Leveraging Machine Learning to Predict Asthma following Vaccination
Seema C. Gull, NaveenKumar C.G., Karuna C. Gull, V. B. Sanskruti, Mahantesh Laddi · 2024
This project seeks to employ more sophisticated approaches to machine learning to determine asthma occurrences after vaccine administration. As vaccine distribution becomes more widespread around the world, it is important to monitor and prevent issues such as asthma. The goals for future research include the creation of risk-adjusted models that may predict the risk of asthma following vaccination, depending on patients’ demographic information and the characteristics of the vaccines. The research uses several sophisticated techniques such as Artificial-Neural-Networks (ANN), Random-Forests (RF), Naïve-Bayes (NB), and logistic regression (LR) to improve the ability to predict and identify prevention measures and healthcare management. Thus, a significant innovation of this work is the comprehensive assessment of these algorithms when exposed to noise, which is a common issue in many medical data sets. The study also showed that Random-Forest and ANN models were significantly more accurate and yielded larger ROC AUC scores than other models; moreover, they were much less sensitive to noise with noise values ranging from 0 to 0.5. This places RF and ANN on a better pedestal to offer a better prediction of post-vaccination asthma than the other models, especially in a scenario where there is the possibility of data contamination. The present findings can thus help create better and more resistant models of prediction that are useful in better health management and treatment of asthma over and after vaccination.