Utilizing an Integrated Feature Selection Technique in Ovarian Cancer to Solve Classification Problem

Abstract

Ovarian cancer is a famous and extremely deadly disease in the female reproductive organ. The fundamental procedure of enhancing the performances of microarray cancer data classification is feature selection or dimensionality reduction. It also diminishes the complexity of classifiers like misclassification, overfitting, etc. Feature selection is usually more essential when the size of features is comparatively high in the dataset. Constructing and choosing the best feature selection and classification algorithm is usually significant for optimizing ovarian cancer classification problems. In this research, we introduce a structure called the Integrated Feature Selection (IFS) technique. Two integrated feature selection techniques were proposed to solve classification issues these are Evolutionary Non-dominated Radial Slots Based Algorithm (ENORA) combined with Evolutionary Algorithm (EA) and Non-dominated Sorted Genetic Algorithm (NSGA2) combined with Harmony Search Algorithm (HSA). Three different classifiers are utilized as machine learning algorithm which evaluated ENORA-EA and NSGA2-HSA. ENORA-EA and NSGA2-HSA selected 55 and 37 most optimal features. 99.60% higher classification accuracy attained with artificial neural network (ANN) and Linear Discriminant Analysis (LDA) classifiers. For the proposed IFS algorithm’s efficiency justification, experimental outcomes were contrasted with miscellaneous algorithms. Particularly, the experimental results show that the proposed two IFS techniques clearly outperformed the other approaches.