Research On Intelligent Optimization Algorithm And Its Application In Syndrome Classification Of Type 2 Diabete

Objective: With the continuous innovation of theory and technology,machine learning has increasingly shown its important role in the field of medicine.However,the existing machine learning models are also faced with problems such as complex operation and insufficient generalization ability to varying degrees.It is of great significance to build a machine learning model with strong generalization ability.Intelligent optimization algorithms are generated based on optimization strategies or behavior patterns in nature or biological systems.They can solve practical optimization problems by simulating these optimization strategies and behavior patterns,and have good optimization performance.This paper studies intelligent optimization algorithms and applies them to feature selection and model parameter optimization of TCM syndrome data set of type 2 diabetes to improve the classification performance of models.Methods:1.To reduce redundant and irrelevant features in the dataset,genetic algorithm is used in feature selection,GA＿FS＿SVM model is proposed and the underlying classification model is a support vector machine algorithm,experiments are conducted on type 2 diabetes Chinese medical evidence dataset with chi-square test,F-test and mutual information design control experiments;2.To address the shortcomings of the Harris Hawk optimization algorithm,an improved Harris Hawk optimization algorithm(BGOHHO)based on Bernoulli chaos mapping,backward learning,Gaussian variation and nonlinear escape energy update strategies was proposed,and the effectiveness of the algorithm was verified using 16 standard test functions.3.The improved Harris Hawk optimization algorithm(BGOHHO)was used to optimize the penalty coefficient C and kernel function parameter G of the support vector machine,and the BGOHHO＿SVM model was proposed to conduct experiments on the type 2 diabetes Chinese medical evidence dataset to carry out a comparative study with SVM optimized by other intelligent optimization algorithms(HHO,PSO,GWO,WOA);4.To further improve the model classification performance,the GA＿FS＿SVM and BGOHHO＿SVM models were fused,and the GA＿BGOHHO＿SVM fusion model was proposed and experiments were conducted on the type 2 diabetes Chinese medical evidence dataset.Results:1.the final experimental results are compared with the chi-square test,F-test and mutual information in the filtered feature selection algorithm.The experimental results show that GA＿FS＿SVM can effectively select the best feature subset and improve the classification performance of the model.2.The BGOHHO algorithm is tested against the Harris Hawk algorithm(HHO),Particle Swarm Optimization algorithm(PSO),Whale optimization algorithm(WOA)and Grey Wolf Optimizer algorithm(GWO)on 16 benchmark test functions,and the experimental results show that the improved The experimental results show that the improved BGOHHO algorithm has a better search capability.3.A comparison study with SVM optimized by other intelligent optimization algorithms(HHO,PSO,GWO,WOA)was conducted to evaluate the model performance by evaluation metrics,and the experimental results show that the SVM optimized by using BGOHHO achieves a better classification performance.4.A comparison study with GA＿FS＿SVM and BGOHHO＿SVM model.The performance of the model was evaluated by means of evaluation metrics and the experimental results showed that the fusion model obtained the best classification performance.Conclusion:In this study,feature selection based on genetic algorithm is performed on the type 2 diabetes Chinese medical evidence dataset,which improves the classification performance of the model while reducing redundant features.In order that the Harris Hawk Optimizer algorithm can be better applied to the parameter search of the model and further improve the classification performance of the model,various strategies are proposed to improve it,and experiments are conducted on the standard test function to obtain the best search performance compared with the classical intelligent Optimizer algorithm,indicating that the improvement strategies proposed in this study can effectively help the Harris Hawk Optimizer algorithm to jump out of the local optimum.The final fusion model,which obtains the optimal classification accuracy,fully demonstrates that the feature selection and parameter optimization experiments designed by this study based on the intelligent optimizer algorithm can effectively improve the classification performance of the model.