Hybrid Intelligent Framework for Adaptive Decision-Making Systems

This study proposes a Hybrid Intelligent Framework that integrates Neural Networks (NN), Fuzzy Logic Systems (FLS), and Evolutionary Computation (EC) to improve adaptive decision-making in dynamic, uncertain, and data-driven environments. The framework combines data-driven pattern learning using a multilayer perceptron, interpretable fuzzy reasoning through Mamdani inference and centroid defuzzification, and evolutionary optimization to tune network weights, membership parameters, and fuzzy rule structures. Two dataset categories were used to assess robustness: simulated decision scenarios and industrial datasets with dynamic operational variables. Data were normalized via min–max scaling and fuzzified using Gaussian membership functions before being processed by the NN–FLS pipeline. EC then minimized a weighted objective that balances prediction error and rule complexity, enabling accurate yet explainable decisions. Performance was evaluated using accuracy, MAE, RMSE, and F1-score, and compared against standalone NN and standalone FLS baselines. The hybrid model achieved the best results, reaching 92.3% accuracy and 0.93 F1-score while reducing MAE to 0.32 and RMSE to 0.48. These findings indicate that hybridizing learning, reasoning, and optimization yields faster adaptation and lower error rates than single-model approaches, supporting scalable deployment in real-world decision-support systems. Confusion-matrix inspection also showed fewer critical misclassifications under changing conditions, supporting suitability for online updates in practice.