Study Of Mining Outliers Based On Interestingness

Today, more and more data were accumulated in companies. For some of them, exceptional data are more valuable than normal ones. The requirement of mining data of such type increases every year in both our country and foreign ones. Wide ranges of study area were formed. Technology about the area have been put into many applications in our modern society and acted as a useful tool. For example, the analysis of prediction of risk management of enterprise, network intrusion detection and so on. This technology need detect exceptional case sensitively and speedily. All these are backgrounds of mining outliers.In recent study of data mining, most study of interesting rules based on associational rules or strong associational rules. The work of outlier detection based on single dataset. For these reasons, this paper brought out a new definition about interestingness that named peculiarity. Combining the conception of peculiarity and outliers, the paper focus on mining outliers in database. A new advanced algorithm GPOD were brought out based on GridLOF and LOF, which could mine outliers out more efficiently. It can also expand to be used in multi-database, which can provide more useful information to users. In the beginning, the paper introduced conceptions and definitions relation to the content of study. Then it gave the definition and applications of interestingness in detail. The technology of outlier detection and clustering were important part of the issue. After this, the process of mining outliers by GPOD and how to get rules among databases were described. A series experiments were executed to test the merits of GPOD and its improvements. Finally, the summary of this paper was given including more work we should do about the study and the main direction of next task.