Hybrid Rf-Pso for predicting premature birth during pregnancy a case study of Dr M Ansari Saleh Hospital Banjarmasin

Preterm birth (PTB) occurs before 37 weeks of pregnancy. Although many factors influence the course of pregnancy, the actual cause of PTB remains unknown, and current medical efforts focus more on reducing the effects rather than preventing them. Therefore, a technology-based approach is needed to predict essential characteristics and develop a prediction model using machine learning algorithms. This study aims to identify the characteristics of PTB and develop a prediction process model using machine learning. Data was collected from Hospital Dr M. Ansari Saleh in Banjarmasin between 2020 and 2022, involving 915 samples. Factors include maternal factors, demographics, nutritional status, and current pregnancy, which are assessed using a machine- learning algorithm. To determine the best accuracy, this study tested the algorithm's performance with data division ratios of 90/10, 80/20, 70/30, and 60/40. Optimum results were obtained with 80/20 division, then optimized using Particle Swarm Optimization (PSO) and Cross-Validation. The Random Forest algorithm chosen for this study achieved the highest accuracy of 96.45%. Important features identified include the history of abortion, a history of Cesarean surgery (CS), and parity. This research contributes to health informatics by developing an exact PTB prediction model using the Random Forest-PSO hybrid technique. The implication of this study, the model improves health services and can help reduce infant and child deaths due to PTB in Indonesia.