FORECASTING THE GLOBAL INNOVATION INDEX BASED ON NEURAL NETWORKS

This article is devoted to researching the possibilities of using neural networks for forecasting the Global Innovation Index (GII). Modern approaches to GII analysis and forecasting are reviewed, their limitations are revealed, and a new method based on deep neural networks is proposed. Scientific works on the forecasting of innovative activity and the forecasting of GII with the application of mathematical methods of machine learning and their possible areas of application in the digital economy are studied. The structure of GII was analyzed and the key factors affecting its dynamics were identified. A neural network architecture optimized for GII forecasting, taking into account its specificities, has been developed. The model is trained on historical GII data and its accuracy is evaluated on a test sample. The effectiveness of the developed model is compared with the traditional methods of GII forecasting. The positive aspects of the proposed approach, which provides more accurate forecasts compared to traditional methods, are established. The possibilities of interpreting the neural network results, allowing to identify key factors influencing the forecast, are explored. The practical aspects of applying the developed model to support decisionmaking in the field of innovation policy are discussed. The directions of further research are determined, in particular, regarding the interpretation of the results of the neural network and its adaptation to changes in the methodology of GII calculation. The developed model demonstrates high accuracy, but it should not be considered as a replacement for expert analysis and traditional methods of evaluating innovation potential; it should serve as an additional tool that complements existing approaches and contributes to making more informed decisions.