University of Tehran Press Industrial Management Journal 3115-7386 18 4 2026 10 01 Presenting a Global Optimization-based Dynamic Fuzzy Inference System (GODFIS) for online data stream prediction 10 47 108605 10.22059/imj.2026.412275.1008301 EN Farzad Bahrami Assistant Prof., Department of Industrial Management, Faculty of Management, University of Arak, Arak, Iran. 0000-0002-3327-648X Mohammad Hosein Alavidoost Ph.D., Department of Industrial Management, Faculty of Management, University of Tehran, Tehran, Iran. Journal Article 2026 03 17 Objective: Real-world data streams exhibit nonlinear and non-stationary characteristics that create significant challenges for effective predictive modeling. While evolving fuzzy systems (EFS) can dynamically update their structure and parameters to address these challenges, existing methods primarily focus on system identification rather than the optimality of learned parameters. This paper aims to bridge this gap by proposing a novel Global Optimization-based Dynamic Fuzzy Inference System (GODFIS) designed to enhance prediction accuracy through optimized parameter estimation. Methodology: The proposed GODFIS framework leverages two key concepts: recursive center-of-gravity and global parameter estimation. Under this framework, we introduce a new Global Optimization-based Evolving Clustering Algorithm (GOECA) to optimize clustering in the response space and accurately determine initial parameters. Furthermore, a novel noise elimination principle is incorporated to maintain response quality and mitigate the negative impact of noise on the system’s knowledge base. Results: To evaluate the performance of GODFIS, extensive experiments were conducted on standard benchmarking datasets. The empirical results demonstrate that the proposed model achieves significantly higher prediction accuracy in solving benchmark problems compared to state-of-the-art and existing evolving fuzzy approaches. Conclusion: By integrating global parameter optimization with robust noise elimination, GODFIS effectively addresses the limitations of conventional evolving fuzzy systems in non-stationary environments. The findings confirm that optimizing learned parameters and refining the clustering process lead to superior and more reliable predictive performance in dynamic, real-world data streams https://imj.ut.ac.ir/article_108605_f927519212c629fdb8f6cf9a616ef348.pdf