Optimizing high dimensional data classification with a hybrid AI driven feature selection framework and machine...

Area of Science:

• Machine Learning
• Data Science
• Bioinformatics

Background:

• Feature selection (FS) is crucial for high-dimensional datasets to improve classification accuracy.
• FS minimizes model complexity, reduces training time, and enhances generalization.
• The curse of dimensionality necessitates effective feature selection strategies.

Purpose of the Study:

• To evaluate and compare various classification algorithms for feature selection.
• To introduce and assess novel hybrid algorithms for enhanced feature identification.
• To demonstrate the impact of feature selection on classification performance metrics.

Main Methods:

• Experiments were conducted on the Wisconsin Breast Cancer Diagnostic, Sonar, and Differentiated Thyroid Cancer datasets.
• Evaluated standard classifiers: K-Nearest Neighbors (KNN), Random Forest (RF), Multi-Layer Perceptron (MLP), Logistic Regression (LR), and Support Vector Machines (SVM).
• Introduced and tested hybrid algorithms: TMGWO (Two-phase Mutation Grey Wolf Optimization), ISSA (Improved Salp Swarm Algorithm), and BBPSO (Binary Black Particle Swarm Optimization).

Main Results:

• The TMGWO hybrid approach achieved superior results in both feature selection and classification accuracy.
• Comparative analysis showed significant improvements in accuracy, precision, and recall with FS.
• TMGWO outperformed other experimental methods in identifying significant features for classification.

Conclusions:

• Hybrid feature selection algorithms, particularly TMGWO, offer significant advantages for classification tasks.
• Effective feature selection is vital for improving model performance and avoiding the curse of dimensionality.
• The study highlights the importance of advanced FS techniques in machine learning applications.