An expert on analytical data from the Indian Vellore Institute of Technology, demonstrated a method for predicting prices for digital currency in real time.
The method is based on the use of machine learning technology with the architecture of a neural LSTM network (long short-term memory). The scientist, Abinhav Sagar, showed the process, which consists of four stages:
1) Real-time tokenToken
– is an accounting unit that is used to represent a digital balance in an asset.Details data collection;
2) Their preparation for training the neural network;
3) Testing the forecast using the LSTM network;
4) Visualization of the results.
A set of data from CryptoCompare, such as price, volume and opening, high and low values, served to train the neural network. As cryptocurrency news reports, the researcher also, as an indicator for estimating forecasts, used the average absolute deviation, measuring the average value of errors in the forecast without taking into account their direction.
Abinhav Sagar posted a link to the code for the entire project on GitHub and described the functions that he used to normalize data values in preparation for machine learning.
Although machine learning has been successful in predicting stock market prices using a variety of time series models, its use in cryptocurrency price forecasting is quite limited. The reason for this is obvious, since cryptocurrency prices depend on many factors, such as technological progress, internal competition, market pressure, economic problems, security problems, political factors, etc. Their high volatilityVolatility
– is a financial term that means a statistical indicator of price change. In the cryptocurrency world, it is used by traders as an indicator for managing financial risks. That is, it is a measure of trade risk and as a financial analytical tool for gambling. Details also speaks of the enormous potential of high profits if reasonable tradingTrading
– is an economic term that means the process of independent trade, independent analysis of the market and the conclusion of trade transactions.Details strategies are used. However, cryptocurrencies remain relatively unpredictable compared to traditional financial instruments.
Interesting in the section: Data keepers: quantum computers versus blockchain
- The service that allows using digital currency to pay for travel services, Travala, has now begun to work with Booking.com.