Abstract

Now a day it is an emerging challenge in data mining to mining richly structures datasets where the objects are\nlinked. Links between objects may demonstrate certain patterns which can be helpful for many types of data mining tasks and\nare usually very hard to capture with traditional statistical models. Several datasets of interest today are best described as a\nlink collection of inter related objects. These may be represent homogeneous networks in which there are single-object type\nand link type (eg. people connected by friends links or the World Wide Web, a collection of linked web pages) or richer\nheterogeneous networks in which there may be multiple object and link types and possibly other semantic information.\nExamples of heterogeneous networks include those in medical domain describing patients, diseases, treatments and contacts,\nor in bibliographic domains such as describing publications, authors and venues. Link mining refers to data mining\ntechniques that explicitly consider these set of links when building descriptive or predictive models of the linked data.\nCommonly link mining tasks include object ranking, collective classification, link prediction, group detection and sub-graph\ndiscovery. It is an exciting and rapidly expanding area. In this article we review link mining tasks, some of the common\nemerging themes and discuss ongoing link mining challenges, open issues and suggest ideas that could be opportunities for\nsolutions. The most conclusion of this article is that providing an idea to usage link mining techniques from link mining to\nhelp to construct the Semantic Web as well as providing future scope in Link Mining.