Research On Diagnosis Method Of Diabetes Based On Machine Learning

Due to the improvement of residents’ living standards and people’s long-term bad living habits,the incidence of diabetes is increasing day by day.The powerful processing and decision-making capabilities of machine learning are conducive to assist in the establishment of corresponding diagnostic models and provide a new way for diabetes diagnosis.Therefore,this paper uses machine learning algorithms to establish a diabetes diagnosis fusion model.The specific work content is as follows:(1)In order to improve the performance of the SVM classification model,a support vector machine parameter optimization model(EOPCALO-SVM)based on the improved Antlion algorithm is proposed.First,the original ant lion optimization algorithm(ALO)is optimized by introducing disturbance factors,Logistic chaotic mapping and elite reverse learning,and proposes an elite reverse learning chaotic ant lion algorithm with disturbance factors(EOPCALO)with better convergence performance.);Secondly,use EOPCALO to optimize the combination of SVM penalty factor and nuclear parameters,and propose the EOPCALO-SVM model;finally,experiments are carried out on 8 data sets of actual diabetes,breast cancer,Wine,etc.,and the classification accuracy rate is as high as 96.91%,Which proves that EOPCALO-SVM has high classification performance and generalization ability.(2)In order to reduce the feature dimension of the data set,a feature selection method based on an improved butterfly optimization algorithm(IBOA-FS)is proposed.First,in order to improve the global and local search capabilities of the algorithm,the original butterfly optimization algorithm(BOA)introduces Logistic-Cubic cascaded chaotic mapping,nonlinear convergence factors and Sigmoid function-based constraint factors to improve;secondly,the butterfly optimization will be improved The algorithm is applied to the feature selection method of the data set;finally,a comparative experiment is carried out on 8 data sets,and a better feature subset can be obtained under the premise of ensuring that the classification accuracy rate is not reduced,which proves that IBOA-FS can effectively select The best feature subset and the improvement of classification performance provide the basis for the subsequent fusion model of diabetes diagnosis.(3)In order to construct the best diagnosis model of diabetes,a fusion model of diabetes diagnosis(IBA-FS-SVM)combining IBOA-FS and EOPCALO-SVM is proposed.First,perform data preprocessing on 8 data sets;secondly,compare the feature subsets and classification accuracy of IBA-FS-SVM with the same type of algorithms on 8 data sets;finally,the experimental results show that it is compared with EOPCALO-SVM,the classification accuracy of the IBA-FS-SVM model is up to99.10%,which proves that the IBA-FS-SVM diabetes diagnosis fusion model has certain advantages and generalization capabilities.