Link Analysis Essay

Link Analysis encompasses multiple distinct tasks [2]. These tasks can be grouped into a small set of overall categories. Fig. 1 shows hierarchy of these link analysis tasks. We have discussed some of these categories below.

i. Link-based Object Classification (LOC) Link-based object classification predicts the object’s category based upon its attributes and links to other objects. An example of link-based object classification is webpage classification. The class of a web page is determined with the help of words on the page, anchor text, links between pages and other attributes of the links and pages. Similarly, link-based object classification can predict the category of paper in the bibliographic domain, based on its citations [4].

The goal is to assign a relative quantitative assessment with all nodes in a network. Web page ranking is one of the most well-known LOR tasks. We can say that LOR is a more fine-grained version of LOC [2]. In social network analysis, Link-based object ranking is primary task. The purpose is to rank individuals in a specific social network according to their strength and importance. This concept is called centrality. The research in LOR mainly target graphs which have single link type and single object type [5]. HITS , PageRank, Weighted PageRank (WPR) and SimRank algorithms are the most popular algorithms of LOR

This is known as deduplication, alias detection object consolidation and record linkage. Deduplication also could be referred as a group detection task itself [17]. Group detection is an object-centric task. It is sometimes also known as community detection. The purpose of group detection is to make cluster of the nodes into groups, which share common characteristics. Many techniques have been presented in many communities to address this problem. Many approaches of group detection are based on stochastic block modeling from social network analysis