Business challenge:
In order to make good decisions, it is necessary to possess ample amount of information. However, there are several examples showing that too much information is as bad as inadequate information; it is called information overload problem. Recommender System has been introduced to solve this problem. It is very popular and useful concept in current digital era. It is an information filtering system that suggests products and services most relevant to the User. Recommender System has been used widely for the products and services, intended for entertainment like music, books, and movies, online games, restaurants and completely based on user ratings. Some other applications are Personalized B2B E-Services, Critique-Based Mobile of insurance companies, and increasing competition among companies resulted in huge range of insurance products. Insurance policies have complex terminology and numerous features. Also, terms and conditions of insurance policies are not effortlessly comprehensible to customers. Therefore, a recommendation system has been required for a long time that hides complexities and recommends best polices to its users.
UFT comes into show:
We proposed an original perception-based utility recommender system for supporting insurance policy related decision making. Unlike existing recommender system, which mainly focuses on relations between products and between customers, proposed system’s recommendations are based on user’s contextual requirement. The proposed system filters insurance policies that match user’s demographic information, determines utility of them according to user preferences and recommend policies with maximum utility to the users. Insurance is very important need in today’s uncertain life but understanding features, terms and conditions of each product is very tedious and time taking process and always influenced by insurance agent’s biases.
Impact Delivered:
Our recommendation system has been tested with approximately 600 potential insurance buyers with different age and income group. The hit-rate of the proposed system was about 92.6%. The recommendation technique is also validated by domain experts. They found the results significantly accurate.
We compared our proposed system with other existing recommender systems and decision support systems for insurance, presented in literature. Other systems mainly use fuzzy logic and data mining tools for recommending policies or policy segments, but do not extract user’s preference.