An Outlier Detection Method based on Error-based Pruning for Identifying Unusual Data in Medical Data Sets

摘要

The task of outlier detection is to find data that are exceptional when compared with the rest large amount of data. Detection and analysis of such data objects is very important in many domains including medical care. However, most existing outlier detection methods suffer from some limitations. To overcome the limitations a new method Error-based Pruning based Outlier Detection (EBPBOD) is proposed. Experimental results on UCI medical data sets demonstrate the availability of EBPBOD. Furthermore a new way to explain these special data is presented. Application to a real medical data set Clinics find some interesting patterns hidden behind the outlying data.