A Novel Method for Optimizing the Hyperparameters of a Fuzzy Rule-Based System Used for Time Series Prediction

Time series prediction is a critical task across various domains, including finance, healthcare, weather forecasting, and energy systems. While traditional approaches, such as statistical methods and modern machine learning techniques, have demonstrated effectiveness, they often suffer from limitations, including strict assumptions about data or high computational requirements. Fuzzy logic offers an alternative framework, leveraging its capacity to manage uncertainty and imprecision, characteristics commonly found in real-world time series data. In this paper, we propose an adaptive fuzzy rule-based system in which the rules are automatically generated from data, and the membership function parameters are optimized using the Chemical Reaction Optimization (CRO) algorithm. To evaluate the performance of the proposed method, experiments are conducted on a time series dataset of USD-to-VND exchange rates. The experimental results demonstrate that the proposed method outperforms several well-known optimization techniques, including grid search, random search, and genetic algorithm, in terms of achieving the lowest mean squared error and the shortest computational time.