CFS-FA Enabled Ensemble Model for Cancer Diagnosis

Abstract

Cancer is increasingly playing a significant role in the increase in the number of fatalities occurring all over the globe, particularly in developing countries. The degree of healing from cancer is increased when it is discovered early. Microarray data have been used to create several cancer classification models, which have been developed using machine learning, which has a large number of dimensions. Applying the machine learning algorithms to the high dimensional microarray dataset will result in an issue known as Small Sample Size (SSS) which can reduce the classification and prediction accuracy. As a result, before classifying the data, it is necessary to minimize the size of the dataset using any accessible approach. An ensemble model can help in diagnosing the disease more effectively as it uses the prediction level from different classifiers. The current study the focuses on implementing the ensemble method for microarray dataset. As a cancer diagnosis tool, we have proposed an ensemble method based on stacking that includes SVM, RF, and Naive Bayes, as well as Correlation Feature Selection (CFS) and Firefly Algorithm (FA). The CFS will be used for dimensionality reduction, the FA will be used for optimization, and the ensemble Support Vector Machine (SVM), Naive Bayes (NB), and Random Forest (RF) will be used for classification. The accuracy, F1 score, precision, and recall of the model will be used for performance analysis. The proposed ensemble model is compared in contrast to the hybrid model such as SVM-CFS-FA, RF-CFA-FA, and NB-CFS-FA to show the effectiveness.