협업 필터링 기법을 이용한 추천 시스템

Abstract

Since advertising market has recently grown rapidly, providing items quickly and accurately to customers is a factor of success. A recommendation system helps to provide content to meet the needs of these individuals. It often takes a long time for each individual to make a decision when they want to choose a product because of the wide range of content provided. To solve this problems, the content that users want should be predicted. A system that allows users to select content within a short time is important. This study shows how a selection system can increase the number of referrals using collaborative filtering techniques in order to provide content to meet users’ needs. In the first step, the selection is divided into main categories and then similarities are sought between targeted customers and users. The most relevant main category was recommended to the customer by first obtaining the predicted affinity for the main categories and by configuring the nearest neighbors. In the second step, the predicted affinity was obtained for the sub-categories and the highest sub-category was recommended. This study verified that a collaborative filtering method based on sub-categories leads to a high number of content selections by the target customers in comparison with collaborative filtering methods based only on the main category.