Clustering Data Fitur Wajah Karakter Wanita Pilihan Konsumen Menggunakan Algoritma K-Means Dan Som

Megawati, Citra Dewi (2019) Clustering Data Fitur Wajah Karakter Wanita Pilihan Konsumen Menggunakan Algoritma K-Means Dan Som. Masters thesis, Institut teknologi sepuluh november surabaya.

Text 07111750050008_Master_Thesis.pdf Restricted to Repository staff only Download (5MB)

Abstract

Permainan saat ini sangat diminati masyarakat. Karakter merupakan salah satu komponen yang penting di dalam suatu permainan. Namun illustrator sulit menentukan bagaimana wajah karakter yang sering dipilih konsumen. Penelitian ini mengenai pengelompokkan data fitur wajah karakter wanita yang sering dipilih oleh konsumen. Data fitur wajah yang dimaksudkan seperti bentuk wajah, alis, mata, hidung, bibir, telinga dan warna kulit. Penelitian ini menggunakan 2 metode algoritma clustering yaitu yang pertama Kmeans clustering dan Self organizing map clustering. Di dalam penelitian ini kedua metode tersebut di partisi menjadi 6 cluster. Hasil dari metode KMeans sebanyak 23% memilih wajah oval dan bentuk hati. Berdasarkan seluruh fitur wajah pada hasil wajah yang sering dipilih oleh konsumen wajah termasuk kedalam wajah simetri karena kombinasi fitur wajah mulai dari mata dan alis seimbang dari kiri kekanan begitu pula dengan bentuk hidung dan bibir. Hasil peneltian dengan metode SOM 24% konsumen memilih wajah diamond dengan berbagai fitur wajah yang melengkapinya, namun golongan wajah tersebut masuk kedalam wajah asimetri karena salah satu fiturnya yaitu hidung bentuknya tidak simetri. Dari hasil penelitian clustering data fitur wajah wanita tersebut sama seperti hasil beberapa penelitian mengenai kecantikan wajah wanita yakni kecantikan bentuk wajah umumnya berbentuk wajah oval. Hasil dari validasi cluster menggunakan DBI clustering menggunakan K-Means lebih optimal dibandingkan metode self organizing maps karena nilai DBI pada metode KMeans lebih rendah dibandingkan metode SOM, lalu dari segi proses pengerjaan juga metode KMeans lebih cepat dibandingkan metode SOM.
===================================================================================================================================
Games have become increasingly popular, and character design is one of the most important elements influencing player preference. However, illustrators often face difficulties in determining which facial characteristics are most appealing to consumers. This study aims to cluster the facial features of female game characters that are most frequently preferred by consumers. The analyzed facial attributes include face shape, eyebrows, eyes, nose, lips, ears, and skin color. Two clustering algorithms, namely K-Means and Self-Organizing Map (SOM), were employed, with both methods partitioning the data into six clusters. The K-Means results showed that 23% of consumers preferred characters with oval and heart-shaped faces. Based on the combination of facial attributes, these faces were generally categorized as symmetrical, with balanced eyebrows and eyes as well as proportionate nose and lip shapes. In contrast, the SOM method indicated that 24% of consumers preferred diamond-shaped faces with diverse facial characteristics. However, this cluster was categorized as asymmetrical because one of its prominent features, particularly the nose shape, lacked facial symmetry. Overall, the clustering results are consistent with previous studies on facial attractiveness, which commonly identify the oval face shape as the most attractive. Cluster validation using the Davies–Bouldin Index (DBI) demonstrated that K-Means outperformed the Self-Organizing Map method, as it produced a lower DBI value, indicating better cluster quality. Furthermore, K-Means required less computational time than SOM, making it both more effective and more efficient for clustering female facial feature data.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	algorithm, Clustering, character, Kmeans, Self organizing maps
Subjects:	T Technology > T Technology (General) > T57.5 Data Processing
Divisions:	Faculty of Electrical Technology > Electrical Engineering > 20101-(S2) Master Thesis
Depositing User:	Citra Dewi Megawati
Date Deposited:	05 Aug 2026 08:48
Last Modified:	05 Aug 2026 08:48
URI:	http://repository.its.ac.id/id/eprint/65081

Actions (login required)

View Item