SYNTHESIS OF ECONOMETRIC AND NEURAL NETWORK MODELS FOR INDICATORS PREDICTION IN RESEARCH AND INNOVATION IN THE RUSSIAN FEDERATION

The article considers general methodology and architecture of hybrid system models for prediction of economic indicators and its implementation in the form of an integrated information system on the example of research and innovation indicators of the Russian economy. The scheme of the distributedinformation-analytical system is demonstrated. The general verification process algorithm of the modelprediction unit is presented, which contributes significantly to the credibility of the forecast results. The object of the study is a unified system of hybrid models, combining econometric and neural network modelsinto a single system of hybrid economic models. The structure of the hybrid forecasting system consists of two subsystems: the subsystem of the distributed econometric forecast models and subsystem of the distributed neural network prediction models. The objective reasons, under which the level best of regression modelsis reached, are identified. The subsystem architecture of the distributed neural network models developedin the programming language Python with the use of the web framework Django is described. The stages of indicators forecasting in a hybrid model are shown. The hybrid models functional structure based on the use of software modules are considered. The use of such a system allows not only to improve the accuracy and quality of the forecasts, but also to apply them in the control loop foreaching the targets.

research and innovation, regression equations systems, short-term prediction models, verificationof prediction, neural network models system, hybrid models system

1. Grishin V.I., Abdikeev N.M. Kolmakov I.B., et al. Sistema rascheta prognoznyh pokazatelej makrojekonomiki Rossii [The calculation system of Russia’s macroeconomic predictive indicators] // Finansovaja analitika. Problemy i reshenija. Nauchno-prakticheskij i informacionno-analiticheskij sbornik [Financial analyst. Problems and solutions. Theoretical and practical and informationanalytical collection], Moscow, Finansy i kredit — Finance and credit, 2010, no. 13 (37), pp. 2–15 (in Russian).

2. Kolmakov I.B., Koltsov A.V., Domozhakov M.V. Osnovy postroenija sistemy kompleksnogo prognoza sfery issledovanij i innovacij vo vzaimosvjazi s makrojekonometricheskimi modeljami jekonomiki Rossii [The foundations of the system of complex forecasting areas of research and innovation interrelated with macroeconometric models of Russian economy]. Innovatika i jekspertiza — Innovation and expertise, 2015, no. 1 (14), pp. 255–275.

3. Abdikeyev N. M. Kognitivnaya biznes-analitika: uchebnik [Cognitive business analytics: textbook]. Moscow, INFRA-M — INFRA-M, 2014, 511 p.

4. Kitova O.V., Kolmakov I.B., Sharafutdinova A.R. Analiz tochnosti i kachestva kratkosrochnogo prognoza pokazatelej social’no-jekonomicheskogo razvitija Rossii [Accuracy and quality analysis of shortterm forecast of social and economic development of Russia]. Vestnik Rossijskogo jekonomicheskogo universiteta im. G.V. Plehanova — Bulletin of Russian Plekhanov University of Economics, 2013, no. 9, pp. 111–119 (in Russian).

5. Kolmakov I.B., Domozhakov M.V. Metodologija prognozirovanija pokazatelej sfery nauchnyh issledovanij i innovacij s pomoshh’ju nejrosetevyh modelej [The methodology of the indicators prediction in research and innovation with the help of neural network models. Menedzhment i biznes-administrirovanie — Management and business administration, 2015, no. 3, pp. 121–127 (in Russian).

6. Amosov O.S., Pashchenko F.F., Muller N.V. Strukturno parametricheskaja identifi kacija vremennogo rjada s primeneniem fraktal’nogo i vejvlet-analiza [Structural parametric identifi cation of time series with the use of fractal and wavelet analysis]. Informatika i sistemy upravlenija — Information science and control systems, 2015, no. 2 (44), pp. 80–88 (in Russian).

7. Rossijskij statisticheskij ezhegodnik. Statisticheskij sbornik. [Russian statistical yearbook. The statistical compilation.] Moscow, Rosstat — Rosstat, 2014, 693 p. (in Russian).

8. Ivanjuk V.A., Pashhenko F.F. Methods and models for the forecasting and management of time series / Proceedings of International work-conference on Time Series (ITISE 2015, Granada, Spain), Granada, 2015, pp. 283–292.

9. Khaikin S. Nejronnye seti. Polnyj kurs. [Haykin S. Neural networks: A Comprehensive Foundation], 2nd edition, translated from English. Moscow, Williams — Vil’jams, 2006, 1104 p. (in Russian).

10. Fors’e Dzh. Bisseks P., Chan U. Razrabotka veb-prilozhenij na Python [Forcier J., Bissex P., Chun W. Python Web Development with Django], translated from English. St. Petersburg, Simvol-Plus — SymbolPlus, 2010, 456 p. (in Russian).