Neural networks to trade cryptocurrencies

The Financial University under the Russian Government has developed software that can guess the moment when market trends turn around with a high level of accuracy. Using convolutional neural networks it is possible to predict the situation in the market, including the cryptocurrency market. 

The project was carried out at the instruction from the Government. Government agencies will benefit from the system as they lay out and clarify economic development scenarios and monitor the state of key industries, including those that are sensitive to the market conditions. Other market players can use AI to develop efficient trading strategies. 

According to professor Vladimir Solovyov, Head of the Data Analysis Department, Decision-Making and Financial Technology at the Financial University, roboadvising is becoming a promising trend while AI opens opportunities for a steady 25% to 30% income in traditional stock markets. 

Working on an efficient algorithm, university researchers dropped the idea of involving a neural network based on reinforcement learning, or a generative adversarial network, which demonstrated results not better than those of humans. Scientists preferred convolutional networks. 

“Convolutional networks showed very good results despite the fact that they were learning on a small amount of data. When the data selection is larger, the quality of decisions will improve,” Vladimir Solovyov stressed. 

The developers plan to apply the software to the cryptocurrency market. Solovyov says that within a month, the Financial University website may start displaying a trend indicator for most common tokens. It is already available in the mobile app. The developers want to make it openly accessible. 

However, not all market players share the same opinion of such tools. Leonid Matveyev, Head of Stock Exchange Analytics at Waves, believes that the mathematical methods of the neural networks will operate with a 50/50 probability. 

“The cryptocurrency market has low liquidity and a bigger actor may push it aside thus drastically changing the circumstances, The currency rate will either soar or plunge. And no neural network can predict this,” Leonid Matveyev notes. “Therefore, it would be pointless to use such a network without a quality algorithm for generating revenue.” 

Earlier, the Financial University started developing a bitcoin aquarium. 

Previous ArticleNext Article