Cluster Analysis in Machine Learning using Progressive Latin Hypercube Sampling

This study adds to the recent research on classification and regression using Latin Hypercube Sampling (LHS) as well as on classification involving Progressive LHS (PLHS) by investigating the application of PLHS in unsupervised machine learning. PLHS is a sampling method where the sample is constructed by sequential accumulation of matrices on an initial Latin Hypercube Design (LHD) matrix to consecutively build up LHD matrices, thus securing the maximization of space filling. The objective is to test the efficacy of this sampling technique, when it is implemented in clustering problems, by comparison with a benchmark sampling method, like random sampling. The outcome indicates the potential of this technique, as the results tend to be in favor of PLHS against RS.