TY  - BOOK
AU  - Comendador, Christine Y.
TI  - Monothetic clustering algorithms based on measures of association / 
PY  - 2010///
KW  - Association measures
KW  - Binary data
KW  - Clustering
KW  - Monothetic analysis
KW  - Undergraduate Thesis
KW  - AMAT200
N1  - Thesis, Undergraduate (BS in Applied Mathematics -  Operations Research) -- U.P. Mindanao
N2  - Cluster analysis is one common technique that is used in multivariate analysis. Its objective is to maximize the dissimilarity of objects within the sane cluster and maximize the dissimilarity of object among different clusters. Monothetic analysis is a class of methods used for clustering binary data. Monothetic analysis algorithms produce a hierarchy of clusters in which each step a group is split into two clusters based on the values of one of the binary variables. However, existing monothetic algorithms use the simplest association measure. This study was done to test whether Yule?s q, Yule's y, Phi coefficient and percent difference could be used as a splitting criterion. The algorithm was modified in such a way that variables having identical values for all observations were ignored to avoid undefined values for the association between variables and by replacing the standard criterion with the selected association measures. The modified algorithms were tested with twenty simulated binary data resulted in identifying the same number of cluster. Nevertheless, the algorithms varied in the selection of splitting variable and in effect, results to different number of iterations performed. Yule?s y ? based MONA performs the fastest algorithm among the modified algorithms and Phi coefficient ? based MONA performs the slowest algorithm in terms of the number of iterations. However, percentage difference ?based MONA yields the fastest algorithm compared to all other algorithms in terms of processing time. Aside from the simulated data sets, an application to the habitat of 16 species to 129 communities was also used in this study. As a result, it confirms the fact that Percent difference ?based MONA performs the fastest algorithm compared to all algorithms in terms of processing time. Also, it confirms the fact that Phi coefficient ?based MONA performs slowest algorithm. Despite the fact that Percent difference ?based MONA is the fastest algorithm on both application to data sets, further study are still needed to verify the performance of these modified algorithms
ER  -