Online Discussion Participation Prediction Using Non-negative Matrix Factorization

This paper studies the online discussion participation prediction (OFPP). Online discussion is an application on the Web that provides a cyberspace for users to exchange or share different information. Finding suitable online discussions on Internet becomes difficult as huge amount of information existed. This led to recommendation systems that provide advices to users. In this paper, a weighted non-negative matrix factorization method is used to discover latent user preferences of online discussions such that prediction of user´s participation can be obtained. Experimental results show that with the prediction of user´s preferences, suitable online discussions can be suggested to the user.