Optimizing high dimensional data classification with a hybrid AI driven feature selection framework and machine learning schema
Amina Salhi, Rayan Alshamrani, Ashrf Althbiti, Atef Ismail, Manar Abd-ElRahman, Basma M. Hassan
Abstract
Feature selection (FS) is critical for datasets with multiple variables and features, as it helps eliminate irrelevant elements, thereby improving classification accuracy. Numerous classification strategies are effective in selecting key features from datasets with a high number of variables. In this study, experiments were conducted using three well-known datasets: the Wisconsin Breast Cancer Diagnostic dataset, the Sonar dataset, and the Differentiated Thyroid Cancer dataset. FS is particularly relevant for four key reasons: reducing model complexity by minimizing the number of parameters, decreasing training time, enhancing the generalization capabilities of models, and avoiding the curse of dimensionality. We evaluated the performance of several classification algorithms, including K-Nearest Neighbors (KNN), Random Forest (RF), Multi-Layer Perceptron (MLP), Logistic Regression (LR), and Support Vector Machines (SVM). The most effective classifier was determined based on the highest level of accuracy. Additionally, this research introduces hybrid algorithms such as TMGWO (Two-phase Mutation Grey Wolf Optimization), ISSA (Improved Salp Swarm Algorithm), and BBPSO (Binary Black Particle Swarm Optimization) for identifying significant features for classification. A comparative analysis was conducted to assess the performance of these hybrid FS algorithms from various perspectives. We also compared the performance of classifiers on datasets with and without FS, measuring improvements in accuracy, precision, and recall. Among the algorithms tested, the TMGWO hybrid approach achieved superior results, outperforming the other experimental methods in both feature selection and classification accuracy.