COMPARATIVE STUDY OF KSVDD AND FSVM FOR CLASSIFICATION OF MISLABELED DATA

Rajani S Kadam, Prakash R. Devale

Abstract: Outlier detection is the important concept in data mining. These outliers are the data that differ from the normal data. Noise in the application may cause the misclassification of data. Data are more likely to be mislabeled in presence of noise leading to performance degradation. The proposed work focuses on these issues. Data before classifying is given a value that represents its willingness towards the class. This data with likelihood value is then given to classifier to predict the data. SVDD algorithm is used for classification of data with likelihood values

Keywords: Confusion Matrix, FSVM, Outlier, Outlier Detection, SVDD

DOI: https://doi.org/10.15623/ijret.2016.0503018