Forecasting the exchange rate of cryptocurrencies: an article in the NBU magazine
Date: 2024-07-04 Author: Henry Casey Categories: CRYPTO PAYMENTS, IN WORLD
The National Bank of Ukraine (NBU) has published a new issue of its scientific journal "Visnyk of the National Bank of Ukraine". This edition covers a variety of topics, including corporate lending drivers and energy and metals price forecasting. Particular attention is paid to a separate article devoted to modeling the price dynamics of cryptocurrencies.
Research on the price dynamics of cryptocurrencies
The authors of the article, Yuri Kleban and Tatyana Stasiuk from the Ostroh Academy National University, conducted a study aimed at studying the problems of modeling the price dynamics of cryptocurrencies. Using data from the popular Ukrainian crypto exchange Binance, the researchers focused on predicting the prices of Bitcoin, Ethereum, Ripple (XRP) and Dogecoin (DOGE).
To analyze the data and predict the prices of crypto assets, the researchers used machine learning methods, including a neural network with long-term memory (LSTM). Data was collected from July 6, 2020 to April 1, 2023. The study showed that the LSTM model demonstrated the best results on RMSE, MAE and MAPE criteria, outperforming traditional models such as ARIMA and Prophet, as well as the naive model.
Advantages and challenges of the LSTM model
The authors note that although the LSTM model is difficult to use, it is a powerful tool for modeling cryptocurrency prices, which are characterized by significant volatility. According to experts, the forecasting accuracy using LSTM is much higher, which makes this model preferable for problems of this kind.
Features of the cryptocurrency market
The article also discusses the features of the cryptocurrency market. The authors emphasize that the cryptocurrency market is in many ways similar to traditional financial assets and is based on the balance of supply and demand. Supply and demand are influenced by factors such as market stability, the price of Bitcoin, the issuance of cryptocurrencies, news and legislative changes.
However, the cryptocurrency market also has its own specific features, such as round-the-clock trading, decentralization, high volatility and algorithmic trading. These features create both opportunities and risks for market participants, including price volatility, threats of cyber attacks, and changes in legislation.
Research results
According to the results of the study, the LSTM model demonstrated significantly lower error and higher accuracy in predicting the price of all four cryptocurrencies compared to the ARIMA, Naïve and Prophet models. The Prophet model showed the worst results among all the models reviewed.
The researchers emphasize that cryptocurrency price modeling is a promising and understudied area of scientific research. Developing mathematical models to predict crypto asset prices is important for creating trading algorithms, bots, and portfolio management tools. With the development of modeling methods and increasing computing power, the accuracy of forecasting will improve.
Expert forecasts
Recall that Binance CEO Richard Teng predicts Bitcoin will rise to $80,000 by the end of 2024. Experts also expect the price of Ethereum to exceed $6,500 due to the influx of funds into spot ETFs based on this asset.
Thus, an article in the NBU magazine emphasizes the importance and promise of research into methods for forecasting cryptocurrency prices, which can significantly affect the development and stability of the cryptocurrency market.
