基于血常规支持向量机应用于肾病模型的预测

Support vector machine to identify nephropathy based on blood routine test

文章利用主成分分析对肾病患者的血常规检测数据进行特征提取,采用支持向量机对肾病患者样本、虚拟正常人样本及其他疾病患者样本建立分类模型。采用高斯径向基函数为核函数,调节核函数参数C及σ以建立最佳支持向量机模型,并且采用留一交叉验证法从虚拟正常人样本中区分出来的识别率为91.35%,而用十折交叉验证法得到该模型从其他疾病中区分出来的识别率为81.71%。结果表明,基于血常规支持向量机法建立的肾病分类模型能较好的把肾病从其他疾病和正常人样本中区分出来。

Principal component analysis （PCA） is used to extract features of blood routine testing results of patients with nephropathy, and the classified model of nephropathy patients and the normals is established based on support vector machine （SVM）. The radical basis function is adopted as a kernel function of SVM, and the model adjusts C and a to build the optimal classifier. The correct classification ratio reaches 91.35% by left-one-out cross validation from normal persons and 81.71% by I 0-fold cross validation from patients with other diseases. The results show that the classified model of blood routine based on PCA-SVM can classify the nephropathy from the other diseases and normal persons.