基于参数优化的SVM分类器在继发性干燥综合征诊断中的应用

发布时间：2018-08-24 16:10

【摘要】：干燥综合征.(siccasyndrome,简称SS)是一种外分泌腺体的自身免疫性慢性疾病,通常被分成原发性干燥综合征与继发性干燥综合征这两个类型。其中,继发性干燥综合征常继发于系统性红斑狼疮(Systemic Lupus Erythematosus,简称SLE)等疾病之中,因此容易被人忽视。为了解决SLE患者并发继发性干燥综合征不容易及时确诊与治疗过程中由于医生主观性依赖较强导致治疗方案有偏差的问题,提出了一种将支持向量机(SupportVectorMachine,简称SVM)与系统性红斑狼疮继发干燥综合征早期诊断相结合的新思路,该方法主要思想是利用SVM分类器对系统性红斑狼疮患者及SLE继发性干燥综合征的患者进行二分类。本文以141例患者病例为研究对象,经过数据筛选处理,分别运用了交叉验证法、网格搜索法、标准粒子群优化算法分别对SVM模型分类器中的惩罚参数C与核函数参数g进行优化选择,最终发现粒子群算法优化SVM参数模型不仅分类的效果最佳,其泛化能力也有很大提升。之后针对粒子群算法易陷入局部最优的问题,提出混沌机制改进粒子群优化算法的方法。最后,分别利用MATLAB软件对上述4种优化方式进行编程的实现,并将结果以直观的图像表示在workspace窗口中,用以表达SVM模型分类器最终的分类正确率。最终对比选出对SLE患者并发继发性干燥综合征疾病诊断分类度的准确率分别为82.3529%、88.2353%、90.1961%、92.1569%。最终结果对比表明:基于混沌机制改进的粒子群算法优化的支持向量机分类模型参数的寻优,相较于交叉验证与网格搜索法,对SVM参数的优化选择更加科学及严谨;且对比标准粒子群优化SVM参数的模型,可以明显看出,基于混沌机制改进的粒子群算法改善了标准粒子群容易陷入早熟现象,从而提高了 SVM分类器对SLE继发干燥综合征疾病分类诊断的精度。
[Abstract]:Sjogren syndrome (. (siccasyndrome,) is an autoimmune chronic disease of exocrine glands, which is usually divided into primary Sjogren syndrome and secondary Sjogren syndrome. Among them, secondary Sjogren syndrome is often secondary to systemic lupus erythematosus (Systemic Lupus Erythematosus,) and other diseases, so it is easy to be ignored. In order to solve the problem that the patients with SLE complicated with secondary Sjogren's syndrome are not easily diagnosed in time and deviated from the treatment plan due to the strong subjective dependence of the doctor. A new idea of combining support vector machine (SVM) with early diagnosis of Sjogren's syndrome secondary to systemic lupus erythematosus is proposed. The main idea of this method is to use SVM classifier to classify patients with systemic lupus erythematosus and patients with SLE secondary Sjogren's syndrome. In this paper, 141 cases of patients were studied. After data screening and processing, cross validation method and grid search method were used respectively. The standard particle swarm optimization algorithm optimizes the penalty parameter C and kernel function parameter g in the classifier of SVM model. Finally, it is found that the particle swarm optimization algorithm not only has the best classification effect, but also greatly improves its generalization ability. Then, aiming at the problem that particle swarm optimization (PSO) is easy to fall into local optimization, a chaotic mechanism is proposed to improve PSO. Finally, the above four optimization methods are programmed by using MATLAB software, and the results are represented in the workspace window as an intuitive image, which is used to express the final classification accuracy of the SVM model classifier. The accuracy of diagnosis of secondary Sjogren's syndrome in SLE patients was 82.3529 and 88.2353, 90.1961and 92.1569, respectively. The final results show that the optimization of support vector machine (SVM) model parameters based on improved particle swarm optimization (PSO) based on chaos mechanism is more scientific and rigorous than cross-validation and mesh search. Compared with the model of standard particle swarm optimization (SVM) parameter optimization, it is obvious that the improved particle swarm algorithm based on chaos mechanism can improve the premature phenomenon of standard particle swarm optimization. Thus, the accuracy of SVM classifier in the diagnosis of Sjogren's syndrome secondary to SLE is improved.
【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】：R593.2;TP18

【参考文献】