Experimented Multi-Objective Teaching-Learning-Based Optimization for Medical Data Analysis

Feature selection (FS) analyzes the most important features to enhance classification accuracy during preprocessing. The objective of FS techniques is to reduce the number of input variables and computational load to achieve optimal classification performance. However, finding the best features is challenging due to high dimensionality, with many features but limited data samples. Metaheuristic optimization algorithms (MOAs) are crucial in creating the best subset of features through exploration and exploitation phases. This study experiments with Teaching-Learning-based Optimization (TLO) as MOA from several diverse binary medical datasets. TLO draws attention to its excellence in improving performance with minimal features. TLO is a fairly straightforward algorithm without algorithm-specific parameters, with rapid convergence and implementation yet effectiveness. In this study, we have also conducted four classifiers to obtain the best model for minimizing the features and maximizing the performance of the model, i.e., decision tree (DT), random forest (RF), k-Nearest Neighbour (KNN), and support vector machine (SVM) with the default parameter of each classifier. As the result, TLO shows its excellence in achieving minimal features, but still outstanding performance with 100% accuracy, sensitivity, specificity, and precision. TLO is successfully applied to varying diverse medical datasets with varying number of features.