Implementasi Reduksi Fitur t-SNE Pada Clustering Gambar Head shape Nematoda

Abstract

In this research, clustering of nematode head shape images is carried out. In processing the picture, a feature extraction method is needed to find important information from the image to be processed. One of the feature extraction that can be used is the wavelet. After the image goes through feature extraction, 5624 features are generated; many features can result in a long computation time. Therefore, it is necessary to make feature reduction to reduce the number of features from 5624 to only 2 or 3 elements, one of the newest feature reduction methods that can be used is t-SNE. In this study, a comparison of the results of cluster quality between those using feature reduction and those not using feature reduction was carried out. Silhouette Index results obtained without feature reduction is 0.046, and after using the t-SNE feature reduction, there is a significant increase to 0.418. Keywords: Clustering; Extraction Features; Reduction Features; t-SNE; Wavelet