Ταξινόμηση διευθύνσεων Bitcoin χρησιμοποιώντας μη εποπτευόμενη μηχανική εκμάθηση
Bitcoin address classification using unsupervised machine learning
View/
Keywords
Bitcoin ; Unsupervised machine learning ; Logistic regression classifier ; StellarGraph ; Blockchain ; Graph classificationAbstract
Bitcoin is a decentralized digital cryptocurrency, introduced in 2008 by Satoshi Nakamoto, providing pseudonymity to its users. Since Bitcoin blockchain data is publicly available, transactions can be modeled to a directed graph, for further analysis. This dissertation presents a novel approach to reduce the anonymity provided, by using Unsupervised Machine Learning on the transactions graph. By using node representation learning, node features can be extracted and used by a Logistic Regression Classifier to predict the label of each graph node. To simplify data access, blockchain data was imported to a MySQL Database. Performance of the complete proposed solution was evaluated, by executing the classifier on a sub-set of the blockchain data, achieving a maximum accuracy of 76.39%.
Postgraduate Studies Programme
Προηγμένα Συστήματα ΠληροφορικήςDepartment
Σχολή Τεχνολογιών Πληροφορικής και Επικοινωνιών. Τμήμα ΠληροφορικήςNumber of pages
42Language
EnglishCollections
Except where otherwise noted, this item's license is described as
Αναφορά Δημιουργού-Μη Εμπορική Χρήση-Όχι Παράγωγα Έργα 3.0 Ελλάδα
Αναφορά Δημιουργού-Μη Εμπορική Χρήση-Όχι Παράγωγα Έργα 3.0 Ελλάδα