Klustering Dengan K-Means Berbasis LVQ Dan K-Means Berbasis OWA

Penulis

Dian Eka Ratnawati, Indriati .

Abstrak

Abstrak

Pada penelitian ini dilakukan pembandingan hasil klustering pada data car evaluation dengan menggunakan K-Means berbasis LVQ (Learning Vector Quantization) dan K-Means berbasis OWA (Ordered Weighted Averaging). Pada kedua metode ini sama-sama mempergunakan K-Means tetapi yang sudah mengalami modifikasi.

Hasil dari penelitian sebelumnya secara terpisah yang membandingkan metode K-Means modifikasi tersebut dengan K-Means konvensional menunjukkan bahwa kedua metode modifikasi tersebut sama-sama lebih baik daripada K-Means konvensional. Tetapi belum pernah ada penelitian yang membandingkan akurasi hasil klustering kedua metode modifikasi tersebut. Sehingga pada penelitian ini dilakukan klustering dengan menggunakan kedua metode tersebut untuk data car evaluation, karena dari penelitian sebelumnya kedua metode tersebut cukup handal dalam melakukan klustering.  Hasil dari ujicoba menunjukkan rata-rata hasil akurasi dimulai yang tertinggi adalah K-Means berbasis LVQ(86.50%), K-Means berbasis OWA(86,16%) kemudian K-Means konvensional (56,50%). Tetapi dengan urutan atribut yang benar dan pemilihan nilai alpha yang tepat yakni 0.8, K-Means berbasis OWA bisa menghasilkan akurasi yang lebih tinggi yakni 93.33%.

 

Kata kunci: K-Means berbasis LVQ, K-Means, K-Means berbasis OWA, bobot


Abstract

In this paper do a comparison with the results of klustering using K-Means based LVQ (Learning Vector Quantization) and K-Means based OWA (Ordered Weighted Averaging). In both of these methods used K-Means but which has been modified.

Results from previous studies have shown that both methods are better than conventional K-Means. But there has never been a study comparing the accuracy of klustering results of the two methods. So in this study conducted klustering using both methods for data car evaluation, because of previous studies both methods are reliable enough to perform klustering In the researchs before it, both method are prefer than conventionalK-Means, but there are no researchs which compare them. So, in the research , we will compare it by using same data that is  car evaluation. In order to know what it is method is the best. The result of research are that in the average , K-Means LVQ(86.50%) is more accuracy than K-Means – OWA(86,16%) and conventional K-Means(56,50). But if the order of selection attributes and alpha values is correct ​​, K Means based OWA can generate higher accuracy that is 93.33 % using the alpha value of 0.8

Keywords: K-Means based LVQ, K-Means, K-Means based OWA, weight

Teks Lengkap:

PDF (English)

Referensi


AHN BS. 2006. On The Properties of OWA Operator Weights Functions with Constant Level of Orness, IEEE Transactions on Fuzzy Systems (4),511–515.

CHENG, CHING-HUSE, JING-WEI LIU, MING-CHANG WU,2007. Owa Based Information Fusion Techniques For Classification Problem, Proceedings of the Sixth International Conference on Machine Learning and Cybernetics, Hong Kong, 19-22 August 2007, -4244-0973-X/07/$25.00 ©2007 IEEE.

CHENG CH, JW WANG, DAN MC WU. 2009. OWA-Weighted Based Klustering Method for Classification Problem, Expert Systems with Application4988-4995.

FAYYAD, USAMA., dkk., 1996. From Data Mining to Knowledge Discovery in Databases. American Association for Artificial Intelligence.

FAUSETT, L., 1994, Fundamentals of NeuralNetwork, Architecture, Algorithms and Applications, Prentice Hall, New Jersey

KANTARDZIC, MEHMED. 2003. Data Mining : Concepts, Models, Methods and Algorithm. John Wiley & Sons. New York.

RATNAWATI, DIAN, MARJI, LAILIL M. 2012. Pengembangan Metode Klasifikasi berdasarkan K-Means dan LVQ, Penelitian DIPA 2012.

RATNAWATI, DIAN, MARJI, 2014. Klustering dengan K-Means dan K-Means modifikasi , proceeding KNSI 2014.

ULYA, MILYATUL, 2011. Modifikasi K-Means Berbasis Ordered Weighted Averaging (Owa) Untuk Kasus Klastering, AGROINTEK Volume 5, No.2 Agustus 2011




DOI: http://dx.doi.org/10.25126/jtiik.201521125