An Application of Machine Learning to Thalassemia Diagnosis  

作　　者：Sitan Liu Sitan Liu(School of Mathematics and Statistics, Guilin University of Technology, Guilin, China;Guangxi Colleges and Universities Key Laboratory of Applied Statistics, Guilin, China)

机构地区：[1]School of Mathematics and Statistics, Guilin University of Technology, Guilin, China [2]Guangxi Colleges and Universities Key Laboratory of Applied Statistics, Guilin, China

出　　处：《Journal of Computer and Communications》2024年第2期211-230,共20页电脑和通信（英文）

摘　　要：Mediterranean anemia is a genetic disease that currently relies heavily on expert clinical experience to determine whether patients are affected. This method is overly reliant on expert experience and is not precise enough. This paper proposes two modeling methods to predict whether patients have Mediterranean anemia. The first method involves using Principal Component Analysis (PCA) to reduce the dimensionality of the data, followed by logistic regression modeling (PCA-LR) on the reduced dataset. The second method involves building a Partial Least Squares Regression (PLS) model. Experimental results show that the prediction accuracy of the PCA-LR model is 87.5% (degree = 2, λ=4), and the prediction accuracy of the PLS model is 92.5% (ncomp = 4), indicating good predictive performance of the models.Mediterranean anemia is a genetic disease that currently relies heavily on expert clinical experience to determine whether patients are affected. This method is overly reliant on expert experience and is not precise enough. This paper proposes two modeling methods to predict whether patients have Mediterranean anemia. The first method involves using Principal Component Analysis (PCA) to reduce the dimensionality of the data, followed by logistic regression modeling (PCA-LR) on the reduced dataset. The second method involves building a Partial Least Squares Regression (PLS) model. Experimental results show that the prediction accuracy of the PCA-LR model is 87.5% (degree = 2, λ=4), and the prediction accuracy of the PLS model is 92.5% (ncomp = 4), indicating good predictive performance of the models.

关 键 词：MULTICOLLINEARITY Statistical Analysis Models Data Mining PCA-LR PLS 

分 类 号：R73[医药卫生—肿瘤]