Enhancing Machine Learning Classifiers for Prostate Cancer Detection through Feature Selection

Prostate cancer causes most death among men, which makes early detection very important. In this work, we use microarray gene expression data, which contains a large amount of information, to help detect prostate cancer early. To handle this complex data, we use different methods like machine learning, feature selection, and dimensionality reduction. Chi<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> Probability Density Function (Chi<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> PDF) and Cuckoo Search (CS) are used to reduce the number of features, while Particle Swarm Optimization (PSO) and Harmonic Search (HS) help choose the most useful ones. We test five classifiers—Gaussian Mixture Model (GMM), SVM with RBF kernel, Expectation Maximization (EM), Linear Regression, and Nonlinear Regression—with and without feature selection. The results show that using SVM-RBF with feature selection gives the highest accuracy for detecting prostate cancer early