Jurnal Publikasi STMIK Pontianak

Data Mining Untuk Dataset Multidimensional Penggunaan Algoritma Forgy, Isodata dan K-Means

Abstrak

At this time, information can be obtained from a pile of very large size of data, hence the need to data mining activities that are still hidden to further processed into useful knowledge in decision making. The process of extraction of information from the collection of data that is stored with the data mining. One of the methods in Data Mining is clustering. Clustering aims to segment the physical or abstract objects in the form of classes or objects similar. Some partitioning clustering are Forgy, ISODATA and K-Means algorithms. On this research analyzed the influence of the changes made on application of the level of accuracy and the resulting clusters, and system parameters that apply. For the purpose of software is built as a media test Forgy, ISODATA and K-Means algorithms. Level of accuracy can be improved by using threshold, maximum number of iterations and increase the value of the parameter. Comparison is made the test of clustering time process, the variance ratio and the value of Sum Squared Error. Experiments are carried out on the number of datasets. This research concludes that the three clustering methods has similar value of variance ratio, and high intraclass similarity and low interclass similarity. The clustering algorithm using K-Means requires the longest time.

Keywords : Clustering, Forgy, ISODATA, K-Means, Cluster Analysis