EXAMINING RISK FACTORS OF ANEMIA IN PREGNANCY USING HYBRID LOGISTIC REGRESSION MODEL AND ROUGH SET THEORY

Abstract

Anemia in pregnancy is a potential danger to the mother and child. Therefore, the risk of anemia in pregnant women requires serious attention from all relevant parties. Considering the numerous negative effects caused by anemia in pregnant women, efforts must be made to prevent and treat anemia in pregnant women by understanding the factors that influence it. This study assesses the risk factors for anemia in pregnant women at Tegal Rejo Community Health Center, Yogyakarta Province. In this paper, a new integrated classification approach with binary logistic regression (LR) analysis and Rough Set Theory (RST) is proposed, in order to examine factors on the incidence of anemia in pregnancy. The proposed model is called the Logistic Regression and Reduction Rough Set (LR3S). In LR3S model, the RST technique is used to detect inconsistent sample and removing inconsistent sample that have probability less than 0.5 before doing LR modelling. To evaluate the development of the resulting model, a comparison was made of the performance of Original Logistic Regression (OLR), LR model after removed outlier namely as Remove Outlier Original Logistic Regression (RO2LR) and LR3S. Using a number of model performance metrics, it is found that LR3S has the best performance for the three models used. Using LR3S model, it is found that CED status, educational level, parity and gestational are significant variable impact on the incidence of anemia.