Partitioning Clustering algorithms for handling numerical and categorical data: a review

TL;DR

This paper reviews various partitioning clustering algorithms designed to handle datasets with both numerical and categorical data, highlighting their approaches and applications in real-world data mining tasks.

Contribution

It provides a comprehensive review of existing partitioning clustering algorithms for mixed data, categorizing their methods and discussing their effectiveness.

Findings

Algorithms effectively handle mixed data types in clustering.

Different approaches have unique advantages depending on data characteristics.

The review highlights gaps and future directions in mixed data clustering.

Abstract

Clustering is widely used in different field such as biology, psychology, and economics. Most traditional clustering algorithms are limited to handling datasets that contain either numeric or categorical attributes. However, datasets with mixed types of attributes are common in real life data mining applications. In this paper, we review partitioning based algorithm such as K-prototype, Extension of K-prototype, K-histogram, Fuzzy approaches, genetic approaches, etc. These algorithm works on both numerical and categorical data. The approaches has been proposed to handle mixed data are based on four different perceptive: i) split data set into two part such that each part contain either numerical or categorical data, then apply separate clustering algorithm on each data set, finally combined the result of both clustering algorithm, ii) converting categorical attribute into numerical attribute and apply numerical attribute clustering algorithm; iii) discrimination of numerical attribute and apply categorical based clustering algorithm; iv) Conversion of the categorical attributes into binary ones and apply any numerical based clustering algorithm