Penentuan Parameter Tingkat Ke-Fuzzy-an Fuzzy C-Means dan Pengaruhnya Terhadap Proses Algoritma

Abstract

Fuzzy C-Means (FCM) is an algorithm in the process of data clustering that has limitations in the form of being sensitive to the parameters used so that in some cases, the final solution provided is not an optimal solution. One of the influential parameters is the fuzziness level of the algorithm. This parameter is a random real number greater than 1. The determination of these parameters is adjusted to the data used and evaluated with the condition that it reaches a minimum number of iterations for convergence, a small objective final value, and a DBI cluster validity value close to 0. In this study, Indonesian automotive sales data received the optimal algorithm fuzzy level parameter at a value of 2 with other fixed parameters, such as the number of clusters is 3, the smallest error expected to be is 0.00001, and the maximum iteration is 100.