Extracting Attributes for Recommender Systems Based on MEC Theory

Abstract

To retain consumer attention and increase their purchasing rates, many online e-commerce vendors have adopted content-based approaches in their recommender systems. However, except for text based documents, there is little theoretic background information guiding the selection of elements. On the other hand, Means-End Chain theory noted deciding elements that dictate product selection include attributes, benefits, and values. This study strives to establish a methodology to identify favorite attributes based on Means-End Chain theory. The experiment is conducted to compare and contrast the performance of the proposed method and two traditional content (attribute) based methodologies. The results show that the proposed system outperforms the two methods by 82% and 68%, respectively.