Forecasting Methodology of Scientific Investigations and Innovations Sphere’s Indicators by Means of Neural Networks

Abstract

The aim of the work is to elaborate the methodology, economic and mathematic models and tool means for short-term forecasting the scientific investigations and innovations sphere’s indicators. It considers the system of neural networks forecasting the economic indicators as a part of hybrid, regressive and intellectual forecasting system, and its implementation as exemplified by the scientific investigations and innovations sphere’s indicators in the Russian Federation. The forecasting data by 72 measures (96%) out of 75 scientific investigations and innovations sphere’s indicators were successfully modeled and received. The conclusion is made on the basis of computer experiment that the use of such a system allows not only to raise the accuracy and quality of the forecast calculations, but to use them also in management outlines for achieving the target indicators. The further investigations are aimed at system parameters’ optimization in order to improve the efficiency without losing accuracy and quality, improving the management service of indicators’ models calculations.

Keywords

• innovations
• investigations and innovations sphere
• the systems of regressive equations
• short-term forecasting models
• forecasting retro verification
• the system of neural networks forecasting models
• the system of hybrid models