An Ensemble Feature Selection Framework of Sonar Targets Using Symmetrical Uncertainty and Multi-Layer Perceptron (SU-MLP)
In Data Mining applications, feature selection methods has become popular area of investigation. It has become priority strategy to get significant knowledge about any chosen domain. In literature many ways are existed for feature selection (Filter, Wrapper, Embedded). With the availability of large characters of data, the feature selection has become necessary in the process of Data Mining. In this article, we presented and examined a new feature selection technique on the basis of Symmetrical Uncertainty (SU) and Multi-Layer Perceptron (MLP), then we estimated the performance of rule-based (Jrip), tree-based (J48), lazy (KNN) algorithm with ensembling techniques such as boosting, bagging on SONAR (sound navigation and ranging) data. In our proposed framework, we divided the most prominent features derived by SU into finite number of clusters. Each cluster formed by our technique has unique features. As an initial step, MLP is applied on all clusters and decided the foremost cluster as per classifier accuracy. The finest cluster of features are examined with Jrip, J48, KNN with ensembling approaches and compared with existing feature selection techniques Gain Ratio Attribute (GR), Information Gain(IG), Chi Squared Feature selection (CHI). Cluster of features originated by the proposed framework has recorded improvement with the most of the classifiers than traditional methods.
KeywordsBagging Boosting Multi-layer perceptron Symmetrical uncertainty SMOTE
- 10.Prasad Potharaju, S., Sreedevi, M.: An improved prediction of kidney disease using SMOTE’. Indian J. Sci. Technol. 9. https://doi.org/10.17485/ijst/2016/v9i31/95634 (2016)
- 11.Potharaju, S.P., Sreedevi, M.: Ensembled rule based classification algorithms for predicting imbalanced kidney disease data. J. Eng. Sci. Technol. Rev. 9, 201–207 (2016)Google Scholar
- 13.Kotaria, V., Changa, K.C.: Fusion and gaussian mixture based classifiers for SONAR data. In: SPIE defense, security, and sensing, pp. 80500U–80500U. International Society for Optics and Photonics (2011)Google Scholar
- 14.Chen, C.H., Lee, J.D., Lin, M.C.: Classification of underwater signals using neural networks. Tamkang J. Sci. Eng. 3(1), 31–48 (2000)Google Scholar