Multiclass Classification Methods: A Review

Abstract

Artificial intelligence is playing vital role in various domains presently. Multiclass classification is an important task of AI. The concept of multiclass classification has been adopted in various domains such as medical science, banking sector, corporate, cyber security etc. Still researchers are applying such concept in different fields and trying to enhance the accuracy and efficiency of the  models.  The efficiency and accuracy of models are depend on various factors of such as dataset, selection of features, selection of algorithms, selection of hyper-parameters. The main objective of this paper is to present the various machine learning algorithms that are used for multiclass classification besides their merits and demerits. The present study also states the various research gaps in the existing multiclass classification models that are need to be resolved. This study will be useful to the researchers who are applying the multiclass classification in a specific domain to classify the data with great accuracy and efficiently.  [ The efficacy of models used for classification, recognition, diagnosis, or clustering of data of different domains is determined by their performance in real-world applications. Evaluating such  models require detail understanding of their underlying equations and features to discern whether they perform "well" or "poorly." Such methods have become increasingly important to researchers in recent years. While a wide range of statistical methods has been applied, there remains a gap in guiding researchers toward the most appropriate method for their specific applications. This paper collects and analyzes the most significant classification methods used in research. It provides a comparative analysis of these methods, focusing on their mathematical foundations, purposes, and limitations. The study highlights that feature selection plays a significant role in classification performance, and the results provide guidance on selecting features for classification, recognition, diagnosis, or clustering tasks.]