Auto-Encoder Based K-Means Clustering Algorithm
Venubabu Rachapudi¹, S. Venkata Suryanarayana², T.Subha Mastan Rao^3
1Venubabu Rachapudi, Department of Computer Science and Engineering, Koneru Lakshmaiah Education Foundation, Vaddeswaram, Guntur (Andhra Pradesh), India.
²S.Venkata Suryanarayana, Department of Information Technology, CVR College of Engineering, Hyderabad (Telangana), India.
³T.Subha Mastan Rao, Department of Computer Science and Engineering, CMR Technical Campus, Hyderabad (Telangana), India.
Manuscript received on 07 March 2019 | Revised Manuscript received on 20 March 2019 | Manuscript published on 30 March 2019 | PP: 1223-1226 | Volume-8 Issue-5, March 2019 | Retrieval Number: E3378038519/19©BEIESP
Open Access | Ethics and Policies | Cite | Mendeley | Indexing and Abstracting
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC-BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)

Abstract: Identifying network, or clusters, in graphs is a task of great importance when analyzing network structure.A telecommunications provider, for instance, would like to identify communities of customers that place a large amount of calls to each other, in order to create more effective, directed marketing campaigns. Linear or non-linear data transformations are widely used processing techniques in clustering.There are wide range of algorithms and methods that can be applied, in order to identify communities in network structures, such as Spectral Clustering, Modularity Maximization, and Hierarchical Clustering etc. Spectral Clustering is one of the traditional algorithms that is suitable for network clustering because it first map the original data points into space such that graph tend to be much more evident, allowing for a subsequent application of standard clustering techniques such as K-Means. Despite its good results, Spectral Clustering presents some computational challenges when applied to very large networks.Recent literature demonstrates that systems and calculations from Deep Learning, for example, layered neural systems based auto-encoders, are appropriate for the task of mapping information focuses into lower-dimensional spaces, which can be helpful for an assortment of further undertakings. In this proposed paper we will make another handling pipeline for non-covering network identification in system structures dependent on K-Means. We will demonstrate that this methodology is like a conventional profound learning auto-encoder in its capacity to obtain reasonable portrayals of the remarkable information in a lower-dimensional space, making it less demanding to play out the information bunching undertaking. We will at that point test its appropriateness for the particular difficulties of network recognition in systems and contrast its execution and the conventional Spectral Clustering approach.
Keyword: Auto Encoder, Deep Learning, Graph Cluster, K-Means Clustering, Spectral Clustering.
Scope of the Article: Deep Learning