N-Centroid Classifier vs. Machine Learning Models for Data Classification

Due to their universal applicability, machine learning models (ML) have been a hot topic during the last two decades. Despite their effectiveness, some ML models exhibit inefficiency, particularly in big data classification. Moreover, some ML models have been seen ineffective on some small datasets. In this regard, due to the growing accessibility of online data, the automatic data classification technique has attracted a lot of study interest. As a result, numerous unique learning strategies have been developed in the text categorization field. The Centroid-Based Classifier (CBC) is one of these most extensively used technique among them. While focusing on enhancing NC classifier, this paper, therefore, aims to briefly investigate the impact of some ML models on small and medium-sized dataset’s classification. Among these models: N-Centroid technique (NC) as a simply-designed classifier, Support Vector Machine (SVM), and Multinomial Bayesian (MNB). Most importantly, this paper introduces an enhanced variation of NC via its integration with two similarity measures, namely, Set Theory Based Similarity Measure (STB-SM) and Improved Cosine Similarity Measure (ISC). The performance of integrated NC classifiers has been seen as promising in terms of both effectiveness and efficiency.