Finding Accuracy in Feature Selection Using Firefly Algorithm with Rough Set theory

Abstract

Feature selection techniques play a vital  role in bioinformatics applications. In addition to the  large group of techniques that have already been  developed in the machine learning and data mining  fields, specific applications in bioinformatics have led to  possess of newly proposed techniques. In this paper, a  method for feature selection is based on Firefly  Optimization (FFO) with Rough Set Theory(RST) is  proposed. Data sets include a large volume of features  with irrelevant and redundant features. Redundant and  irrelevant features reduce accuracy. The main aim of this paper is to select a subset of relevant features. A statistical metric-based feature selection technique has  been proposed in order to reduce the size of the  extracted feature vector. The proposed method shows  the improvement significantly in terms of performance  measure metrics: accuracy, sensitivity, specificity,  computation time and so on. FFO technique is applied  to determine the features globally according to the light  intensity. Then the selected features are grouped  together to make a subset and applied RST to find the  optimized feature. This optimized feature is used to  analyze the protein information in the organisms and  improve the feature selection accuracy and reduce the  computation time in protein data analysis.