A method for forming a multi-tiered neural network forecasting system with the possibility of reconfiguration

   Purpose of research. Improving the accuracy of forecasting by identifying logical connections in unstructured datasets and forming a multi-tiered structure of a specialized neural network computing system.

   Methods. A parallel algorithm for determining the fragmented structure of the training sample is proposed, which is used to isolate fragments containing training data based on the logical dependencies of the sample. Based on the generated fragmented sample, a method for assembling neural networks has been developed, which is used to form an effective structure of a cascade forecasting system.

   Results. Forecasting the results of the unofficial team competition of the International Student Sports Festival 2023 was chosen as the main experiment. A fragmented training sample has been formed on the basis of which a cascade of neural network modules has been built. Four cascade variants were tested in experiments, which showed a significant increase in prediction accuracy compared to single-module analogues. To significantly improve the performance of a neural network system with ultra-short-term forecasts, the hardware implementation of cascades based on the decisive field of FPGA is considered. The structure of the complex with the possibility of its reconfiguration is proposed.

   Conclusion. The use of artificial neural networks in forecasting is promising, but it may face problems of inaccuracy of results due to insufficient computing power and collisions in training samples. One of the proposed solutions to the problem is cascading specialized neural network modules. Positive results were demonstrated by both the software and hardware implementation of the system based on the proposed cascade. The evaluation of the hardware implementation demonstrates the possibility of acceleration, compared with the software implementation, which may be necessary when conducting ultra-short-term forecasts. The proposed methods and algorithms have demonstrated their correctness.

1. Baltagi B.H., Li D. Prediction in the panel data model with spatial correlation. Springer, Berlin: Advances in Spatial Econometrics: Methodology, Tools and Applications. 2004; 13: 283-295.

2. Kholodilin, K.A., Siliverstovs, B., Kooths, S. A dynamic panel data approach to the forecasting of the GDP of German Länder. Spatial Economic Analysis. 2008; 3: 195-207.

3. Belouschenko Ya. A. Statistical methods of forecasting in economics. Kerch; 2016. 44 p. (In Russ.).

4. Reutskaya E.A. A comprehensive diagnostic system for ski racers in order to determine the prospects and predict predisposition to high sports results. Nauka i sport: Sovremennye tendentsii = Science and sport: Modern trends. 2022. 10(4): 79-87. (In Russ.).

5. Zhukov S.V., Zelensky A.V. The use of neural networks in building optimal training trajectory in biathlon. Uchenye zapiski universiteta im. P.F. Lesgafta = Scientific notes of the P.F. Lesgaft University. 2017; (5): 81-88 (In Russ.).

6. Yushkin V.N. Problems of sports forecasting. Uchenye zapiski universiteta im. P.F. Lesgafta = Scientific notes of the P.F. Lesgaft University. 2021; (4): 473-478. (In Russ.).

7. Krutikov A.K., A parallel algorithm for the formation of a fragmented training sample for a specialized forecasting system based on the assembly of neural network modules. Vestnik Permskogo natsional'nogo issledovatel'skogo politekhnicheskogo universiteta. Elektrotekhnika, informatsionnye tekhnologii, sistemy upravleniya = Bulletin of the Perm National Research Polytechnic University. Electrical engineering, information technology, control systems. 2024; (49): 61-81. DOI: 10.15593/2224-9397/2024.1.04 .

8. Rozyhodzhaeva G.A., Rozyhodzhaeva D.A. Features of the formation of a training sample and training of a neural network with incomplete input data in solving private medical problems. Nauchnoe obozrenie. Biologicheskie nauki = Scientific Review. Biological sciences. 2017; (5): 28-32. (In Russ.).

9. Dorokhova Yu.V., Goncharuk N.S. The use of the method of expert assessments in forecasting social processes. Srednerusskii vestnik obshchestvennykh nauk = Central Russian Bulletin of Social Sciences. 2014; (3): 76-81 (In Russ.).

10. Zhukov V.G., Bukhtoyarov V.V. On the use of artificial neural networks with radial basis functions in the tasks of detecting anomalies in network traffic. Reshetnevskie chteniya = Reshetnev readings. 2013. 2(17): 285-286 (In Russ.).

11. Konakoglu B., Cakir L. Generalized Regression Neural Network for Coordinate Transformation. Montenegro, International Symposium on Advancements in Information Sciences and Technologies (AIST). 2018. P. 66-74.

12. Malyutin A. Can't count the gold? Forbes. Available at: http://www.forbes.ru/forbes/issue/2004-08/20514-ne-schest-zolota (acceddes: 21. 05. 2024). (In Russ.).

13. Yasnitskiy L.N., Pavlov I.V., Cherepanov F.M. Forecasting the 2014 Olympic Games in the unofficial team competition using artificial intelligence methods. Sovremennye problemy nauki i obrazovaniya = Modern problems of science and education. 2013; (6). (In Russ.).

14. Deep Learning Toolbox. Available at: https://exponenta.ru/ neural-network-toolbox (accessed: 02. 04. 2024).

15. Krutikov A.K., Meltsov V.Y., Strabykin D.A. Evaluation the Efficiency of Forecasting Sports Events Using a Cascade of Artificial Neural Networks Based on FPGA. Proceedings of ElConRus-2022. St. Petersburg: ETU LETI; 2022. P. 355-360.

16. FPGA. Available at: https://alt.ru/Mikroshemi.html (accessed: 01. 02. 2024).

17. Levin I.I., Dordopulo A.I., Kalyaev I.A., Doronchenko Yu.I., Razladkin M.K. Modern high-performance computing systems with reconfigurable architecture based on FPGAs Xilinx Virtex–7 and Virtex UltraScale. Russian Supercomputing Days 2015. 2015; 435-447 (In Russ.).

18. Dordopulo A.I., Kalyaev I.A., Levin I.I., Semernikov E.A. High-performance reconfigurable computing systems of a new generation. In: Trudy Mezhdunarodnoi superkomp'yuternoi konferentsii s elementami nauchnoi shkoly dlya molodezhi «Nauchnyi servis v seti Internet: ekzaflopsnoe budushchee» = Proceedings of the International Supercomputer Conference with elements of a scientific school for young people "Scientific service on the Internet: an examflop future". Moscow: Izd. MGU; 2011. P. 42-49 (In Russ.).

19. Kachanovsky Yu.P., Kachanovsky Yu.P., Korotkov E.A., Korotkov E.A. Data preprocessing for neural network training. Fundamental'nye issledovaniya = Fundamental research. 2011; (12-1): 117-120. (In Russ.).

20. Guzik V.F., Kalyaev I.A., Levin I.I. Reconfigurable computing systems. Rostov-on-Don; 2016. 471 p. (In Russ.).