A Machine Learning-Based Decision Support System Design to Predict Maternal Health Risk Groups During Pregnancy

İrem Şenyer Yapıcı; Rukiye Uzun Arslan

Article Details

A Machine Learning-Based Decision Support System Design to Predict Maternal Health Risk Groups During Pregnancy

Authors:

İrem Şenyer Yapıcı | Rukiye Uzun Arslan

Publisher:

Karyay Karadeniz Yayımcılık

Main Topics:

Karar Desteği ve Grup Destek Sistemleri | Bilgi Sistemleri (Diğer)

Journal:

Black Sea Journal of Engineering and Science Vol.7 Issue 3, pp.509-520

Document Type:

Article

Publication Date:

5/2024

Document Language:

Turkish

Document ID (Org No.):

2001315152

pISSN:

2619-8991

Keywords:

Abstract:

 Early detection of maternal health risks during pregnancy and appropriate interventions are vital for maternal and infant health. In this process, the use of machine learning ML algorithms that can automatically analyze complex relationships and patterns from large datasets plays a critical role. By revealing hidden information in large data sets, ML algorithms offer the opportunity to more accurately identify pregnancy-related risk factors. In this context, this study focuses on increasing the success of predicting maternal health risk levels during pregnancy, especially in high-risk pregnancies. For this purpose, firstly, important critical features that will increase the success are identified and the most effective computer-based decision support system is designed by using six different machine learning algorithms. By applying the Chi-Square test in combination with the SelectKBest method, it was determined that the most critical characteristics in the dataset were age, systolic blood pressure and diastolic blood pressure. In addition, synthetic minority oversampling technique SMOTE was utilized to address the imbalance in the dataset. The performance of the ML algorithms used in the proposed model is analyzed by hold-out performance evaluation method. In the light of the findings, it was determined that the use of the SMOTE technique has a positive effect on improving model performance in predicting risk levels in pregnancy. In the proposed model, the highest classification performance for each classification algorithm was obtained for the high risk class. Among the algorithms used, the extreme gradient boosting algorithm was found to have the best performance with a 97 success rate. Overall, the results confirm that the proposed model is highly effective in detecting high-risk pregnancies. This finding shows that the proposed ML-based decision support system has the potential to provide significant support to specialist physicians in making more accurate diagnoses and performing the necessary interventions more quickly during pregnancy.

 Early detection of maternal health risks during pregnancy and appropriate interventions are vital for maternal and infant health. In this process, the use of machine learning ML algorithms that can automatically analyze complex relationships and patterns from large datasets plays a critical role. By revealing hidden information in large data sets, ML algorithms offer the opportunity to more accurately identify pregnancy-related risk factors. In this context, this study focuses on increasing the success of predicting maternal health risk levels during pregnancy, especially in high-risk pregnancies. For this purpose, firstly, important critical features that will increase the success are identified and the most effective computer-based decision support system is designed by using six different machine learning algorithms. By applying the Chi-Square test in combination with the SelectKBest method, it was determined that the most critical characteristics in the dataset were age, systolic blood pressure and diastolic blood pressure. In addition, synthetic minority oversampling technique SMOTE was utilized to address the imbalance in the dataset. The performance of the ML algorithms used in the proposed model is analyzed by hold-out performance evaluation method. In the light of the findings, it was determined that the use of the SMOTE technique has a positive effect on improving model performance in predicting risk levels in pregnancy. In the proposed model, the highest classification performance for each classification algorithm was obtained for the high risk class. Among the algorithms used, the extreme gradient boosting algorithm was found to have the best performance with a 97 success rate. Overall, the results confirm that the proposed model is highly effective in detecting high-risk pregnancies. This finding shows that the proposed ML-based decision support system has the potential to provide significant support to specialist physicians in making more accurate diagnoses and performing the necessary interventions more quickly during pregnancy.

 Early detection of maternal health risks during pregnancy and appropriate interventions are vital for maternal and infant health. In this process, the use of machine learning ML algorithms that can automatically analyze complex relationships and patterns from large datasets plays a critical role. By revealing hidden information in large data sets, ML algorithms offer the opportunity to more accurately identify pregnancy-related risk factors. In this context, this study focuses on increasing the success of predicting maternal health risk levels during pregnancy, especially in high-risk pregnancies. For this purpose, firstly, important critical features that will increase the success are identified and the most effective computer-based decision support system is designed by using six different machine learning algorithms. By applying the Chi-Square test in combination with the SelectKBest method, it was determined that the most critical characteristics in the dataset were age, systolic blood pressure and diastolic blood pressure. In addition, synthetic minority oversampling technique SMOTE was utilized to address the imbalance in the dataset. The performance of the ML algorithms used in the proposed model is analyzed by hold-out performance evaluation method. In the light of the findings, it was determined that the use of the SMOTE technique has a positive effect on improving model performance in predicting risk levels in pregnancy. In the proposed model, the highest classification performance for each classification algorithm was obtained for the high risk class. Among the algorithms used, the extreme gradient boosting algorithm was found to have the best performance with a 97 success rate. Overall, the results confirm that the proposed model is highly effective in detecting high-risk pregnancies. This finding shows that the proposed ML-based decision support system has the potential to provide significant support to specialist physicians in making more accurate diagnoses and performing the necessary interventions more quickly during pregnancy.

 Gebelik döneminde anne sağlığı risklerinin erken tespiti ve uygun müdahalelerin yapılması, anne ve bebek sağlığı açısından hayati bir önem taşımaktadır. Bu süreçte, büyük veri kümelerinden elde edilen karmaşık ilişkileri ve desenleri otomatik olarak analiz edebilen makine öğrenme MÖ algoritmalarının kullanımı son derece kritik bir rol oynamaktadır. MÖ algoritmaları, büyük veri setlerindeki gizli bilgileri açığa çıkararak, gebelikle ilişkili risk faktörlerini daha doğru bir şekilde belirleme imkanı sunmaktadır. Bu bağlamda gerçekleştirilen bu çalışmada, gebelik sürecinde anne sağlığı risk seviyelerinin özellikle yüksek riskli hamileliklerin tahmininde başarının arttırılmasına odaklanılmıştır. Bunun için öncelikle başarımı artıracak önemli kritik özellikler belirlenmiş ve altı farklı makine öğrenme algoritması kullanılarak en etkili bilgisayar temelli karar destek sistemi tasarlanmaya çalışılmıştır. Ki-Kare testi SelectKBest yöntemiyle birlikte uygulanarak, veri setindeki en kritik özelliklerin yaş, sistolik kan basıncı ve diyastolik kan basıncı olduğu tespit edilmiştir. Yanı sıra veri setindeki dengesizliği gidermek için sentetik azınlık aşırı örnekleme tekniğinden SMOTE yararlanılmıştır. Önerilen modelde kullanılan MÖ algoritmalarının başarımları hold-out performans değerlendirme yöntemiyle analiz edilmiştir. Elde edilen bulgular ışığında, SMOTE tekniğinin kullanılmasının gebelikteki risk seviyelerinin tahmininde model başarımlarını artırmada olumlu bir etkiye sahip olduğu belirlenmiştir. Önerilen modelde her bir sınıflandırma algoritması için en yüksek sınıflandırma başarımı yüksek risk sınıfı için elde edilmiştir. Kullanılan algoritmalar arasında, 97 başarı oranıyla en üstün performansa sahip olanının ekstrem gradyan arttırma algoritması olduğu tespit edilmiştir. Genel olarak elde edilen sonuçlar, önerilen modelin yüksek risk taşıyan gebeliklerin tespitinde son derece etkili olduğunu doğrulamaktadır. Bu bulgu, önerilen MÖ temelli karar destek sisteminin uzman hekimlere gebelik sürecinde daha doğru teşhisler koyma ve gerekli müdahaleleri daha hızlı bir şekilde gerçekleştirme konusunda önemli bir destek sağlama potansiyeline sahip olduğunu göstermektedir.

Article Details

A Machine Learning-Based Decision Support System Design to Predict Maternal Health Risk Groups During Pregnancy

Authors:

İrem Şenyer Yapıcı | Rukiye Uzun Arslan

Publisher:

Karyay Karadeniz Yayımcılık

Main Topics:

Karar Desteği ve Grup Destek Sistemleri | Bilgi Sistemleri (Diğer)

Journal:

Black Sea Journal of Engineering and Science Vol.7 Issue 3, pp.509-520

Document Type:

Article

Publication Date:

5/2024

Document Language:

Turkish

Document ID (Org No.):

2001315152

pISSN:

2619-8991

Pregnancy risk classes

Maternal health

Machine learning

Feature selection

SMOTEGebelik risk sınıfları

Anne sağlığı

Makine öğrenmesi

Özellik seçimi

SMOTE

Issues

Article Details

Authors: İrem Şenyer Yapıcı | Rukiye Uzun Arslan

Publisher: Karyay Karadeniz Yayımcılık

Main Topics: Karar Desteği ve Grup Destek Sistemleri | Bilgi Sistemleri (Diğer)

Journal: Black Sea Journal of Engineering and Science Vol.7 Issue 3, pp.509-520

Document Type: Article

Publication Date: 5/2024

Document Language: Turkish

Document ID (Org No.): 2001315152

pISSN: 2619-8991

Issues

Authors:

İrem Şenyer Yapıcı | Rukiye Uzun Arslan

Publisher:

Karyay Karadeniz Yayımcılık

Main Topics:

Karar Desteği ve Grup Destek Sistemleri | Bilgi Sistemleri (Diğer)

Journal:

Black Sea Journal of Engineering and Science Vol.7 Issue 3, pp.509-520

Document Type:

Article

Publication Date:

5/2024

Document Language:

Turkish

Document ID (Org No.):

2001315152

pISSN:

2619-8991