Comparative study of selected clustering algorithm based on Monte Carlo simulation study.

Clustering is a technique that involves partitioning data to group similar data points in a larger dataset with unknown groups. It is very useful in managing large datasets that have no predetermined classification, making it a crucial tool for data analysis. Based on clustering algorithm, the results of clustering or object segmentation depend on the combination of distance and linkage methods used. Therefore, it is important to investigate which clustering method(s) outperform others for different distance measures and linkage methods. A simulation study for 5000 iteration is conducted in evaluating the statistical performance of several distance and linkages methods. The average value of cophenetic correlation coefficient is computed for two different distributions: multivariate normal distribution and gamma distribution using MATLAB programming in measuring the quality of clustering solution. In addition, simulated data are deployed under p = 3 and p = 5, with sample sizes n = 30, 50, and 100. Overall, these results can serve as a helpful guideline for researchers when choosing appropriate distance and linkage methods to make better decisions. [ABSTRACT FROM AUTHOR]