UPSI Digital Repository (UDRep)
|
UPSI Digital Repository (UDRep)
|
|
|
|
||||||||||||||||||||||||
| Abstract : Perpustakaan Tuanku Bainun |
| Postpartum depression (PPD) represents a significant mental health concern that affects the relationship between mothers and their infants worldwide, including in Indonesia. Specifically, in Banjarmasin Indonesia, it has been reported that about 43% of postpartum mothers are diagnosed with the symptoms which become a consideration scenario for scientific study. It has been noted that health institutions in Banjarmasin, Indonesia, are deficient in an effective detection tool, primarily due to the expensive, time-intensive, and insufficient nature of existing methods. The conventional screening approach, which relies on questionnaire instruments, fails to yield a meaningful dataset of PPD that can be analyzed scientifically and accurately. A purposive sampling method was applied, yielding a sample of 317 respondents from healthcare institutions in Banjarmasin City. A hybrid resampling technique combining Bootstrapping and the Synthetic Minority Oversampling Technique (SMOTE) was implemented to address the majority/minority class imbalance. Three experimental approaches were adopted using (1) all Postpartum Depression Risk Factors (PPDRF) features, (2) features selection using statistical tests, and (3) automatic features selection using Relief and Backward Elimination (BE). Four Machine Learning (ML) algorithms which are Naive Bayes (NB), Logistic Regression (LR), Random Forest (RF), and Adaptive Boosting (AdaBoost) were employed to compare the performance of the PPD determination model. The findings revealed that the consistently identified key PPDRF features are mode of delivery, birth weight, fear of delivery, disposition during pregnancy, family relationships, and support from family and husband. On the other hand, the model of RF integrated with BE demonstrated the highest performance, with an accuracy of 91.62% and an Area Under Curve (AUC) of 0.966. In contrast, the statistical test for RF yielded the lowest result, with an accuracy of 73.38% and an AUC of 0.752. The research demonstrated that an ML-based model is capable of accurately identifying PPD in mothers. The results of this research could assist healthcare institutions, particularly in Banjarmasin Indonesia, in improving their clinical decision-making processes. |
| This material may be protected under Copyright Act which governs the making of photocopies or reproductions of copyrighted materials. You may use the digitized material for private study, scholarship, or research. |