Contribution to Collaborative Filtering Based on Soft Computing to Enhance Recommender System for e-Commerce

Abstract

Recommender Systems (RSs) are used by an ever-increasing number of e-commerce sites to recommend items of interest to the users based on their preferences. Collaborative filtering is one of the most regularly used techniques in RSs that help the users to catch the items of interest from a massive numbers of available items. This technique is based on the idea that a set of like-mind users can help each other to find valuable information. The major challenge in recommender systems is that the user ratings or grades are very often uncertain or vague because it is based on user's tastes, opinions, and perceptions. Fuzzy sets appear to be a proper paradigm to handle the uncertainty and fuzziness of human decision making activities and to successfully model the normal sophistication of human behavior. Because of these motives, this paper adopts type-2 fuzzy linguistic approach to efficiently describe the user ratings and weights to precisely rank the relevant items to a user. The proposed method permits users to express their ratings in qualitative form, converts such preferences to their corresponding quantitative form using the concept of type-2 fuzzy logic, maps the values that represent the preferences with the retrieved items from the database, and finally recommends products that best satisfy the consumer's likings. Empirical evaluations show that the proposed technique is feasible and effective.