Clustering of Female Avatar Face Features Consumers Choice using KMeans and SOM Algorithm

Abstract

Games are interactive activities that is popular with consumers. However, illustrators are unable to determine of characters face many chosen with consumers. The paper proposes a clustering of face features often chosen by consumers for their female avatars. The face features involved in this research are the face shape, eyebrows, eyes, nose, lips, ears and skin color. The paper uses two methods for clustering, KMeans and SOM. These methods used class partitioning based on shape similarities. Using the K-Means results, 14% chose a triangular face, then 16% chose an diamond face shape, 23% chosen a heart face shape, 6% chosen a round face shape, 23% chosen a oval face shape, 18% chosen a square face shape. Results from SOM shows that 18% chose a triangular face, then 11% chose an diamond face shape, 24% chosen a heart face shape, 22% chosen a round face shape, 12% chosen a oval face shape, 13% chosen a square face shape. It shows that KMeans has better performance than SOM in determining the female avatar face classes that are wanted by customers.