Deteksi Manipulasi Pasar Non-fungible Token (NFT) Menggunakan Graph Neural Network

Ade, Indriawan (2023) Deteksi Manipulasi Pasar Non-fungible Token (NFT) Menggunakan Graph Neural Network. Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 6026211015-Master_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 April 2025. Download (3MB) | Request a copy

Abstract

Adopsi aset digital berbasis blockchain seperti cryptocurrency dan non-fungible token (NFT) semakin meluas di berbagai kalangan masyarakat. Namun, tingginya antusiasme tersebut tidak diiringi oleh literasi keuangan yang memadai, khususnya mengenai risiko investasi di kedua jenis aset digital tersebut. Hal ini membuka peluang terjadinya fraud dan scams di sektor tersebut. Penelitian ini mengusulkan sebuah pendekatan untuk mendeteksi manipulasi di pasar NFT berdasarkan data histori transaksi NFT dari platform Kaggle yang diekstrak dari jaringan blockchain Ethereum menggunakan Graph Neural Network. Sebuah representasi graph dari dataset tersebut dibangun untuk proses pelabelan data menggunakan algoritma Depth-First Search (DFS) dan mengekstrak fitur-fitur yang relevan menggunakan beberapa algoritma graph lainnya seperti PageRank, Degree Centrality, dan Greedy Modularity Communities. Kemudian dataset tersebut dimodelkan ke dalam sebuah graph yang heterogen untuk menjadi input beberapa model klasifikasi yang dibangun, yang berbasis Multilayer Perceptron (MLP), Graph Convolutional Neural Network (GCN), dan Heterogeneous Graph Convolutional Neural Network (HeteroGCN). Hasil pengujian memperlihatkan bahwa model klasifikasi berbasis HeteroGCN memiliki performa yang lebih baik dibanding model yang lain. Sementara itu, model-model klasifikasi yang mengambil input dari data yang direpresentasikan dalam bentuk graph secara umum memiliki kinerja yang lebih baik daripada model klasifikasi yang mengambil input dari data yang berformat tabular yang menggunakan algoritma Random Forest, Decision Tree, dan k-Nearest Neighbor. Selanjutnya, perbandingan beberapa skenario fitur menunjukkan bahwa algoritma PageRank dan Degree Centrality memiliki peran yang signifikan untuk mengklasifikasikan transaksi yang terindikasi sebagai bagian dari manipulasi pasar.
================================================================================================================================
The use of blockchain-based digital assets such as cryptocurrencies and non-fungible tokens (NFTs) is becoming more common in various segments of society. However, the high level of enthusiasm is not matched by dequate financial literacy, particularly in terms of investment risks in both types of digital assets. This creates new opportunities for fraud and scams in the industry. The study proposes a method for detecting manipulation in the NFT marketplace using NFT transaction history data from the Kaggle platform extracted from the Ethereum blockchain network and Graph Neural Network. A graph representation of the dataset is built for the process of labeling data using the Depth-First Search (DFS) algorithm and extracting relevant features using several other graph algorithms such as PageRank, Degree Centrality, and Greedy Modularity Communities. The dataset is then modeled as a heterogeneous graph to serve as the input for several classification models developed, including Multilayer Perceptron (MLP), Graph Convolutional Neural Network (GCN), and Heterogeneous Graph Convolutional Neural Network (HGCN) (HeteroGCN). The test results show that the HeteroGCN-based classification model performs better than other models. Meanwhile, classification models that take input from data represented as graphs generally perform better than classification models that take input from tabular formatted data which use Random Forest, Decision Tree, and k-Nearest Neighbor algorithms. Furthermore, a comparison of several feature scenarios reveals that the PageRank and Degree Centrality algorithms play an important role in classifying transactions that are suspected of being part of market manipulation.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	fraud detection, features extraction, graph neural network, manipulasi pasar, NFT
Subjects:	T Technology > T Technology (General) > T57.5 Data Processing T Technology > T Technology (General) > T58.5 Information technology. IT--Auditing T Technology > T Technology (General) > T58.6 Management information systems T Technology > T Technology (General) > T58.62 Decision support systems
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Information System > 59101-(S2) Master Thesis
Depositing User:	Ade Indriawan
Date Deposited:	15 Feb 2023 02:35
Last Modified:	20 Feb 2023 04:07
URI:	http://repository.its.ac.id/id/eprint/96810

Actions (login required)

View Item