Recurrent fuzzy neural networks for time-series prediction

Abstract

Time series forecasting is a fundamental challenge in many scientific and industrial domains, especially when dealing with chaotic and highly nonlinear systems. Traditional forecasting models, including Artificial Neural Networks (ANNs) and statistical methods, often struggle to capture the complex temporal dependencies and uncertainty inherent in these datasets. To address these limitations, hybrid approaches that integrate fuzzy logic and Recurrent Neural Networks (RNNs) have emerged as promising alternatives. This thesis presents a comparative study of two hybrid models: the Multi-Functional Recurrent Fuzzy Neural Network (MFRFNN) and the Recurrent Neurofuzzy System ReNFuzz-LF. The effectiveness of these models is evaluated across multiple datasets, including Electric Loads, Lorenz chaotic system, Box–Jenkins Gas Furnace, Wind Speed Prediction, Google Stock Price Prediction and Air Quality Index (AQI).