Binary Learning Cooking Algorithm with Relief for Feature Selection of High-dimensional Gene Expression Data

Feature selection (FS) has a significant impact on machine learning. The objective is to find the most important features while maintaining high classification accuracy by identifying the possible fewest features from the high-dimensional dataset. Researchers prefer to use the filter-wrapper model by considering different metaheuristic methods for accurate feature selection. But in some cases, the proposed algorithms become less productive and achieve high time complexity. This paper presents a hybrid model of Relief with Binary Learning Cooking Algorithm (BLCA) to select optimal features. The algorithm builds on the Learning cooking algorithm (BLCA), a new optimization method inspired by the learning process "How to learn cooking". We tested how well this method responds to FS challenges in high-dimensional datasets using six cancer datasets, many with different ranges of features. We compared different existing FS algorithms with R-BLCA using accuracy, precision, sensitivity, specificity, and F-Score. The results show that the proposed algorithm performs better than other FS methods with an accuracy of 99. 91% for the SRBCT dataset.