基于PSO-SVM的糖尿病并发症预测研究

目的:运用数学模型粒子群优化(PSO)算法与支持向量机(SVM)(PSO-SVM)评估糖尿病并发症的发病风险,为糖尿病的临床诊断提供有效的数据信息.方法:以SVM为人工智能算法,通过PSO算法对其参数进行优化,利用K-fold交叉验证法将部分数据用于模型的训练,建立以16项指标数据作为输入变量、以糖尿病肾病、糖尿病性视...

Study on the prediction of the complication of diabetes mellitus based on PSO-SVM

Objective:To assess the onset risk of the complication of diabetes mellitus by using the mathematical model particle swarm optimization(PSO)algorithm and support vector machine(SVM)(PSO-SVM)so as to provide effective information for the adjuvant therapy of clinical diagnosis of diabetes mellitus.Methods:SVM was used as artificial intelligence algorithm,and its parameters were optimized by PSO algorithm,and K-fold cross-validation method was adopted to use part of data for training model,so as to established a prediction model of diagnosis for complication of diabetes mellitus which used 16 index data as input variables,and used 3 kinds of complications included diabetic nephropathy,diabetic retinopathy and nervous lesion as output variables.Results:After the test data were inputted into prediction model for diagnosing diabetic complication post training,the accuracy of the classification of test set,which was finally obtained,was 84%.Conclusion:The established prediction model of diagnosis based on PSO-SVM has a favorable auxiliary effect on the clinical diagnosis for diabetic complications.