|
| Title | Machine learning methods and findings | Shortcoming and pitfalls |
|
| Supervised learning techniques for analysis of neonatal data [22] | ANN, NB, SVM, and LR have been used. The highest accuracy is 89% with SVM | The dataset is small with only demographic features. There is also a high-class imbalance |
| Prediction of neonatal deaths in NICUs: development and validation of machine learning models [23] | ANN, RF, CHART, SVM, and ensembles have been used. The highest accuracy is 94% with SVM | The dataset is imbalanced |
| An artificial neural network model for neonatal disease diagnosis [24] | ANN has been used. The accuracy is 75% | The dataset contains 94 rows which is too small. Clinical features, which are very essential in predicting neonatal disease, were not used |
| Developing a fuzzy expert system to predict the risk of neonatal death [25] | A fuzzy model inference system has been used. The accuracy is 90% | Comparisons were not made with similar machine-learning techniques. The reason for using fuzzy was not also adequately justified |
| Machine learning techniques for neonatal apnea prediction [26] | DT, SVM, and RF have been used. The highest accuracy is 88% with RF | The small dataset and the selected machine-learning models were not adequately justified |
| Medical decision support using machine learning for early detection of late-onset neonatal sepsis [17] | SVM, NB, and its variants TAN and AODE, K-nearest neighbor, CART, RF, LR, and LBR. Machine learning models outperformed physicians | Bias may be introduced in the method of conversion of temporal variables |
| Machine learning approach for predicting underfive mortality determinants in Ethiopia: evidence from the 2016 Ethiopian demographic and health survey [27] | RF, LR, and KNN have been used. The highest accuracy is 67.2% with RF | Clinical features were not used and the accuracy was low |
|
