Feature selection in support vector machines for outlier detection
Deepika Rani Bansal, Brij Kishore · 2018 Second International Conference on Electronics, Communication and Aerospace Technology (ICECA) · 2018
This paper proposes a new association measure to be used in mRMR framework for feature selection. Also a filter-type approach is suggested unlike successive selection in all previous mRMR methods. Three different combinations of association are evaluated for classification accuracy and error. The reduced dataset is used in support vector machines applied to two-class classification in datasets particularly designed to benchmark methods for outlier detection. The proposal reduces classification error to acceptable levels.