ISSN : 2319-7323




INTERNATIONAL JOURNAL OF COMPUTER SCIENCE ENGINEERING


Open Access

ABSTRACT

Title : A SURVEY ON LDA APPROACH IN PREDICTING LINK BEHAVIOR IN A SOCIAL NETWORK
Authors : MD ABDUL NAVEED MASTAN, S.RAVI KISHAN
Keywords : Social Networks, Link Prediction, Learning, Topic Modeling.
Issue Date : September 2013
Abstract : Social network sites (SNS) in recent times are focusing mainly on user interactions. These SNS are attracting the attention of academic and industry researchers who are intrigued by their accordance and reach rapidly. Mainly data mining techniques have been very effective in using the content and graph structure that was available to solve various problems such as friendship link prediction, estimating the percentage of their friendship…etc. Topic models are one among the most effective approaches to discover latent topic analysis and text data mining. One desirable feature of a social network is to be capable to suggest potential friends to its existing users and the approach must be proved to be effective in improving the predictions. Topic modeling approach provides an easy way to analyze large volume of data and the topic modeling techniques like Latent Dirichlet Allocation (LDA) to uncover latent structure in user interests which have to be explored is going to be implemented. By using LDA, the users are predicting their friends and with the how much amount of percentage ratio they are becoming friends. In this review, it has been identified that LDA has a limitation of topic correlation modeling which can be overcome by using CTM (correlated topic model) and it can work better than Arm (Association Rule Mining) for list of 4 or more communities while the tagging can be effectively done when both LDA and association rules are used together.
Page(s) : 183-188
ISSN : 2319-7323
Source : Vol. 2, No.5