Search

Byline: Do Hyun Ahn, Jae Kyeong Kim, Il Young Choi, Yoon Ho Cho With the advance of e-commerce, the importance of recommender systems grows larger than before. Collaborative filtering is the most successful recommendation method, but its widespread use has exposed some problems such as sparsity and scalability. To overcome such limitations, this paper proposes hybrid recommendation methodologies based on web usage mining and dimensionality reduction techniques to enhance the recommendation quality and performance of current collaborative filtering-based recommender systems. Web usage mining populates the rating database by tracking customers' shopping behaviours on the web, thereby leading to better quality recommendations. The product taxonomy and SVD (Singular Value Decomposition) are used to improve the performance of searching for nearest neighbours through dimensionality reduction of the rating database. Experiments on real e-commerce data show that the proposed methodologies provide higher quality recommendations and better performance than existing collaborative filtering methodologies.