Журнал «Современная Наука»

Russian (CIS)English (United Kingdom)

MOSCOW  +7(495)-755-19-13

A+ R A-

Preprocessing of statistical data to improve the quality of the forecast by a neural network

E-mail Print

A. Gilmanov, A. Gusev, A. Okunev,  (Perm State National Research university)

Series "Natural & Technical Sciences" # 03  2018
The article describes the method of functional preprocessing of statistical data to improve the forecast obtained with the help of neural networks. We consider a fairly wide range of functions that can be used to pre-process statistical data. The advantage of neural networks for forecasting using data preprocessing is shown, in terms of forecast stability. The forecast error is considered as a random variable for which: statistical estimates for the mathematical expectation and for the standard deviation are calculated, and a selective coefficient of variation is calculated to determine the most stable forecast model.

Keywords: functional preprocessing, forecast, stability of the neural network model, coefficient of variation.


Read the Full Article in Russian …

©  A. Gilmanov, A. Gusev, A. Okunev, Journal "Modern science: actual problems of theory and practice".



Перепечатка материалов допускается только в некоммерческих целях со ссылкой на оригинал публикации. Охраняется законами РФ. Любые нарушения закона преследуются в судебном порядке.
© ООО "Научные технологии"