Design of a multi-stage hybrid optimizer combining Bayesian and evolutionary algorithms for modeling and optimizing seizure-related information in EEG signals

The prediction of epileptic seizures from EEG signal analysis is modeled as a complex, high-dimensional, multi-objective optimization problem. In this study, we propose and design a novel multi-stage hybrid optimizer that integrates Bayesian optimization with a family of evolutionary and metaheuristic algorithms, including Genetic Algorithm (GA), Evolutionary Programming (EP), Evolution Strategies (ES), Ant Colony Optimization (ACO), Artificial Bee Colony (ABC), and Harmony Search (HS).This optimizer is structured to operate across sequential stages of optimization, where each stage is designed to progressively refine the solution space by selecting the optimal subset of EEG channels and time blocks related to the preictal phase. The mathematical model aims to maximize the posterior probability of seizure occurrence while minimizing the temporal distance to the actual seizure, thus formulating a well-defined trade-off in the objective space. Unlike traditional methods, the proposed hybrid framework models and controls the optimization trajectory by combining the global exploration capability of evolutionary algorithms with the adaptive search of Bayesian optimization. Experimental results demonstrate that the optimizer successfully converges toward high-informational regions of EEG data that occur early in the preictal phase. A detailed performance comparison and sensitivity analysis further validate the effectiveness and robustness of the proposed approach, offering insights into the design of more generalizable hybrid optimizers for complex modeling tasks.

کلیدواژه ها:

Multi-Objective Optimization ، Hybrid Optimizer ، Bayesian optimization ، Evolutionary Algorithms ، Posterior Probability ، Seizure Prediction