Predicting preterm birth using machine learning methods.

Area of Science:

• Obstetrics and Gynecology
• Medical Informatics
• Computational Biology

Background:

• Preterm birth is a major cause of neonatal mortality and morbidity.
• The causes of preterm birth are complex and multifactorial.
• Accurate prediction of preterm birth risk is crucial for timely intervention.

Purpose of the Study:

• To develop and compare machine learning models for predicting preterm birth risk.
• To evaluate the efficacy of various algorithms including XGBoost, CatBoost, logistic regression, SVM, and decision trees.
• To identify the most effective model for preterm birth risk assessment.

Main Methods:

• Data from 50 maternity ward patients were analyzed.
• Machine learning models were trained to predict delivery timing (preterm vs. term).
• Performance metrics including accuracy, precision, recall, and F1-score were used for comparison.

Main Results:

• Linear Support Vector Machines (SVM) with boosted parameters achieved the highest performance (82% accuracy, 83% precision, 86% recall, 84% F1-score).
• Boosted logistic regression showed comparable results (80% accuracy, 82% precision, 82% recall, 82% F1-score).
• Other models, including decision trees, performed less effectively, potentially due to dataset size.

Conclusions:

• Machine learning models, particularly linear SVM, are effective for assessing preterm birth risk.
• The linear SVM model demonstrated the highest efficacy among the tested algorithms.
• These findings support the use of machine learning for improving preterm birth prediction and management.