The relationship between prediction accuracy and pre-information in collaborative filtering system
Sun Ok Kim · 2010
Abstract This study analyzes the characteristics of preference ratings by dividing estimatedvalues into four groups according to rank correlation coecient after obtaining pref-erence estimated value to user’s ratings by using collaborative ltering algorithm. Itis known that the value of standard error of skewness and standard error of kurtosislower in the group of higher rank correlation coecient. This explains that the prefer-ence of higher rank correlation coecient has lower extreme values and the di erencesof preference rating values. In addition, top n recommendation lists are made afterobtaining rank tting by using the result ranks of prediction value and the ranks ofreal rated values, and this top n is applied to the four groups. The value of top nrecommendation is calculated higher in the group of higher rank correlation coecient,and the recommendation accuracy in the group of higher rank correlation coecient ishigher than that in the group of lower rank correlation coecient. Thus, when usingstandard error of skewness and standard error of kurtosis in recommender system, rankcorrelation coecient can be higher, and so the accuracy of recommendation predictioncan be increased.Keywords: Collaborative ltering, rank correlation coecient, recommender system,top n.