Performance Analysis of Parametric and Non-Parametric Machine Learning Algorithms for Multi-Class Classification

Performance Analysis of Parametric and Non-Parametric Machine Learning Algorithms for Multi-Class Classification

Mahziyar Ghaedi1 Arman Galedari2 Mahdi Moghimi3

1) Ph.D. Student, School of Mechanical Engineering, Iran University of Science and Technology, Tehran, Iran Energy Conversion Department Email:
2) M.Sc. Student, School of Mechanical Engineering, Iran University of Science and Technology, Tehran, Iran Energy Conversion Department Email:
3) Associate Professor, School of Mechanical Engineering, Iran University of Science and Technology, Tehran, Iran Energy Conversion Department Email:

IOI Code : ABEB-ABKCDD
Publication : 4th International Conference on Recent Advances in Engineering, Innovation and Technology - Belgium(eitconf.com)
Abstract :
    In this study, the multi-class classification of a dataset consisting of 1,000 samples, 9 features, and 3 classes was investigated using several machine learning algorithms. After data standardization, the dataset was divided into training and validation subsets, and the performance of Maximum Likelihood, Bayesian, Euclidean Distance, Mahalanobis Distance, K-Nearest Neighbors (KNN), Weighted K-Nearest Neighbors (WKNN), and Support Vector Machine (SVM) classifiers was evaluated. The models were assessed using Accuracy, Precision, Recall, F1-score, and cross-validation metrics. The results showed that the Maximum Likelihood classifier achieved the best performance, with an accuracy of 84.67% and an F1-score of 0.7612, whereas the Mahalanobis Distance classifier exhibited the weakest performance. In addition, the non-parametric methods and the Support Vector Machine demonstrated satisfactory classification performance. Finally, the trained models were employed to classify previously unlabeled data. The findings indicate that the statistical characteristics of the dataset and the degree of class overlap significantly influence classification performance, and that selecting an appropriate machine learning algorithm can substantially improve the accuracy and robustness of multi-class classification systems.

Keywords : machine learning parametric methods non parametric methods multi class classification validation training