MIXED VARIABLES CLASSIFICATION: A COMPREHENSIVE REVIEW

Abstract

Classification involving mixed categorical and continuous variables presents unique challenges due to the differences in data scales and statistical properties. These types of data are commonly encountered across various research fields. In this study, we introduce a systematic review of classification methods designed to handle mixed-type variables. A total of 52 journal articles were selected through a systematic review process based on the related topics, which were identified and screened from the Web of Science (WoS) as well as Scopus databases. The selected articles were reviewed and categorised into two main groups of classification methods: parametric and non-parametric approaches. Among parametric approaches, methods based on the location model were among the most frequently used. In contrast, non-parametric approaches, such as classification trees and machine learning methods, were often employed when distributional assumptions were not met. This review also presents the use of classification methods for mixed-type variables across diverse fields such as medical, psychology, agriculture, and biology. Therefore, this review may serve as a useful reference for researchers working with mixed-type data and emphasises the importance of ongoing methodological development to address the complexity of such data in real-world applications.