[
ABSTRAKPenelitian ini mengusulkan tiga algoritma meta-heuristik berbasis Fuzzy K-modes
untuk clustering binary data set. Ada tiga metode metaheuristik diterapkan, yaitu
Particle Swarm Optimization (PSO), Genetika Algoritma (GA), dan Artificial Bee
Colony (ABC). Ketiga algoritma digabungkan dengan algoritma K-modes.
Tujuannya adalah untuk memberikan modes awal yang lebih baik untuk K-modes.
Jarak antara data ke modes dihitung dengan menggunakan koefisien Jaccard.
Koefisien Jaccard diterapkan karena dataset mengandung banyak nilai nol . Dalam
rangka untuk melakukan pengelompokan set data real tentang supplier otomotif di
Taiwan, algoritma yang diusulkan diverifikasi menggunakan benchmark set data.
Hasil penelitian menunjukkan bahwa PSO K-modes dan GA K-modes lebih baik
dari ABC K-modes. Selain itu, dari hasil studi kasus, GA K-modes memberikan
SSE terkecil dan juga memiliki waktu komputasi lebih cepat dari PSO K-modes
dan ABC K-modes.
ABSTRACTThis study proposed three meta-heuristic based fuzzy K-modes algorithms for
clustering binary dataset. There are three meta-heuristic methods applied, namely
Particle Swarm Optimization (PSO) algorithm, Genetic Algorithm (GA) algorithm,
and Artificial Bee Colony (ABC) algorithm. These three algorithms are combined
with k-modes algorithm. Their aim is to give better initial modes for the k-modes.
Herein, the similarity between two instances is calculated using jaccard coefficient.
The Jaccard coefficient is applied since the dataset contains many zero values. In
order to cluster a real data set about automobile suppliers in Taiwan, the proposed
algorithms are verified using benchmark data set. The experiments results show
that PSO K-modes and GA K-modes is better than ABC K-modes. Moreover,
from case study results, GA fuzzy K-modes gives the smallest SSE and also has
faster computational time than PSO fuzzy K-modes and ABC fuzzy K-modes., This study proposed three meta-heuristic based fuzzy K-modes algorithms for
clustering binary dataset. There are three meta-heuristic methods applied, namely
Particle Swarm Optimization (PSO) algorithm, Genetic Algorithm (GA) algorithm,
and Artificial Bee Colony (ABC) algorithm. These three algorithms are combined
with k-modes algorithm. Their aim is to give better initial modes for the k-modes.
Herein, the similarity between two instances is calculated using jaccard coefficient.
The Jaccard coefficient is applied since the dataset contains many zero values. In
order to cluster a real data set about automobile suppliers in Taiwan, the proposed
algorithms are verified using benchmark data set. The experiments results show
that PSO K-modes and GA K-modes is better than ABC K-modes. Moreover,
from case study results, GA fuzzy K-modes gives the smallest SSE and also has
faster computational time than PSO fuzzy K-modes and ABC fuzzy K-modes.]
T44406-Yuliana Portti.pdf :: Unduh