Application of Lévy and sine cosine algorithm hunger game search in machine learning model parameter optimization and acute appendicitis prediction

• Published in: Expert systems with applications Vol. 269; p. 126413
• Main Authors: Qu, Shizheng, Liu, Huan, Zhang, Hanwen, Li, Zhuoshi
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 15.04.2025
• Subjects: Artificial intelligence; Disease prediction; Hunger game search algorithm; Machine learning; Parameter optimization; Hunger game search algorithm; Parameter optimization; Artificial intelligence; Machine learning; Disease prediction
• ISSN: 0957-4174; 1873-6793
• DOI: 10.1016/j.eswa.2025.126413

If acute appendicitis is not treated in time, it will develop complications such as perforation and gangrene, which will lead to death in severe cases, so it is necessary to make an accurate diagnosis early. Traditional inspection methods are time-consuming and rely heavily on operator experience. To address these challenges, this paper proposes a machine learning model optimized by integrating sine cosine algorithm ideas and introducing Levy flight-improved hunger games search. The data of 354 patients in Zhongshan Hospital Affiliated to Dalian University in the past five years were collected as the data set of this study. In order to solve the problems of local optimum traps and slow convergence speed that may occur in the optimization process of the original hunger game search algorithm, learning from the idea that the amplitude is large in the early stage and decreases with iteration of the sine cosine algorithm, the original search strategy is replaced by a periodic oscillation search pattern, and the Levy flight strategy is embedded on the hunger parameter. It guides the global search in the early stage of the algorithm, and focuses on the known best position and accelerates the convergence in the later stage. The improved algorithm was compared by IEEE Congress on Evolutionary Computation 2017 test functions and optimizing different machine learning models, such as Support Vector Machine, Random Fores, and K Nearest Neighbor. The accuracy of the proposed model is improved by 0.01429–0.25171. Experiments show that the performance of the proposed method is better than other models, which can provide a reliable basis for the prediction of acute appendicitis.