Outlier Detection Method Based On Classification

Classification based statistical modeling is often affected by outliers,therefore outlier detection is important issue in classification statistical model.This paper devotes to the outlier detection and classification in spherical and non spherical data,to improve the classification accuracy of the model by outlier detection method.The data X with outliers is transformed into the form of linear regression Y = X?+? and outliers are detected by means of variable selection and -test.The performance of the simulation data and actual data illustrated that the proposed model can effectively identify outliers in the data.The outliers that deviate from the data center are modified to normal ones,then the procedure classification can effectively improve the accuracy of data classification.