Thesis (BS Applied Mathematics) -- University of the Philippines Mindanao, 2005

Clustering of the morphological characteristics of sago palm was performed the data used were the morphological attributes of 72 samples gathered in Agusan del Sur by Uy (2004). The data contained ten morphological attributes of sago palm in which seven are numerical and three are categorical. This study explored the potential of the k-prototypes algorithm proposed by Huang (1998) to cluster the morphological sago palm characteristics, which are a mixed type of data. The data were subdivided into three group stages namely, early bole, late bole and inflorescence. Then samples within each group stage were subjected to clustering. There were different numbers of clusters obtained in each group stage. Results showed that some clusters consisted of either small or large sago palms, meaning the numerical attributes have effect on clustering. In addition, categorical attributes were also affecting the clusters formed. Specifically, the members in each cluster have only one type of suckering pattern (type 1, type 2) and grow in one soil condition (waterlogged, flooded). The result of this study indicated that varieties of sago palm could not be determined only by combining the categorical attributes since both numerical and categorical attributes of sago palm differ. This study also indicated that k-prototypes algorithm is capable of clustering the morphological characteristics of sago palm.