FORECASTING FINANCIAL MARKETS USING CONVENTIONAL NEURAL NETWORK

Task. In the modern world, associated in particular with the development of high technologies, new, previously unused tools for analyzing big data are being developed, the use of which has also captured the sphere of financial markets. To predict price movements in financial markets, neural networks have been successfully used, which, unlike other algorithms, are not programmed, but self-learning. In the process of training, a neural network is able to identify complex dependencies between input and output data and, on their basis, predict new data. The article discusses one of the ways to use neural networks to predict financial markets, in particular, the development of a mathematical model for a convolutional neural network used to recognize the state of the financial market and predict future moments of trend reversal. In general, the developed mathematical model and machine learning algorithm for predicting financial market conditions based on a convolutional neural network are working, although they require additional refinement. Improving the accuracy of predictions of the described models is the main direction of further research. Model. Along with the traditional methods of forecasting (fundamental analysis and technical analysis), the article highlights the methods of data mining, the most famous of which is the method using neural networks. Conclusions. The results obtained by the authors indicate a high potential for using this technology. However, the process of training neural networks is quite expensive both in terms of computing resources and time. Note that this study is only one of the steps to building an effective tool for forecasting the stock market. Practical value. The practical importance of the study is to identify the optimal moments of opening a position - buying assets on an uptrend and implementing them on a downtrend, which should be made as close as possible to the moment of the next change in the market position in order to maximize the investor's profit.

financial market forecasting, machine learning, convolutional neural network, mathematical model, algorithm