基于RF和KNN的三种肝炎分类模型的建立

The establishment of three kinds of hepatitis of classification model based on the RF and KNN

目的建立甲、乙、丙三种肝炎的计算机分类模型。方法以病毒学检查为指标确定三种肝炎病例,以患者和健康体检者血常规检查指标、生化检查指标为原始数据建立数据库,其中甲肝病例186例,乙肝病例835例,丙肝病例129例,健康志愿者438人。分别采用随机森林和K-最邻近法建立甲、乙、丙三种肝炎的分类模型。结果随机森林筛选出了9个(ALT、GGT、AST、ALB、BUN/Crea、CPT、MO%、TBIL、Cl-1)相对重要的变量,该模型内部预测准确率、测试集的预测准确率分别是92.59%、91.56%,KNN模型内部预测准确率、训练集、测试集的预测准确率分别是93.95%、96.89%、90.23%。结论所建的分类模型对三种肝炎患者和健康人有较好的识别能力。

Objective To establish the computer classification model of three kinds of hepatitis A, B and C. Methods Three hepatitis cases were determined by using virology examination. Database was set up using blood routine examination and biochemical in- dexes of patients and healthy people as the original data. There were 186 cases of hepatitis A,835 cases of hepatitis B, 129 cases of hepa- titis C and 438 healthy volunteers. Three kinds of hepatitis A, B, C of classification model were respectively established by using random forest method and K - nearest neighbor method. Results Nine of relative important variables （ALT,GGT,AST,ALB,BUN/Crea,CPT, MO% , TBIL, C1 - 1 ） were screened out using random forest. The internal prediction accuracy of the model was 92.59%. The forecast ac- curacy of the test set was 91.56%. The internal prediction accuracy of KNN model was 93.95%. The forecast accuracy of the training set was 96.89% and the forecast accuracy of the test set was 90.23%. Conclusion The establishment of the classification model has better recognition ability on patients of three kinds of hepatitis and healthy people.