A METHOD FOR COMMUNITY RECOMMENDATION IN SOCIAL NETWORKS

Abstract

THERE has been an active interest in analyzing Twitter data in last few years, specifically for user recommendation, community discovery, and community recommendation. In this paper we consider community recommendation for Twitter data. The existing approaches for community recommendation have some limitations. The recommended community may consist of users that are not currently active. Such users may not be of much help to the given user. Moreover the given user may be interested to get recommendations of those users who belong to her locality. In order to address these issues, we associate with a user some spatio-temporal fatures like activity state (temporal), and location (spatial), besides topic of interest. We have designed an algorithm for community recommendation by taking in to account these features of the users. We have conducted several experiments using Twitter data that consists of 37995 unique tweets obtained from 125 users. We have compared our algorithm with some existing algorithms. The results indicate that our algorithm gives significantly better results.