A Maximum Relevance Minimum Redundancy and Random Forest based feature selection and classification of gene expression data

Feature selection is essential for improving the accuracy of models and interpretability in the field of machine learning. This research paper investigates the application of Minimum Redundancy Maximum Relevance (mRmR) feature selection on gene expression datasets to enhance the performance of classification models. The study focuses on the critical task of identifying relevant genes that contribute significantly to the classification of biological samples. mRmR, known for its ability to select informative and non-redundant features, is employed to rank and select a subset of genes based on their relevance to the classification task. The performance of the mRmR-enhanced Random Forest model is evaluated and compared against a baseline model without feature selection. Along with random forest classifier, SVM and k-NN are also used for classification and their accuracies are being evaluated. The experimental results demonstrate the efficiency of the mRmR feature selection method in improving the classification performance of the three classification models on gene expression data.