The Online Social Networks Have Existed Computer Science Essay

Published: Last Edited:

This essay has been submitted by a student. This is not an example of the work written by our professional essay writers.

Online social networks have existed since the beginning of the Internet. Online social networking sites like Orkut, YouTube, Facebook and Flickr are among the most popular sites on the Internet. Users of these sites form a social network, which provides a powerful means of sharing, organizing, and finding content and contacts. Popular online social networking sites like Flickr, You-Tube, Facebook, Orkut, and Live Journal rely on an explicit user graph to organize, locate, and share content as well as contacts. The popularity of these sites provides an opportunity to study the characteristics of online social network graphs at large scale. Online social networks are already at the heart of some very popular Web sites. As the technology matures, more applications are likely to emerge. It is also likely that social networking will play an important role in future personal and commercial online interaction, as well as the location and organization of information and knowledge. Social Networks are inherently dynamic entities new members join them, old members stop participating, new links emerge as new contacts are built, and old links become obsolete as the members stop interacting with one other. Much of the research in mining social networks is focussed on using the links in order to derive interesting information about the social network such as the underlying communities, or labelling the nodes with class labels. However, in most social networking applications, the links are dynamic, and may change considerably over time. For example, in a social network, friendship links are continuously created over time. Therefore, a natural question is to determine or predict future links in the social network. The prediction process may use either the structure of the network or the attribute-information at the different nodes.

Link Prediction

In the recent years, the significant popularity in internet has being gained by different types of information sharing systems, including the web, online social network. Social network are modelled as graph structure for social network analysis. Social networks are dynamically structured which grows over time where the addition of new edges, signifying the appearance of new interactions in the social structure and nodes represent people or other entities embedded in a social context, and edges represent specific types of connectivity between them. Link prediction has been attracting researchers in the field of large scale networks or graphs. Link prediction deals with predicting these new edges or links that would be added in the near future based on the current structure of the graph. Basically, link prediction is a sub-field of social network analysis or in other words, that it’s a fundamental data mining task for various application domains, including recommender systems, information retrieval, automatic web hyperlink generation, record linkage, and communication surveillance and other social networks. Link prediction is the problem of predicting which new edges will be added in the near future based on past snapshots of a social network. Social network analysis has been applied to two types of data: persistent relationships (friendships, affiliations etc.) and discrete events (meetings, publications, communications, transactions, etc.). Previously work related to link prediction has been mostly focused on a static network, where a partial network structure is known and the objective is to predict the hidden links, even if the underlying data is known to change over time. Dynamic interactions over time introduce another dimension to the challenge of mining and predicting link structure. Here considering the task of link prediction at time T then the goal is to predict the links at time T + 1. Since static graphs treat all links as appearing at the same time, they do not capture key temporal characteristics such as duration of contacts, inter-contact time, recurrent contacts and time order of contacts along a path. For this reason, they give us an overestimate of the potential paths connecting pairs of nodes and they cannot provide any information about the delay associated with the information spreading process.

Problems in Link Prediction

The Biggest challenge in large graphs is covering the entire connected component. One can only obtain the set of links into or out of a specified node at a particular time step. Efficiently analyzing the graph is important in online social network since the graphs are large and highly dynamic in nature. Common algorithms for searching graphs include breadth-first search (BFS) and depth-first search(DFS). Searching over an entire connected component is not feasible.

In Link prediction 'Accuracy' is the measure to evaluate efficiency of the algorithm, ‘Mathematical Proof of Correctness' is not present. Accuracy is defined as "what percentage of the 'predicted links', will really exist in the future".

Link prediction face following problems

link existence problem: is there exist a link in future

link type problem: what type of link exist

link cardinality: more than one link between same pair of nodes in a social network

link weight: links have different weights associated with them

We first concentrate on the first two problems. We try to find the mechanism to find the link existence problem and the type of link. We can extend these two problems easily to the link cardinality and link weight problem.


Link prediction refers to the task of predicting the edges or links that will be added to a social network in the future based on past snapshots of the network. Previously most of the work related to link prediction is based on a static network, where the only partial view of the network with a limited number of links is known. Prediction is performed on these links to find the hidden or missing links in the network ignoring the fact that the data is known to change over time. Dynamic consideration of the link existence or link interactions over time is a new area to explore in data mining and predicting link. Since in static graph all the links are analyzed at the same time in a single snapshot where the temporal characteristic are not considered such as how long the nodes are in contact with each other. Dynamic graph have a sequence of temporal information associated with the nodes and edges of the graph such as how many times the nodes interact with their associated nodes and at what time interval was the interaction or the occurrence of an event happened. By considering this temporal information we can extract some important temporal attribute such as interaction time, as the value of temporal attribute change the strength of interaction also changes which help in predicting the links in near future. Many researchers in recent years have worked on the dynamic link prediction considering time as a key feature. Still there is a need of more accurate and an efficient link prediction measure based on time feature as the network expands exponentially.

The Objectives of the Thesis

The main objective of this thesis is to propose a novel approach Link Score for predicting link based on the temporal information which consider time aware feature for link prediction. Here are the objectives of this thesis in order to accomplish its aim:

To study Social network and Link analysis & work done related to it.

To study graph theory & work done related to it.

To develop a new time aware based link prediction algorithm to solve the problem of predicting link in future.

To implement the proposed algorithm and test the algorithm on the Cod mat datasets and compare the result with other existing method.

Scope of Dissertation

The scope of this research is mentioned below:

Developing a new technique for solving the Link prediction problem in social networks

The implementation of the proposed work.

Analyze the proposed algorithm by testing it on a dataset containing different number of

nodes with different edges or links.

Dissertation Organization

Organization of this thesis is as follows:

Chapter 1 Introduction: This chapter presents the overview of thesis, motivation, objectives of the thesis & organization of the thesis.

Chapter 2 Social network : Ooverview of social network and its background is describes in this chapter,also basic introduction of graph theory related to social network & links in graph is explained along with various application and basic terminologies.

Chapter 3 Link Prediction: This chapter discuss about link prediction, provides details of existing link prediction techniques. Analysis of link prediction and link detection is given also a short introduction to temporal link prediction, dynamic network analysis with temporal statistics is mentioned.

Chapter 4 Literature Survey for Link Prediction: This chapter describes literature survey of the existing work done in link prediction and also related work in time aware based link prediction.

Chapter 5 Proposed method: This chapter describes the proposed methodology, proposed algorithm and introduce a new index as time path index for evaluating path weight in the graph.

Chapter 6 Experimental details & Result: This chapter checks the performance of the LScore algorithm and the results are compared with the existing methods.

Chapter 7 Conclusions and Future Work: This chapter finally concludes the thesis stating that LScore outperformed other the method which are not considering temporal features for link prediction with better accuracy and also future directions are given at the end of the chapter.