BITCOIN TRANSACTION FORECASTING WITH DEEP.NETWORK REPRESENTATION LEARNING
ABSTARCT :
Bitcoin and its decentralized computing paradigm for digital currency trading are one of the most disruptive technology in the 21st century. This paper presents a novel approach to developing a Bitcoin transaction forecast model, DLForecast, by leveraging deep neural networks for learning Bitcoin transaction network representations. DLForecast makes three original contributions. First, we explore three interesting properties between Bitcoin transaction accounts: topological connectivity pattern of Bitcoin accounts, transaction amount pattern, and transaction dynamics. Second, we construct a time-decaying reachability graph and a time-decaying transaction pattern graph, aiming at capturing different types of spatial-temporal Bitcoin transaction patterns. Third, we employ node embedding on both graphs and develop a Bitcoin transaction forecasting system between user accounts based on historical transactions with built-in time-decaying factor. To maintain an effective transaction forecasting performance, we leverage the multiplicative model update (MMU) ensemble to combine prediction models built on different transaction features extracted from each corresponding Bitcoin transaction graph. Evaluated on real-world Bitcoin transaction data, we show that our spatial-temporal forecasting model is efficient with fast runtime and effective with forecasting accuracy over 60% and improves the prediction performance by 50% when compared to forecasting model built on the static graph baseline.
EXISTING SYSTEM :
The proposed system evolves on-the-fly and is capable of predicting how likely the two accounts will make transactions in the near future. Third but not the last, we apply the Multiplicative Model Updates (MMU) ensemble to combine prediction models trained over different transaction features extracted from the bitcoin transaction graph. The ensemble ensures the stable yet competitive performance of the proposed Bitcoin transaction forecasting system. We achieve accuracy of over 60% on the future transaction forecasting and improve the performance by more than 50% when compared to the forecast model built on the static graph baseline
DISADVANTAGE :
A straightforward way to construct the Bitcoin transaction graph is considering sender addresses and receiver addresses as nodes, sender-receiver pairs as edges, and Bitcoin amount as weight. Since one address may involve in multiple transactions, there can be multiple single-direction edges from the sender vertex to the receiver vertex over time. Besides, the role of the sender and the receiver can also switch. As no graph mining algorithm can process such complicate repeated, weighted and directed connectivity between nodes, it is natural to simplify the problem.
PROPOSED SYSTEM :
Most of the existing work on the statistical analysis of Bitcoin transaction data falls into this category.
In recent years, utilize the Bitcoin transaction graph data to make Bitcoin price prediction. However, none of the existing work, to the best of our knowledge, has developed a DNN-model-based transaction forecasting system. Example predictions include the likelihood of making a transaction between two accounts, or which account is the most likely to conduct a transaction with a given account.
ADVANTAGE :
These features are either focused only on the (spatial) graph scale or on the (temporal) time-changing scale.
Different from these papers, we combine different spatial and temporal features to capture the dynamics in Bitcoin transactions.
Due to the high dynamics of the Bitcoin transaction and the changing transaction pattern, which embedding feature has the best ability to capture transaction pattern varies over time.
In a dynamic environment, we iteratively choose transaction forecasting models constructed from embedding from different Bitcoin transaction features without knowledge of the future
|