An Unsupervised Approach For Selection of Candidate Feature Set Using Filter Based Techniques
Sai Prasad Potharaju · DergiPark (Istanbul University) · 2018
Clusteringis an unsupervised Data Mining approach.In this research article, wehave proposed an unsupervised approach using filter based featureselection methods and K-Means clustering technique to derive thecandidate subset. Initially, score of each feature is recorded usingtraditional filter based methods, then normalized the dataset usingMin-Max technique, then formed the unsupervised dataset. K-Meansalgorithm is employed on the dataset to form the clusters offeatures. To decide the strong subset, Multi Layer Perceptron(MLP) isapplied on each cluster. Based on the minimum Root Mean Square (RMS)error rate given by MLP best cluster is selected. This framework iscompared with traditional methods over six well known datasets havingthe total features in between 34 and 90 using various classificationalgorithms. The proposed method has shown competitive performancethan few of the traditional methods.