An enhanced alpha evolution moss growth optimizer for prognostic prediction in spontaneous intracerebral hemorrhage

Spontaneous Intracerebral Hemorrhage (SICH) is a critical condition with high mortality rates, requiring prompt and effective prognostic assessment. This study aims to improve SICH outcome prediction by developing the Alpha Evolution Moss Growth Optimization (AEMGO) algorithm for feature selection in high-dimensional medical datasets. Current prognostic models, such as clinical scoring systems, often fail to leverage imaging data fully and struggle with non-linear relationships, necessitating advanced machine-learning solutions. AEMGO innovatively integrates an Alpha evolution strategy to strengthen global exploration. Additionally, it uses an adaptive β parameter to dynamically balance the search process. AEMGO is designed to overcome common issues in traditional meta-heuristic algorithms, such as premature convergence and the exploration–exploitation imbalance, thus enabling the effective identification of key prognostic features. The efficiency of AEMGO is validated against the IEEE CEC 2017 benchmark suite, demonstrating its ability to optimize solution stability and diversity. Furthermore, we adapt AEMGO to its binary version (bAEMGO) and combine it with a Support Vector Machine (SVM) to create the bAEMGO-SVM method. This hybrid approach is applied to feature selection in a comprehensive dataset, including CT images, clinical data, and laboratory results, to enhance the predictive model’s accuracy. Our findings reveal that the bAEMGO-SVM model achieves competitive performance on the collected dataset, achieving an accuracy of 87.5% and sensitivity of 88.73%—improvements of up to 8.125% in accuracy over standard SVM—while demonstrating faster convergence compared to competitors. This study provides a robust tool for enhancing prognostic prediction of SICH patients, with potential for clinical implementation to support timely and personalized treatment decisions.