Handling long tail in recommendation systems for niche markets using fuzzy classification

Abstract

Recommendation systems provide the best-desired items to a user based on the user’s interest measured from user activities. Online movie search is a very popular activity of the internet users. There are many systems, which are suggestive for searching the movie based on the fact that users previously have viewed any other similar types of movies. Items or movies are selected based on their features. However, all of the attributes of the items are not equally important. All popular items are not liked by all users and niche items are available to the classified users. Therefore, the degree choice by the users based on the attributes of items of the niche markets should be considered. In this work, the fuzzy rating has been used for a niche market’s items. In this research, we use Mamdani Rule-Based Fuzzy Inference Technique for movie recommendation. This methodology helps the movie viewers to watch a movie after knowing that how much utility the user will get from this movie. We find that, proposed fuzzy rating reduces the long tail of the movie list. This helps users to consume a specific item. The movies will be recommended to the specific users because segmentation and the fuzzy membership of the attribute helps to achieve the popularity in niche market. We design a data-set, where the single movie get multiple ratings for different attributes. We find our proposed model can handle long tail problem of the items of niche markets.