Precision in Obstetric Care - A Machine Learning Approach with CatBoost and Grid Search Optimization

Aug 3, 2026

Upload by Noah

This study improves fetal health classification using machine learning by fine-tuning CatBoostClassifier through Grid Search optimization. The work focuses on accurate diagnosis from Cardiotocogram (CTG) data, producing substantially higher predictive performance and enabling more consistent prenatal care decisions. Results show an overall accuracy of 96%, with strong identification of Normal and Pathological cases. Suspect-case classification remains challenging, indicating opportunities for further refinement. Optimized hyperparameters also yield the lowest loss and best generalization.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages7

Reading Time18 min

AccessFree to download

Abstract

I. Introduction

Fetal health monitoring and CTG

Challenges of CTG interpretation

Prior machine learning approaches and limitations

Study objective and methodology overview

Precision in Obstetric Care: A Machine Learning Approach with CatBoost and Grid Search Optimization Marselina Endah Hiswati1, Mohammad Diqi2*, Izattul Azijah3, Yeyen Subandi4, Azzah Fathinah5, Rahayu Cahya Ariani6 1,2,5Department of Informatics, Universitas Respati Yogyakarta, DI Yogyakarta, Indonesia 3Department of Midwifery, Universitas Respati Indonesia, DKI Jakarta, Indonesia 4Department of International Relations, Universitas Respati Yogyakarta, DI Yogyakarta, Indonesia 6Department of Public Health, Universitas Respati Yogyakarta, DI Yogyakarta, Indonesia [Email:](Email:1marsel.endah@respati.ac.id)[1](Email:1marsel.endah@respati.ac.id)[marsel.endah@respati.ac.id](Email:1marsel.endah@respati.ac.id),2*[diqi@respati.ac.id](diqi@respati.ac.id), [3](3iza@urindo.ac.id)[iza@urindo.ac.id](3iza@urindo.ac.id), [4](4yeyensubandi@respati.ac.id)[yeyensubandi@respati.ac.id](4yeyensubandi@respati.ac.id), [5](522220025@respati.ac.id)[22220025@respati.ac.id](522220025@respati.ac.id), [6](6225050016@urindo.ac.id)[225050016@urindo.ac.id](6225050016@urindo.ac.id) (Received: 3 Aug 2024, revised: 27 Aug 2024, accepted: 28 Aug 2024) Abstract This study focuses on improving how we classify fetal health using machine learning by fine-tuning the CatBoostClassifier with Grid Search. Our main achievement in this research is significantly boosting the accuracy of fetal health classification based on Cardiotocogram (CTG) data. Finding the best hyperparameters has created a more precise and reliable diagnostic tool for making informed prenatal care decisions. The model reached an impressive overall accuracy of 96%, especially excelling in identifying Normal and Pathological cases. However, it faced some challenges in classifying Suspect cases, suggesting room for further improvement. These results highlight the potential of machine learning to enhance the reliability of fetal health assessments , which could lead to better outcomes in clinical settings. The success of Grid Search in this study is evident, as the optimized parameters led to the highest accuracy and lowest loss values, proving its effectiveness in fine-tuning the model. Keywords: CatBoostClassifier, Fetal Health Classification, Grid Search Optimization, Machine Learning in Obstetrics, Diagnostic Accuracy. I. INTRODUCTION Fetal health monitoring is a critical step in prenatal care, aimed at ensuring the safety and health of both the mother and fetus during pregnancy [1] . One of the most commonly used diagnostic tools for monitoring fetal health is the Cardiotocogram (CTG), which measures fetal heart rate and uterine contractions [2] . The use of CTG enables doctors to assess whether the fetus is experiencing stress or oxygen deprivation, which may require immediate medical intervention [3] . Therefore, the ability to accurately interpret CTG results is essential for making correct and timely clinical decisions [4] . However, the interpretation of CTG can be subjective and vary among observers, which can affect the consistency and accuracy of diagnoses [5] . Incorrect or delayed diagnoses can have fatal consequences, such as premature birth or even fetal death [5]. Conversely, accurate and prompt diagnoses can save lives and reduce the risk of long-term complications for both mother and child [2] . In this context, the development of predictive models that can automatically classify the health status of the fetus based on CTG data has the potential to improve health outcomes by providing a more objective and consistent diagnostic tool [3] . Machine learning models, especially those designed to handle complex and categorical data, offer a promising way to overcome these challenges. Although past studies have tried using machine learning algorithms like Random Forest, Support Vector Machines (SVM), and Neural Networks to analyze CTG data, these methods come with their own set of limitations [6] . For instance, while Random Forest and SVM can be accurate in some sit

FAQs

How does the study improve fetal health classification from CTG data?

It fine-tunes CatBoostClassifier using Grid Search to optimize hyperparameters, improving classification performance on CTG features.

What dataset and classification categories are used?

The study uses a public CTG dataset with 2,113 records and three fetal health categories: Normal, Suspect, and Pathological.

What performance results does the optimized model achieve?

The model reaches an overall accuracy of 96%, performing especially well on Normal and Pathological cases, while Suspect-case classification is less effective.

Aug 3, 2026

Upload by Noah