Developing the Neural Network Forecasting System of Dynamics on the Russian Market of Futures Contracts

In this paper, we consider the forecasting system development is based on the neural network. The financial market has been chosen as the object of the research. The subject of research is the method of construction and training of Kohonen Self-Organizing Map (further - SOM) as universal instrument of modeling processes and solving applied forecasting and classification problems. The test instrument is the RTS index futures contract with execution in June of 2017 (further - RTS). The work purpose is to create neural network system for efficient forecasting of financial market. In the work process model of SOM has been developed for exchange data analysis on the basis of their correlation. This work consists of eleven sections, five of which are the main part of the work having the mathematical description of the neural network model and its practical application. The Introduction introduces the general concept of artificial intelligence, as well as the relevance of using the chosen neural network model. The Task Statement describes the purpose of the study and the result that should be obtained. Section Mathematical formalization contains the mathematical formulas underlying the neural networks. The Description of the Kohonen-like Neural Network Model includes a general representation of the model of self-organizing maps work. The Learning Algorithm of the Kohonen Network considers the learning algorithm and the neural network model using in this work. The result of the software and estimate of effectiveness in practical application is presented in the section Experimental Studies. Conclusion is devoted to the main conclusions are based on the results of the system. The Attachment contains the program code and its description. The software is implemented in C # in the Visual Studio development environment.