Comparison of dimensionality reduction and clustering methods for SARS-CoV-2 genome

This paper aims to conduct an analysis of the SARS-CoV-2 genome variation was carried out by comparing results clustering using several algorithms and distribution sequence in each cluster. The used are K-means, Gaussian mixture models, agglomerative hierarchical clustering, mean-shift DBSCAN. However, algorithm has a weakness grouping data that very high dimensions such as data, so dimensional reduction process is needed. In this research, dimensionality principal component (PCA) autoencoder method with three models produce 2, 10, 50 features. main contributions achieved were scheme performance experiment on hyper parameters for method. Based experiments conducted, PCA DBSCAN achieve highest silhouette score 0.8770 clusters when two need more iterations converge. On testing Indonesian than half them enter one cluster rest distributed other clusters.