改进的GA-SVM算法在继发性肾炎诊断中的研究与应用

发布时间：2018-05-02 07:23

本文选题：狼疮性肾炎 + 遗传算法　；参考：《昆明理工大学》2017年硕士论文

【摘要】：系统性红斑狼疮(Systemic Lupus Erythematosus,SLE)是一种病因未明的自身免疫性疾病,该疾病一般伴着多种多样的自身抗体且以多个系统损害共同形成的。在确诊的SLE患者中约有80%的患者患有狼疮性肾炎(Lupus Nephritis,LN)。LN是SLE继发性疾病中发病率最高的,同时由于其发病原因复杂多样、影响因素众多而变得非常难以确诊。因此,对LN疾病的诊断具有非常重要的医学意义。医学上有两种诊断方法,一种是通过诊断标准进行确诊,一种是通过病症表现的积分进行辅助诊断,但是这两种方法都存在一些弊端,前一种方法容易出现误诊和漏诊,后一种方法使用起来比较麻烦,工作量太大。因此本文用了一种基于支持向量机(Support Vector Machine,SVM)模型的智能算法诊断方法。该方法通过搜集处理大量的LN疾病相关数据,把不同的病例分别用不同的数字表示,通过SVM模型进行智能预测,最终达到辅助诊断的目的。但是由于SVM算法在实际应用过程中,例如惩罚参数的选择和核参数的选择方面也存在着某些不足。因此,本文提出了基于遗传算法(Genetic Algorithm,GA)优化SVM参数和核函数参数的模型(即GA-SVM模型)和一种基于(c,g,ε)的GA-SVM优化算法模型在LN中的应用,进而在此基础上提出了一种改进的GA-SVM优化模型。使用三种智能算法分别对同一数据集进行仿真预测,对比三种模型的仿真结果,基于改进的GA-SVM算法在误差值、均方误差和线性相关.度方面都取得了很大的改进,更加适用于LN疾病的诊断预测。
[Abstract]:Systemic lupus erythematosus (systemic lupus erythematosus) is an unknown autoimmune disease. About 80% of SLE patients with lupus nephritis have the highest incidence of SLE secondary diseases. At the same time, it is very difficult to diagnose the disease because of its complex causes and many influential factors. Therefore, the diagnosis of LN disease has very important medical significance. There are two diagnostic methods in medicine, one is to make a diagnosis by diagnostic criteria, the other is to make an auxiliary diagnosis through the integral of the manifestation of the disease. However, there are some drawbacks in the two methods. The former method is prone to misdiagnosis and missed diagnosis. The latter method is more troublesome to use, the workload is too large. Therefore, an intelligent algorithm diagnosis method based on support vector machine support Vector machine model is used in this paper. By collecting and processing a large number of LN disease related data, the different cases are represented by different numbers, and the intelligent prediction is carried out by SVM model. Finally, the purpose of auxiliary diagnosis is achieved. However, SVM algorithm also has some shortcomings in practical application, such as the selection of penalty parameters and kernel parameters. Therefore, this paper presents a model (GA-SVM model) based on genetic algorithm (GA) to optimize the parameters of SVM and kernel function (i.e., GA-SVM model) and a model of GA-SVM optimization algorithm based on genetic algorithm (蔚) in LN. Then an improved GA-SVM optimization model is proposed. Three intelligent algorithms are used to predict the same data set, and the simulation results of the three models are compared. The improved GA-SVM algorithm is based on error, mean square error and linear correlation. Great improvements have been made in degree, which is more suitable for the diagnosis and prediction of LN disease.
【学位授予单位】：昆明理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：R692.3;TP18

【参考文献】