Cluster Based Architecture and Algorithm to Improve the Design of Social Networks

Cluster Based Architecture and Algorithm to Improve the Design of Social Networks

Saurab Dutta, Payel Roy
DOI: 10.4018/IJVCSN.2017070103
(Individual Articles)
No Current Special Offers


In a social network people are connected by relationships, business purpose or transaction activity. The increasing demand of social network analysis and how to improve the architecture is of utmost importance for the organizations who are regularly trying to improve the service through social network analysis. Social network analysis views social relationship in terms of network theory. Social networks connect people at very low cost and this network acts as a customer relationship management tool for increasing sales of organization in terms of goods and services. Different models are proposed and utilized in different platforms. In this model, the authors have proposed a cluster-based structure to improve performance of social networks.
Article Preview

1. Introduction

1.1. Clustering Approach

Clustering is an expertise in which one or more computers work mutually as a single case by sharing their resources. It is used mostly for redundancy of the services that is hosted on it. In a cluster all the hosts are work together. If any of the host stops working all the services that is hosted on it comes online instantly on to another host with no downtime. A cluster is used for load balancing also apart from the redundancy; in this case whatever the load comes to the cluster is divided by all the hosts in it.

Clustering is an approach where several storage hubs are networked within themselves so that they can form a subnetwork. One cluster head is there which the identification point for that subnetwork is. The cluster head may be one of the storage hubs of that subnetwork. In our paper the cluster head is the main hub of that location. The advantage is if we can identify the cluster head then the total subnetwork can be explored. As it is evident that social network is expanding very often so we need to form small group of subnetworks within the huge network otherwise the management of the network may become a hectic task. Figure 1 shows that every main hub is connected with other main hubs in some relation.

So as storage hubs are increased to provide better and faster performance to the customers, the management of social network is getting more and more importance. The organizations like flipcart, Snapdeal, eBay, etc., are trying to implement efficient techniques to reach their goal of no. 1 organization in social networks. Doing clustering the stores will be able to pass their inventories to other stores as the stores is nearby. So, cost will be low as well “Out of Stock” concept will be evaporated.

Figure 1.

Every main Hub is connected with other main Hub with some relation


1.2. KNN Angorithm

K-Nearest Neighbour (KNN) is a non-parametric lazy learning algorithm. When we say a technique is non-parametric, it denotes that it does not create any hypothesis on the fundamental data distribution. This is pretty helpful, as in the actual world, most of the realistic data does not follow the typical theoretical assumptions made. The non-parametric algorithms like KNN come to the release there. KNN is also a lazy algorithm (Figure 2). It does not exercise the training data points to do any simplification. In other words, there is no precise training phase or it is very negligible. This means the training phase is very fast. The lack of simplification means that the KNN maintains all the training data. All the training data is required during the testing phase. This is in distinction to other methods like SVM where we can abandon all non-support vectors without any difficulty. Most of the lazy algorithms particularly KNN constructs decision based on the entire training data set.

In social network the relations are of utmost importance. So do detect the most valuable neighbor we propose to modify KNN (K Nearest Neighbor Algorithm) where based on the past association and interaction as well as on cost comparison.

Figure 2.

K-NN Algorithm for finding nearest node


This algorithm is suitable for trained data as the goal is to find out the closest neighbor based on certain criteria so that we can find the most favorable case. So we are actually implementing predictive analysis each sub hub is connected to its respected main hub or main warehouse. So from each data warehouse supply can be done to any data mart.

Complete Article List

Search this Journal:
Open Access Articles: Forthcoming
Volume 10: 4 Issues (2018)
Volume 9: 4 Issues (2017)
Volume 8: 4 Issues (2016)
Volume 7: 4 Issues (2015)
Volume 6: 4 Issues (2014)
Volume 5: 4 Issues (2013)
Volume 4: 4 Issues (2012)
Volume 3: 4 Issues (2011)
Volume 2: 4 Issues (2010)
Volume 1: 4 Issues (2009)
View Complete Journal Contents Listing