Identifikasi Influential Actors pada Jaringan Transaksi Bitcoin dengan Pendekatan Analisis Graf

Audina, Amadhea Trizza (2026) Identifikasi Influential Actors pada Jaringan Transaksi Bitcoin dengan Pendekatan Analisis Graf. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5025221183-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (6MB) | Request a copy

Abstract

Pertumbuhan pesat mata uang kripto, khususnya Bitcoin, turut diiringi oleh peningkatan penyalahgunaannya untuk aktivitas ilegal seperti pencucian uang, penipuan, dan transaksi darknet market, sehingga investigasi forensik terhadap jaringan transaksi Bitcoin menjadi semakin penting. Transaksi Bitcoin membentuk jaringan hubungan antaralamat yang dapat direpresentasikan sebagai graf, sehingga memungkinkan analisis atribut struktural suatu alamat berdasarkan posisinya dalam jaringan. Penelitian ini bertujuan untuk mengeksplorasi kontribusi atribut struktural berbasis sentralitas dalam mendukung identifikasi influential actors pada jaringan transaksi Bitcoin. Evaluasi dilakukan untuk mengukur efektivitas sentralitas (degree centrality, betweenness centrality, closeness centrality, dan SIIMCO) dalam mengindikasikan keberadaan alamat berlabel illicit berdasarkan posisi strukturalnya pada jaringan, serta menguji kontribusi sentralitas sebagai fitur tambahan pada klasifikasi menggunakan model XGBoost. Penelitian menggunakan dataset Elliptic++ yang direpresentasikan sebagai graf transaksi Bitcoin dengan dua skenario pengujian, yaitu keseluruhan graf dan time step dengan proporsi alamat illicit tertinggi (time step 25). Hasil penelitian menunjukkan bahwa closeness centrality unggul dibandingkan ketiga sentralitas lainnya yang digunakan pada penelitian ini. Pada analisis graf, closeness mencapai recall 8,75% pada keseluruhan graf dan recall 33% dengan precision 75,41% pada time step 25, melampaui kinerja seluruh sentralitas lainnya pada kedua skenario pengujian. Pada implementasi XGBoost, penambahan fitur closeness centrality menghasilkan performa terbaik dengan PR-AUC 0,9205 serta meningkatkan PR-AUC sebesar 1,4% dibandingkan model tanpa fitur sentralitas. Dengan demikian, penelitian ini menunjukkan bahwa pemanfataan atribut struktural berbasis sentralitas mampu memberikan perspektif tambahan dalam mengindikasi keberadaan influential actors atau alamat illicit yang menempati posisi strategis dalam jaringan transaksi Bitcoin, serta memberikan informasi tambahan bagi proses klasifikasi aktivitas illicit menggunakan pendekatan machine learning.
====================================================================================================================================
The rapid growth of cryptocurrency, particularly Bitcoin, has been accompanied by increasing misuse for illicit activities such as money laundering, fraud, and darknet market transactions, underscoring the importance of forensic investigation into Bitcoin transaction networks. Bitcoin transactions form a network of relationships between addresses that can be represented as a graph, enabling the analysis of structural attributes based on each address's position within the network. This study aims to explore the contribution of centrality-based structural attributes in supporting the identification of influential actors within Bitcoin transaction networks. Evaluations are conducted to measure the effectiveness of centrality metrics (degree, betweenness, closeness, and SIIMCO) in indicating the presence of illicit addresses based on their structural positions, as well as to analyze their contribution as additional features in an XGBoost classification model. The Elliptic++ dataset is represented as a Bitcoin transaction graph and examined under two scenarios: the complete graph and the time step with the highest proportion of illicit addresses (time step 25). The results demonstrate that closeness centrality is the most effective metric among those evaluated. In graph-based analysis, closeness centrality achieved a recall of 8.75% on the complete graph and a recall of 33% with a precision of 75.41% on time step 25, outperforming other metrics across both scenarios. In the XGBoost implementation, incorporating closeness centrality as an additional feature achieved the highest performance with a PR-AUC of 0.9205, representing a 1.4% improvement compared to the model without centrality features. These findings indicate that centrality-based structural attributes provide additional perspectives for indicating strategically positioned illicit addresses and influential actors within Bitcoin transaction networks, while also improving illicit activity classification through machine learning approaches.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Analisis Graf, Bitcoin, Influential Actors, Sentralitas, XGBoost, Bitcoin, Centrality, Graph Analysis, Influential Actors, XGBoost
Subjects:	Q Science > QA Mathematics > QA166 Graph theory
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55201-(S1) Undergraduate Thesis
Depositing User:	Amadhea Trizza Audina
Date Deposited:	27 Jul 2026 00:56
Last Modified:	27 Jul 2026 00:56
URI:	http://repository.its.ac.id/id/eprint/137391

Actions (login required)

View Item