Research On The Prediction Method Of Diabetic Complications Based On Feature Selection And Weighted Fusion Classification

Diabetes is one of the most dangerous chronic diseases at present.With the prolongation of diabetes patients,it can damage other organs of the body and lead to many serious complications.In general,complications will cause huge physical problems to the patients,and even endanger the patients’ life safety.In order to detect and prevent complications in diabetic patients earlier and diagnose complications with as high accuracy as possible,this study proposes a predictive method for diabetic complications based on feature selection and weighted fusion classification.Firstly,in order to reduce the dataset dimension,this paper proposes a feature selection method that combines flexible mutual information and recursive feature elimination to select the optimal feature subset from the original dataset.In addition,missing values and class balance were also handled for the data in the dataset.Regarding missing values in data,a random forest iterative imputation method that can be populated for different data types is introduced.To balance the classes of the dataset,adaptive synthesis techniques are used to increase minority class samples.Secondly,considering the limited comprehensive ability of a single classification algorithm to predict complex diabetic complications datasets,this paper proposes a classification algorithm based on a weighted fusion of neural networks and gradient boosting.The deep table learning classification method and the classification gradient boosting classification method are selected as the base classifiers,and the two base classifiers are trained in parallel through the training set.Use the test set to weight the predicted AUC values of the two base classifiers to obtain the final prediction result and improve the performance of classification prediction.Combining the weighted fusion method with the data processing method,a prediction method of diabetes complications based on feature selection and weighted fusion classification was constructed.Finally,on the diabetic complications dataset,the proposed feature selection method and the prediction method of weighted fusion classification are experimentally verified and the results are analyzed based on the two complications of diabetic nephropathy and cardiovascular disease.Through the prediction of diabetic nephropathy and cardiovascular disease,and through experimental comparison,it is verified that the method proposed in this paper has better accuracy and stability in the prediction of diabetic complications.