Study Of Outlier Data Mining Algorithm Based On Web Service Security

Data mining is a potential and useful information extracting process from abundant, imperfect, indistinct and random data. During this process, some data objects differing from the normal data models may be included in database, they are called outliers. Searching for outliers is one of the data mining technologies, which can also be called outlier analysis, outlier detection, exceptional mining, events detection, minimum mining, deviation detection etc. Outliers may be "dirty data", also be meaningful events. From the angle of knowledge discovering, especially in some application, the irregular events may be more interesting and more valuable than the regular ones. The exceptional detection can provide some vital information, which is helpful to find some true but unexpected information. Therefore, outlier data detection and analysis is a significant research.This thesis puts emphasis on outlier data mining based on web service log. It introduces systematically some basic theories of data mining and discusses the outlier detection methods based on statistics, distance and deviation. The article mainly discusses outlier data mining algorithms based on distance, namely index-based, nested-loop and cell-based, and analyzes the merits and defects from the angle of time and space complexities. Considering the inefficiency and the trouble to set parameters may make it difficult to implement such a method. Therefore, the outlier defining method based on distance sum is put forward. By making use of the approximate detection algorithm based on sample, outlier data detection of web service log is implemented. An effective result is obtained in the experiment.