Principal Component Analysis (PCA)

robust PCA

Tukey’s biweight correlation

Pearson Correlation

K-means Cluster Analysis

Prevention

Environmental management

In this research work, the pattern of spatial cluster had been identified for torrential rainfall data within the context of Peninsular Malaysia, which experiences heavy pour annually. Hence, a robust Principal Component Analysis (PCA) technique was employed in this study in order to address problem related to non-balance cluster(s) across patterns of rainfall stemming from skewed rainfall data. To analyze the observations made, Tukey’s biweight correlation was applied. For PCA components extraction, the optimum breakdown point was determined based on the proposed method. In order to strike a balance for extraction of number of components, as well as to hinder insignificant spatial scale or low-frequency variation, the simulation data recorded a breakdown point at 70% cumulative percentage of variance. The study outcomes revealed that the robust PCA gave better enhancement than the Pearson-based PCA did for cluster average number and quality. The findings indicate that ten rainfall patterns obtained are quite definite and clearly display the dominant role extended by the complex topography and exchange monsoons of the peninsular.