Abstract:

This study presents a fuzzy inference system for insurance advisors to identify and suggest appropriate policies to potential or existing clients which can minimize the subjective prejudice of the insurance advisors. The study applies an expert system framework and adapted Ketataet al. proposed method to integrate rules induced by expert and dataset. The dataset is a real life dataset from a major Islamic insurance operator which consists of five types of insurance policies under the Family Plan. Five influential variables are identified which are age, gender, marital status, salary and job class. These inputs are transformed into fuzzy variables using triangular membership functions and then used to construct the fuzzy rule base. Apart from machine learning, an expert with more than 10 years of experience is also engaged to determine common rules used in the practice of identifying suitable policies to prospects. Rules from both sources are compared based on similarity measures. Those with similarity measures above threshold are merged. The accuracy of the rule base is also measured.