Evaluating different clustering procedures using Philippine sardines data

Date of Publication

8-2006

Document Type

Bachelor's Thesis

Degree Name

Bachelor of Science in Statistics Major in Actuarial Science

Subject Categories

Statistics and Probability

College

College of Science

Department/Unit

Mathematics and Statistics

Thesis Adviser

Shirlee R. Ocampo

Defense Panel Chair

Imelda E. De Mesa

Defense Panel Member

Regina M. Tresvalles
Francis H. Campeña

Abstract/Summary

The thesis delves into clustering Philippine sardines inferred from various clustering procedures using morphometric and meristic characteristics of Sardinella species. A total of 199 Philippine sardines comprising of Agoncillo and Talisay both representing Sardinella tawilis, Sardinella albella, Sardinella brachysoma, Sardinella longiceps and Sardinella lemuru were collected and a total of 35 biometric variables (3 meristic and 32 morphometric) were measured from these five species. In addition, clustering of these sardines was also performed after reducing the data by principal component analysis. Furthermore, the study explored the problem of identifying the relationship of S. tawilis which is the only Philippine sardine known to live in freshwater among the other Philippine sardines that thrive in marine water. Six hierarchical clustering procedures such as single linkage, complete linkage, centroid, average, and Ward and one nonhierarchical procedure were explored together with three distance measures. All procedures indicate that Talisay and Agoncillo species always belong to one cluster representing S. tawilis. The results obtained from most clustering algorithms indicate that the closest relative of S. tawilis is S. albella, followed by S. brachysoma, S. longiceps and S. lemuru as the most distant. Furthermore, clustering results after employing principal component analysis suggest that 35 variables can be reduced to 12 representative variables. This implies that the 12 representative variables would be sufficient to use in cluster analysis instead of the original 35 variables. This is helpful result because collection of data in studying taxonomy of Philippine sardines will be less costly and less time consuming.

Abstract Format

html

Language

English

Format

Print

Accession Number

TU13522; CDTU013522

Shelf Location

Archives, The Learning Commons, 12F, Henry Sy Sr. Hall

Physical Description

ix, 121, [50] leaves : ill. (some col.) ; 28 cm. + 1 computer optical disc.

Keywords

Sardines--Philippines; Cluster analysis

Upload Full Text

wf_no

Embargo Period

4-2-2025

This document is currently not available here.

Share

COinS