A Design of Group Recommendation Mechanism Considering Opportunity Cost and Personal Activity Using Spark Framework

Abstract

Group Recommendation is a method of recommending a specific item (e.g. product, service) to a group formed of several members. Least Misery, one of the representative group recommendation method, can recommend items considering group dissatisfaction, but has a drawback of low recommendation accuracy and the other method, Average methods has high accuracy but cannot consider the group dissatisfaction. In this paper, we developed a group recommendation method that improves the recommendation accuracy by measuring the Opportunity Cost (Opportunity Cost is the largest value of the remaining item that is discarded when you select a specific item) and personal activity, taking into account group dissatisfaction. Hadoop-Spark Framework was used in the experiment to distribute large scale of data safely and process efficiently. In Experiment result the proposed group recommendation method improved the recommendation accuracy by 27% while considering the dissatisfaction of the group members compared to the Least Misery method.