Analysis of Recommendation Algorithms for E-Commerce-英文文献.pdf

Analysis of Recommendation Algorithms for E-Commerce Badrul Sarwar, George Karypis, Joseph Konstan, and John Riedl fsarwar, karypis, konstan, riedl g@ GroupLens Research Group / Army HPC Research Center Department of Computer Science and Engineering University of Minnesota Minneapolis, MN 55455 ABSTRACT Pro ducts that the neighb ors like are then recommended to Recommender systems apply statistical and knowledge dis- Neo, as he will probably also like them. Collab orative l- covery techniques to the problem of making pro duct recom- tering has b een very successful in b oth research and prac- mendations during a live customer interaction and they are tice. However, there remain imp ortant research questions achieving widespread success in E-Commerce nowadays. In in overcoming two fundamental challenges for collab orative this pap er, we investigate several techniques for analyzing ltering recommender systems. large-scale purchase and preference data for the purp ose of The rst challenge is to improve the scalability of the col- pro ducing useful recommendations to customers. In par- lab orative ltering algorithms. These algorithms are able ticular, we apply a collection of algorithms such as tradi- to search tens of thousands of p otential neighb ors in real- tional data mining, nearest-neighb or collab orative ltering, time, but the demands of mo dern E-commerce systems are and dimensionality reduction on two dierent data sets. The to