融合改进河马优化算法与Stacking集成的乳腺癌诊断模型OA

Breast cancer is a major disease that seriously threatens people's lives and health,and its early and accurate diagnosis is the key to improving patients'survival rates.Traditional machine learning models exhibit inefficiencies in hyperparameter tuning and are prone to local optima,resulting in slow convergence speed,low optimization accuracy and poor robustness.The insufficient optimization of base models in ensemble learning methods makes it difficult to fully exploit the complementary advantages among models.To address these challenges,a Stacking ensemble model(IHO-Stacking)integrating improved hippopotamus optimization(IHO)algorithm is proposed.The population initialization mechanism is reconstructed by means of Latin hypercube sampling,and the hippopotamus optimization algorithm is improved by combining with the Lévy flight strategy and Gaussian mutation mechanism.The IHO algorithm is used to globally optimize the hyperparameters of base learners including extreme gradient boosting(XGBoost),categorical boosting(CatBoost)and random forest(RF),and integrate the advantages of multiple models by means of the Stacking framework.The experimental results show that the IHO-Stacking model can realize outstanding performance on the Wisconsin breast cancer dataset,with an accuracy of 0.985 4,malignant sensitivity improved to 0.986 2,and benign specificity increased to 0.990 7.It significantly outperforms traditional optimization methods and single models,and can provide strong technical support for the precision and standardization of early breast cancer screening.