An integrated method for feature selection using fuzzy information measure

The information is flooded with heterogeneous data sources and it generates over 2.5 quintillion bytes every day from communication devices, social media, consumer transactions, online behaviour and streaming services. To overcome this difficulty irrelevant and redundant data are to be removed using the technique called feature selection. The goal of the feature selection is to find the minimum set of attribute. The results are implemented by MATLAB and WEKA tool for feature selection and classification respectively. This research work is validated using different datasets namely Pima Diabetic, Breast Cancer, Ecoli, Iris, Sonar and Student which are available in UCI repository. Model performance is evaluated by using Precision, Recall and F-Measure performance metrics. The selected subset features are used to compare different feature ranking methods like Gain Ratio, Relief, Chi Square and OneR. The experimental inference reveals that the proposed algorithms are efficient in selecting minimum features for the feature subset and gives higher accuracy rate.