Interest prediction in social networks based on Markov chain modeling on clustered users
Xianghan Zheng,Dongyun An,Xing Chen,Wenzhong Guo
DOI: https://doi.org/10.1002/cpe.3572
2015-07-14
Concurrency and Computation: Practice and Experience
Abstract:Effective user interest prediction is significant for service providers in a set of application scenarios such as user behavior analysis and resource recommendation. However, existing approaches are either incomplete or proprietary. In this paper, user interest prediction based on the Markov chain modeling on clustered users is proposed with the following procedure: collect dataset from 4613 users and more than 16 million messages from Sina Weibo; obtain each user's interest eigenvalue sequence and establish single‐Markov chain model; and implement user clustering algorithm for the multi‐Markov chain construction in order to divide users into a set of predefined interest categories. The proposed solution is capable of predicting both long‐term and short‐term user interests based on a suitable selection of the initial state distribution, λ. The proposed solution also proves that short‐term interests are consistent with long‐term interests if the influences of social or user‐related events that cause interruptions (e.g., earthquake and birthday) are not considered. Furthermore, experiments show that the proposed solution is feasible and efficient and can achieve a higher accuracy of prediction than that of the other approaches such as Support Vector Machine (SVM) and K‐means. Copyright © 2015 John Wiley & Sons, Ltd.