In a typical social network site, a sender initiates an interaction by sending a message to a recipient, and the recipient can decide whether or not to send a positive or negative reply. Typically a sender tries to find recipients based on his/her likings, and hopes that they will respond positively. We examined historical data from a large commercial social network site, and discovered that a baseline success rate using such a traditional approach was only 16.7%. In this paper, we propose and evaluate a new recommendation method that considers a sender’s interest, along with the interest of recipients in the sender while generating recommendations. The method uses user profiles of senders and recipients, along with past data on historical interactions. The method uses a weighted harmonic mean-based aggregation function to integrate "interest of senders" and "interest of recipients in the sender". We evaluated the method on datasets from the social network site, and the results are very promising (improvement of up to 36% in success rate).

Item Type:	Refereed Conference Paper
Keywords:	recommender systems, social network analysis
Research Division:	Information and Computing Sciences
Research Group:	Artificial intelligence
Research Field:	Artificial intelligence not elsewhere classified
Objective Division:	Information and Communication Services
Objective Group:	Information systems, technologies and services
Objective Field:	Application software packages
UTAS Author:	Kim, YS (Dr Yang Kim)
ID Code:	94649
Year Published:	2010
Deposited By:	Information and Communication Technology
Deposited On:	2014-09-15
Last Modified:	2014-12-09
Downloads:	0