Research On The Algorithm And Application Of Link Discovery

Link Discovery(LD) is a new research field in data mining.It is aimed at identifying known,complex,and multi-relations patterns automatically from large-scale and heterogeneous datasets.More importantly,it is to find unknown but very useful links which indicate the potential threatening activities.Link discovery is proposed for anti-terrorism mainly.Now link discovery has been applied into a wide range of research topics,including law investigation,fraud detection,network analysis,telecommunication,and so on.Recently the research of link discovery is concerning about the methods,system design,applications,confidence intervals and performance valuation of link discovery.The dissertation mainly focuses on the methods and applications of link discovery on the basis of the review of recent works.The main contents are described as follows:Firstly,the history of the development of data mining is described in detail. Then the dissertation discusses about the necessity and inevitability of the appearance of the new field of link discovery.The thesis also discusses about the basic theory and applications of Link discovery(LD),and analyzes the main methods, classical systems.Secondly,the thesis analyzes the problems of the classical novel node discovery in the unsupervised link discovery in detail.The problem is solved by K-neighbor(KNN) method based on the distance to find the abnormal data,but due to the global threshold,KNN is weak in dealing with the dataset which contains different density.Further,KNN is not considering about the weight of each char-acter, and will have a bad influence in the application.On this basis,the dissertation suggests a new method of unsupervised link discovery based on the relative density.The detailed steps of the algorithm are given and the experiments are also proposed.Thirdly,the new method of unsupervised link discovery has been applied into the field of anti-money laundering.According to the complexity of money laundering activity and the limitation of the traditional data mining methods,the thesis uses the proposed new method to find the doubtful account so as to improve the efficiency of investigation.On this basis,a framework of an anti-money laundering system based on link discovery mining technique has been proposed,and the func- tions of the system modules have also been discussed.The dissertation proposes a new method of link discovery according to its problems on the basis of the current works,and also uses the new method to solve the problem of anti-money laundering.This research has brought a new thinking to link discovery,and an effective tool to anti-money laundering.