JIITA, Vol.10 No.3 pp.1359-1369 (2026), DOI: 10.22664/ISITA.2026.10.3.1359
D.Vaishali and S.Lakshmi
Abstract. Classification algorithms generally require an adequate and representative set of training data to establish an appropriate decision boundary. In many machine learning applications, such as automated identification, data collected from multiple sources describing the same physical phenomenon may contain complementary information. In these cases, data fusion can provide deeper insights into the phenomenon than any single source alone.The objective of this paper is to investigate the fusion of multiple machine learning classifiers and to analyze various data fusion schemes. The performance of four classifiers—multilayer perceptrons (MLP), support vector machines (SVM), decision trees (DT), and k-nearest neighbors (KNN)—is first evaluated individually. Their results are then compared when fused in different combinations. Findings indicate that the best performance for pattern recognition applications is achieved by combining the SVM, MLP, and DT classifiers.
Keywords; Classifier Fusion, Multiple Classifier System(MCS), multilayer perceptrons (MLP), the support vector machines (SVM), the decision tree (DT) and K–Nearest Neighborhood (KNN)
Fullpaper: