Yati Sri Hayati
(2008)
PERKIRAAN JUMLAH KELOMPOK PADA ANALISIS KELOMPOK METODE CENTROID DAN WARD DENGAN MENGGUNAKAN GAP STATISTIC.
Thesis thesis, UNIVERSITAS AIRLANGGA.
Abstract
Cluster analysis is one of multivariate data analysis techniques which has purpose to cluster many objects into one cluster which has the more similarity among its objects than with other objects in another cluster. One of the most challenging problems in cluster analysis is choosing the optimal number of clusters in a dataset, mainly if the researcher less understands about the characteristic of group of data. Tibshirani,Walther and Hastie (2001) proposed The Gap Statistic method for estimating the number of clusters. Index of wealthy family is one of classifications which can be processed with cluster analysis technique. According to Soenarnatalina (2006), indicators of a wealthy family were: (1) health, (2) education, (3) housing and sanitation, (4) social and culture, and (5) economics. The purpose of this research was to: (1) estimate the number of optimum clusters in centroid clustering method using the gap statistic, (2) estimate the number of optimum clusters in ward clustering method using the gap statistic, (3) compare the accuracy level of group number estimation using statistic gap method (to centroid and ward clustering methods). This research used secondary data from the results of the research done by Soenarnalita (2006) titled Index development of wealthy family in East Java. The result of this research showed that there were five optimum clusters when using statistic gap in centroid clustering method and five optimum cluster when using statistic gap in ward clustering method. It can be concluded that statistic gap estimation method can be used both in centroid and ward clustering data because it gives the same result. Based on this result, it is suggested to compare the group number estimation using statistic gap method with other estimation methods for further research.
Actions (login required)
 |
View Item |
