PROJECT TITLE :

l-Injection: Toward Effective Collaborative Filtering Using Uninteresting Items - 2017

ABSTRACT:

We tend to develop a completely unique framework, named as l-injection, to deal with the sparsity problem of recommender systems. By fastidiously injecting low values to a particular set of unrated user-item pairs in an exceedingly user-item matrix, we have a tendency to demonstrate that prime-N recommendation accuracies of various collaborative filtering (CF) techniques will be significantly and consistently improved. We first adopt the notion of pre-use preferences of users toward a vast quantity of unrated items. Using this notion, we establish uninteresting things that have not been rated however but are probably to receive low ratings from users, and selectively impute them as low values. As our proposed approach is methodology-agnostic, it can be simply applied to a selection of CF algorithms. Through comprehensive experiments with 3 real-life datasets (e.g., Movielens, Ciao, and Watcha), we demonstrate that our answer consistently and universally enhances the accuracies of existing CF algorithms (e.g., item-primarily based CF, SVD-primarily based CF, and SVD++) by a pair of.five to five times on average. Furthermore, our resolution improves the running time of those CF ways by 1.two to 2.3 times when its setting produces the best accuracy.