Claim Missing Document
Check
Articles

Found 1 Documents
Search
Journal : Journal of Current Research in Blockchain

Analysis of Blockchain Transaction Patterns in the Metaverse Using Clustering Techniques Saputra, Jeffri Prayitno Bangkit; Putri, Nadya Awalia
Journal of Current Research in Blockchain Vol. 1 No. 1 (2024): Regular Issue June
Publisher : Bright Institute

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.47738/jcrb.v1i1.10

Abstract

This study investigates the application of various clustering techniques on a metaverse transaction dataset to identify patterns and groupings. The clustering algorithms evaluated include K-Means, DBSCAN, Gaussian Mixture Model (GMM), Mean Shift, Spectral Clustering, and Birch. The performance of these algorithms is assessed using three metrics: Silhouette Score, Davies-Bouldin Index, and Calinski-Harabasz Index. Among these algorithms, K-Means demonstrated the best overall performance, achieving the highest Silhouette Score (0.4702) and Calinski-Harabasz Index (151946.29), as well as the lowest Davies-Bouldin Index (0.6600), indicating well-defined and compact clusters. DBSCAN, while flexible, showed lower performance with a Silhouette Score of 0.1673, a Davies-Bouldin Index of 1.0084, and a Calinski-Harabasz Index of 4231.19. GMM achieved a Silhouette Score of 0.2453, a Davies-Bouldin Index of 1.3626, and a Calinski-Harabasz Index of 23011.20. Spectral Clustering had a Silhouette Score of 0.1668, a Davies-Bouldin Index of 2.0986, and a Calinski-Harabasz Index of 11830.24. Birch achieved a Silhouette Score of 0.2363, a Davies-Bouldin Index of 1.4967, and a Calinski-Harabasz Index of 21375.76. Mean Shift could not provide valid performance metrics. Visualizations, including histograms, box plots, and count plots, provided additional insights into the distribution of numerical features and cluster characteristics. This study highlights the need for tailored clustering approaches and suggests future research directions in hybrid models as well as the impact of feature selection and scaling methods on clustering outcomes.