Predictive analysis of pediatric gastroenteritis risk factors and seasonal variations using VGG Dense HybridNetClassifier a novel deep learning approach.

Abstract

Pediatric gastroenteritis is a major reason for sickness and death among children worldwide, especially in places where healthcare and clean sanitation are scarce. Conventional methods of diagnosis overlook possible risks and seasonal trends, which results in patients receiving treatment too late and more of them being hospitalized. The study sets out to create a new deep learning method that boosts the initial prediction, proper classification, and seasonal trends of pediatric gastroenteritis through the use of hybrid convolutions. The VDHNC model was formed by merging the strong feature learning of VGG16 with the efficient information sharing feature of DenseNet. To create the model, data about clinical, demographic, and environmental aspects of pediatric patients were used. The dataset was preprocessed by using imputation, normalization, managing outliers, and using SMOTE to balance classes. Further validation was performed by analyzing the model performance using one-way ANOVA and pairwise t-tests with several baselines such as SVM, Random Forest, and XGBoost. The VDHNC model was able to achieve a high accuracy of 97%, and was more precise, recalled more information, and reported a higher AUC-ROC score than any other model. The model was able to discover signs of seasonal gastroenteritis, which assisted in predicting future outbreaks. A statistical test proved that VDHNC was better than the other approaches with a p-value of less than 0.05. VDHNC proves reliable when it comes to early detection and assessment of risk in pediatric gastroenteritis cases. The solidness and ease of understanding in this model suggest it can be helpful for making real-time public health decisions and planning hospital resources.