Departmental Bulletin Paper An Improved Algorithm for Mutual Friends Recommendation Application of SNS in Hadoop


102015-03-24 , 法政大学大学院情報科学研究科
In Social network system, existing “people you might know” or “Mutual friends” recommending applications are commonly utilized to list two-hop mediate relationships that one could have with another in order to tighten the bonds among groups. However, as the number of SNS users increase dramatically, the relationship data get so huge that the performance of Mutual Friends recommendation system becomes an urgent problem considering the developers’ requirements. Here we propose a sorting algorithm in Hadoop----a parallel computing framework, to enhance the efficiency of “Mutual friends” recommendation process by taking advantage of the novel map reduce model. In the revised application, original sorting algorithm of intermediate data, merge sort, is replaced by a more time saving sorting approach which introduces a B-Tree like data structure, 2-3 Tree to store the user friendship data and conducts the sorting process. As number of user increases, the revised user defined map reduce functions perform better than the conventional design in time consuming aspect.

Number of accesses :  

Other information