PREDICTING ANEMIA FOR PREGNANT WOMEN USING DIFFERENT MACHINE LEARNING ALGORITHMS

Abstract

Anemia appears as a common medical condition which affects pregnant women internationally since it produces serious health issues for mothers and their developing babies. The main cause of anemia comes from red blood cell and hemoglobin deficiencies specifically related to iron deficiency while vitamin B12 deficiency and chronic diseases contribute to its progression. Tangible methods for early diagnosis matter to deliver proper management of pregnancy-related issues yet they prove hard to achieve in areas with limited resources. The study tackles the requirement for advanced anemia in pregnancy predictive methods by implementing machine learning algorithms to strengthen diagnostic performance. A team compiled the dataset at sulaimaniya maternity hospital which contained socio-demographic information and medical histories and blood test results from pregnant females. The collected dataset presented data on age as well as occupation and gestational age together with dietary habits and chronic diseases and blood parameters measured as hemoglobin and iron levels. The study optimizes model performance through the use of multiple feature selection methods including PCA, SelectKBest, Chi-Square and RFE. The study trained and evaluated eight machine learning models which included Random Forest (RF), Gradient Boosting (GB), AdaBoost, Extra Trees, Support Vector Machines (SVM), Logistic Regression (LR), K-Nearest Neighbors (KNN) and Decision Trees (DT) and Naive Bayes (NB). Each model received evaluation through assessments of accuracy, precision and recall and F1-score measures. The integration of feature selection led to substantially better model performance because RF, Extra Trees, SVM and LR demonstrated near-perfect results. The performance of KNN remained significantly lower than other models because it demonstrated weak abilities in selecting features. Among the evaluated algorithms SelectKBest and Chi-Square along with RFE demonstrated superior performance than PCA for obtaining 100% accuracy levels. This research shows that machine learning systems have potential in predicting anemia in pregnant women thus promoting improved diagnostic strategies in underprivileged healthcare settings. This research utilizes maternity hospital information to demonstrate the advantages of data-based approaches in maternal healthcare and presents machine learning as a tool for better healthcare results