Product Completion with Fuzzy Association Analysis: Recommendation System Application for an Insurance Company

Recommendation systems have been widely accepted and have attracted significant attention both from practitioners and academicians. Utilising past behavioural patterns or indicators of customers, recommendation systems determine items or contents that would be related to a user. From this perspective, association rule mining provides extracting rules for related items according to the sequence of past behaviour of the customers. Thus, recommendation systems can be enhanced by using these ‘valuable’ rules to find out what kind of items customers might prefer and what could be shown that is not provided before. This study includes the following: (i) The way of association rule mining can be used to improve state-of-the-art association rule mining algorithms (ii) an ensemble of the extracted rules can be used as an alternative way for current state-of-the-art recommendation approaches providing fuzzy set theory and (iii) a case study to prove the way for implementation of fuzzy ensemble association rule mining. The case study is conducted using an insurance dataset, and results are reported for building the infrastructure of ensemble-based recommendation systems.