Research On Incremental Outlier Detection Method Based On Rough Entropy

The purpose of outlier detection is to find the abnormal behavior points in the data set.And outlier detection is widely used in public security,intrusion detection and other fields.Although the existing research has achieved a lot of results in these fields,there are still problems such as parameter dependence,low inspection accuracy,and ineffective processing of uncertain data.The outlier detection method based on rough entropy is a new method for outlier detection proposed in recent years.The rough entropy is used to detect outliers in the method,so this method can effectively deal with uncertain data.However,in the actual detection process,more comprehensive information may not be obtained under the time limit constraint,this often makes the detection of outliers impossible to complete in one step.On the basis of the incremental cognitive theory,a new method of the incremental outlier detection based on rough entropy is proposed in this thesis.The main research is as follows:(1)Based on the algorithm of incremental cognitive theory,an improved algorithm is proposed.The basic method of incremental cognition is studied,a new cognition judgment condition is proposed,and the original incremental cognition algorithm is improved.(2)On the basis of the incremental cognitive theory,an incremental outlier detection method is constructed.Based on the idea of cognitive priority relationship in incremental cognition,a new cognitive priority relationship is proposed,and a phased outlier detection method is constructed,that is,the incremental outlier detection method.(3)Based on rough entropy,the incremental outlier detection method is proposed.The original rough entropy is improved,and an outlier detection method based on improved rough entropy is proposed.Combined with the incremental outlier detection method,the incremental outlier detection method based on rough entropy is proposed.