AN INTELLIGENT SYSTEM FOR TEMPERATURE FORECASTING BASED ON A MODIFIED FUZZY NEURAL NETWORK

Abstract

In the setting of ambiguous weather forecasting, neural networks outperform traditional approaches in identifying nonlinear temperature dynamics with uncertainty. However, in order to improve neural network accuracy, an intelligent adaptation for a specific location that is implemented through an intelligent forecasting system with a modified fuzzy neural network with an attention mechanism adapted for conditions of weather forecasting ambiguity registered on various meteorological websites is required. The study describes the design and test simulation of an intelligent system for a temperature forecast based on a modified fuzzy neural network with an adaptive attention mechanism, highlighting significant forecasting aspects such as nonlinear temperature dynamics based on repository data, using modified authors’ software. The findings of the system developed demonstrate its robustness and a decrease in the root mean square error of its forecast by three on average compared to the recurrent neural networks in the setting of undefined and ambiguous weather forecast.