Performance Analysis of a Micromodel-Based Multinomial Classifier

Supervised learning methods demand large scores of labelled training data to achieve high predictive accuracy for classification problems. In contrast, the less accurate unsupervised learning models can work on raw data but fail to provide any useful insight. Micromodel-based multinomial classifier is a one-stop solution for increased performance with less data and less time. This paper investigates the effectiveness of micromodels for multiclass classification as against the existing supervised and unsupervised machine learning models and artificial neural networks. The model built using One-class Neural Network (OC-NN) to classify students' resumes into four different streams for admission into postgraduate programmes in top-ranked universities, outperforms some of the most powerful machine learning models.