TWO LEVEL CLUSTERING UNTUK PENINGKATAN KUALITAS HASIL CLUSTERING MENGGUNAKAN FUZZY SUBTRACTIVE CLUSTERING DAN SELF-ORGANIZING MAP; TWO LEVEL CLUSTERING FOR QUALITY IMPROVEMENT USING FUZZY SUBTRACTIVE CLUSTERING AND SELF-ORGANIZING MAP

Lisangan, Erick Alfons, Sri Hartati

2015 | Disertasi | FMIPA

Clustering is one of the pattern recognition technique and has been used in various fields. Recently, clustering algorithms combined statistical methods and artificial intelligence. Self-Organizing Map (SOM) is clustering algorithm that apply the concept of neural network. The disadvantages of SOM is the structure of neural network and number of neurons or clusters must defined first. Moreover, the initial value of neuron weights is randomized so produce different clustering results. FSC-SOM is designed to solve the disadvantages of SOM, such as define the number of clusters and initial value of neuron weights. Fuzzy Subtractive Clustering (FSC) determine the cluster centers based on the density of data. FSC is implemented to find the number of clusters and the cluster centers which become the parameter of SOM. FSC-SOM is expected to improve the quality of FSC because the determination of the cluster centers are processed twice i.e. searching for data with high density at FSC then updating the cluster centers of FSC at SOM. FSC-SOM was tested using 10 datasets that measure clustering results using the cluster validation indexes, i.e. F-Measure, entropy, Silhouette Index, and Dunn Index. The quality of clustering results from FSC, SOM, and FSC-SOM is compared each other. The result of this research showed that FSC-SOM can improve the cluster center of FSC with SOM in order to obtain the better quality of clustering results, both external and internal validity. The quality of clustering results by FSC-SOM at least equal to the quality of clustering results by FSC on all datasets.

Kata Kunci : Clustering; Fuzzy Subtractive Clustering; Self-Organizing Map