Deep Learning Model for Integration of Clustering with Ranking in Social Networks

Abstract

Now a day Deep Learning has become a promising and challenging research topic adaptable to almost all applications. On the other hand Social Media Networks such as Facebook, Twitter, Flickr and etc. become ubiquitous so that extracting knowledge from social networks has also become an important task. Since both ranking and clustering can provide overall views on social network data, and each has been a hot topic by itself. In this paper we explore some applications of deep learning in social networks for integration of clustering and ranking. It has been well recognized that ranking systems without taking cluster effects into account leads to dumb outcomes. For example ranking a database and deep learning papers together may not be useful. Similarly, clustering a large number of things for example thousands of users in social networks, in one large cluster without ranking is dull as well. Thus, in this paper, based on initial N clusters, ranking is applied separately. Then by using a deep learning model each object will be decomposed into K-dimensional vector. In which each component belongs to a cluster which is measured by Markov Chain Stationary Distribution. We then reassign the objects to the nearest cluster in order to improve the clustering process for better clusters and wiser ranking. Finally, some experimental results will be shown to confirm that the proposed new mutual enforcement deep learning model of clustering and ranking in social networks, which we now name DeepLCRank (Deep Learning Cluster Rank) can provide more informative views of data compared with traditional clustering.