Hater maturity oriented k-nearest neighbor collaborative filtering algorithms

Huge data volume with high sparsity in Collaborative filtering (CF) recommender motivates the paper to utilize information underlying sparsity to reduce data. An aspect of sparsity is different amount of rated items for a given user, defined as Simple rater maturity (SRM). It is reasonable that ratings from users with higher SRM are more credible. Hypothesis of Rational authorities bias (H-RAB) is proposed to capture this factor, stating that higher prediction accuracy can be attained to emphasize ratings from more mature referential users. Modifications based on H-RAB are designed in two aspects: RAB-aware weight scaling is a fine tuning method by scaling original similarities with SRM; RAB-aware data reduction is a more audacious one that suggests pruning all referential users with less SRM than a given maturity threshold, and particularly applied to kappa-Nearest neighbor CF algorithms. Experimental results from positive and negative tests on three public available CF datasets justify the soundness of modifications and validity of H-RAB.