Summary: | 碩士 === 國立中山大學 === 電機工程學系研究所 === 106 === The fast development of E-commerce causes the urgent need of various recommender systems that help consumers to find interesting products by extracting knowledge from the previous interaction information of users. Collaborative filtering recommender systems traditionally recommend products to users solely based on the user-item rating matrix and are simple, convenient to use. However, some issues have long been concerned, and researchers have been trying hard with different solutions to make collaborative filtering more practical and useful. In this paper, we focus on two main issues, data sparsity and scalability. Data sparsity is related to the sparse ratings in the useritem rating matrix and it can lead to inaccurate recommendations, while scalability is related to the huge number of products and users involved in E-commerce, which may cause an unacceptably long delay before valuable recommendations are acquired. We propose a novel approach to deal with these two issues. Word2Vec is employed to build item vectors, one item vector for each product, from the comments made by users on their previously bought goods. Through the user-item rating matrix, user vectors of all the users are then obtained. Dimensionality reduction and clustering techniques are applied to reduce the time complexity related to the large numbers of items and users. Recommendation work is then done with the resulting clusters. Finally, reverse transformation is performed and a ranked list of recommended items is offered to each user. With the proposed approach, the inaccuracy caused by the sparse ratings in the useritem rating matrix is overcome and the processing time for making recommendations from an enormous amount of data is much reduced. Experimental results of real data sets are shown to demonstrate the effectiveness of our proposed approach.
|