Enhanced slime mould algorithm with chaotic and orthogonal optimization-based learning for improved severity prediction accuracy in malaria patient outcomes.

Malaria remains a critical health challenge in developing countries, particularly in Africa, where it disproportionately affects vulnerable populations. Accurate malaria severity prediction is important for proper treatment and improved patient survival rates. This study proposes an improved Slime Mould Algorithm (iSMA) to address the limitations of the standard SMA, including slow convergence, poor initialization, local optima entrapment, and imbalance between exploration and exploitation. The iSMA incorporates four key innovations: a Chaotic Initialization Strategy (CIS) to generate a diverse initial population, enhancing solution diversity; an enhanced Opposition-Based Learning (eOBL) mechanism to strengthen exploration; an Orthogonal Learning (OL) strategy to improve the balance between exploration and exploitation while accelerating convergence; and a Restart Strategy (RS) to reinitialize underperforming individuals, effectively preventing entrapment in local optima. Integrating the Random Forest (RF) feature selection process with the Support Vector Machine (SVM) classifier, forming the proposed new model RF-iSMA-SVM, classifies malaria into severe and no-severe cases. The iSMA performed better than nine other recent metaheuristic algorithms, as was broadly exhibited in the benchmark tests conducted on CEC2022 functions. When testing the RF-iSMA-SVM model on malaria datasets from health centers in Sierra Leone, the proposed model demonstrated superior predictive performance compared to recent conventional methods. The results yield an accuracy value of 0.981, specificity of 0.827, sensitivity of 0.974, and a Matthews Correlation Coefficient (MCC) score of 0.794. This shows the significant improvement over the standard techniques used for predicting the severity of malaria, which indicates great potential for healthcare practices and decision-making. These results support the model's reliability and effectiveness in improving the prediction of malaria severity. This study offers a novel practical tool for malaria case management by combining advanced optimization techniques with machine learning. It could enhance healthcare decision-making and improve outcomes for affected populations.