- Статьи
- Economy
- Exchange Oracle: Russia has created a neural network for predicting stock market crises
Exchange Oracle: Russia has created a neural network for predicting stock market crises
Russian developers have created a neural network that can predict a short-term stock crisis with an accuracy of more than 83% a day before the event, the press service of the Higher School of Economics reported. The system combines three different machine learning architectures. Izvestia investigated how effective AI technologies are in predicting economic shocks and what indicators should be taken into account when choosing investment tools.
Initial assessment
AI tools cannot be an absolute solution in predicting future crisis periods, as classical approaches based on regression solutions are important, Tamara Teplova, professor at the HSE Faculty of Economics and one of the authors of the neural network, clarified in an interview with Izvestia.
— In our published research, we emphasize the importance of AI algorithms as an early warning system. Combining these approaches, it is possible to create powerful algorithms for early warning of risk growth in the market. However, this does not give us the opportunity to predict "black swans", otherwise life would be uninteresting and boring, the expert believes.
Using a neural network to predict market events is interesting enough in itself, although it is not new. They have been "tormenting" the data and looking for stable patterns for quite some time and with varying success, Anton Tabakha, chief economist at the Expert RA rating agency, told Izvestia.
"The main problem with such models is that market participants change their behavior, which in turn changes patterns and makes it more difficult to identify them, not to mention that the neural network is unable to assess "black swans", that is, man—made or external events that trigger crises," he said. — Therefore, the experiment is interesting, but it does not replace more traditional risk assessment tools, such as credit ratings.
Today, AI technologies are rapidly developing and being implemented in many areas where other methods of analysis and forecasting were previously used, Artem Lyukshin, PhD in Economics, Associate Professor of the Department of State and Municipal Finance at Plekhanov Russian University of Economics, confirmed in an interview with Izvestia. When using AI technologies to predict economic shocks, it is worth remembering that they do not have absolute accuracy and their testing period is short. As a rule, AI technologies are based on historical data and current financial and economic indicators. If there is an economic crisis unlike any seen before, it will not be easy to make predictions using AI.
"It is better to use AI technologies in conjunction with other qualitative and quantitative methods of analysis and forecasting and make a decision based on the totality of the results obtained," the expert advised. — The above also applies to the choice of investment instruments based on AI technologies. In addition, it is worth considering factors such as the purpose of the investment, the level of risk assumed, the timing and profitability, and the amount of initial investment.
The use of neural networks for economic forecasts can help in decision-making, Igor Balynin, associate professor at the Financial University under the Government of the Russian Federation, told Izvestia. In particular, it can be useful in risk management in identifying objectively emerging trends and highlighting problem areas. At the same time, verbal interventions, including those that are far from always true, are now having an increasing impact on short-term fluctuations in the financial market.
— Neural networks can be useful in the initial identification of certain risks, the establishment of individual relationships, as well as for selecting a list of potential options for managing them. At the second stage, in order to obtain a reasonable response to the proposals of the neural network, an expert economic assessment with the involvement of a human will be required," the expert emphasized. — Therefore, I believe that a neural network can be useful as one of the participants in a brainstorming session, but not as a financial decision maker.
Format Search
When choosing investment instruments, it is necessary, first of all, to correlate your own expectations from the market with the instruments presented on it, Tamara Teplova said. Over the past two or three years, the average investor's profile in the Russian market has changed a lot, so factors such as social media, the opinions of well-known analysts and bloggers increase the weight of the average private investor in the decision-making system.
—Classical practice says that the basic way to evaluate an asset is to look at the issuer's management, and then open its financial statements and calculate something," she recalled. — However, information distortions are currently increasing, and now it is important to anticipate the reaction of other market participants who like to discuss assets online and focus on the opinions of other users of online investment platforms. By combining fundamental analysis and online opinion analysis, they will already be able to reduce the risk of incorrect expectations from market fluctuations.
When choosing investment instruments, it is important to understand who exactly carries it out and at least the following parameters: risk appetite, the allowable investment period, the amount of available funds, Igor Balynin noted.
— Currently, the classic rule "profitability is higher where the level of risk is higher" does not always work. In my opinion, the safest investment of available funds (within 1.4 million rubles in one bank) at the moment is a bank deposit. Currently, there is an opportunity to earn about 17-18% per annum on the funds placed in it. Accordingly, having invested 100 thousand rubles today, in a year it will be possible to withdraw 117-118 thousand rubles," the expert reminded.
In his opinion, the most risky assets are currently located in unfriendly countries, including their currencies: this is where they are most likely to lose their investments.
"For example, those who invested in US dollars in August 2024 not only failed to earn, but lost about 10% only on changes in exchange rates, and taking into account banks' exchange fees, the losses are even higher," he noted. — Accordingly, those who invested 100 thousand rubles in US dollars a year ago can now withdraw less than 90 thousand rubles. Therefore, I believe that when forming an investment portfolio, the assets of unfriendly countries should not be considered regardless of the level of risk assumed.
Izvestia sent requests to the Ministry of Energy, the Ministry of Finance and the Ministry of Finance, but no responses had been received at the time of publication.
Переведено сервисом «Яндекс Переводчик»
