An extended proximity relation and quantified aggregation for designing robust fuzzy query engine

Abstract

In this article, we propose a novel model of a robust fuzzy query engine that addresses vagueness in data and users’ requirements. It aims to assist users in recommending similar products or services by retrieving the most suitable entities when the limitations of queries and recommendation approaches are recognized. The proposed fuzzy engine model considers various complex aspects, including imprecise preferences explained by linguistic terms, uncertain data in datasets, connections among elementary requirements, and the lack of historical data necessary for personalized recommendations. To achieve this goal, we propose a similarity matching based on the extended proximity relation and an adapted conformance measure for elementary requirements. In this direction, a new monotonicity property for proximity relation is introduced to ensure consistency in similarities among ordinal categorical data (including binary data) expressed by crisp or fuzzy numbers. Therefore, the conformance measure used to evaluate the similarity between user requirements and attribute values is expressed as a fuzzy number. Next, we propose a quantified aggregation of elementary requirements by strictly monotone fuzzy relative quantifiers. The flexibility is further extended by a convex combination of possibility and necessity measures. A hotel selection experiment is being carried out to explore the potential of the proposed fuzzy query engine. Finally, the limitations and usefulness of the proposed approach are addressed.