An improved neighborhood-restricted association rule-based recommender system
Rage Uday Kiran, Masaru Kitsuregawa · Australasian Database Conference · 2013
Association rule mining is an actively studied topic in recommender systems. A major limitation of an association rule-based recommender system is the problem of reduced coverage. It is generally caused due to the usage of a single global minimum support (minsup) threshold in the mining process, which leads to the effect that no association rules involving rare items can be found. To confront the problem, researchers have introduced Neighborhood-Restricted rule-based Recommender System (NRRS) using the concept of multiple minsups. We have observed that NRRS is computationally expensive to use and can recommend uninteresting products to the users. With this motivation, this paper proposes an improved NRRS using the relative support measure. We call the proposed system as NRRS++. Experimental results show that NRRS++ can provide better recommendations and is runtime efficient than NRRS.