Improving the accuracy of m-distance based nearest neighbor recommendation system by using ratings variance

Abstract

M-distance based recommendation system (MBR) is a nearest neighbor basedrecommendation method which uses the average of ratings given to an item as the attribute ofthat item. This attribute is used to determine similar items. Then, the average of the ratinggiven to the similar items to an item of the active user determines the rating of that item. Inthis paper, to decrease the error of MBR, by combining the following ideas, eight MBR-basedrecommendation systems are proposed: (a) Using the variance of item ratings in addition tothe average of item ratings, as two attributes of an item, for determining similar items in anitem-based nearest neighbor method; (b) Using the variance of user ratings in addition to theaverage of user ratings, as two attributes of a user, for determining similar users in a user-basednearest neighbor method; (c) Using a weighted average method for combining the ratings ofsimilar items or similar users; (d) Using ensemble learning. Experimental results on realdatasets show that our proposed EVMBR and EWVMBR which use ensemble learning havethe least error. The error of the suggested EWVMBR is at-least 20% lower than that of MBR,Slope-One, P-kNN, and C-kNN.