| Title |
A Study on the Estimation of Hemoglobin based on Regression using Physical and Health Information |
| Authors |
홍상훈(Sang-Hoon Hong) ; 홍광석(Kwang-Seok Hong) |
| DOI |
https://doi.org/10.5573/ieie.2021.58.9.42 |
| Keywords |
Hemoglobin; non-invasive measurement; regression analysis; big data |
| Abstract |
In this paper, we propose a regression analysis based hemoglobin estimation method using body (height, weight, and waist circumference) and health information (diastolic blood pressure, triglyceride and ALT(SGPT), etc.) included in health examination big data. Prior studies have confirmed the possibility of estimating hemoglobin using physical and health information, but it has limitations such as the model was created and evaluated using train data without verification through test data, and used 1 or 7 combinations of independent variables. In this paper, to compensate for the above problems, statistically analyzed the correlation with hemoglobin using big data of medical examinations containing a total of 4,026,293 physical and health information. and Linear and multiple regression analyses were performed to estimate hemoglobin for segmentation classes based on the total number of people and features (gender, age group and smoking state). Furthermore, multiple regression analyses using two to seven independent variables were performed and the results were compared. Experiments show that the lowest error rate was calculated at 6.149% and 6.010%, respectively, when linear and multiple regression analyses were performed for classes subdivided into three features (gender, age group and smoking state). |