold start. The use of incomplete preference relations can improve these limitations, providing successful results in such situations.