ADAPTIVE MULTI-LAYER LEARNING FRAMEWORK FOR OUTBREAK PREDICTION

Abstract

In pandemic situations, it is crucial to analyse outbreaks to identify infections and predict their patterns. Machine Learning models are a powerful tool for outbreak analysis and provide effective insights for pandemic management. So far, most of the work has focused on specific pandemics or outbreaks with similar characteristics. But as history shows, every outbreak has unique traits or symptoms that don't match those in previous studies, and that’s where prediction models often fail. The goal of this study is to provide a framework that is more generic and can be applied to any kind of spread that may turn into an epidemic or pandemic. The major challenge is to find a common framework that can be applied to all pandemics or epidemics. In this work, six pandemic datasets were collected and pre-processed using data cleaning, min-max normalisation, and data augmentation methods, such as Synthetic Minority Over-sampling Technique (SMOTE), to obtain a balanced dataset. Further, a hybrid machine learning model with a three-layer architecture is applied for outbreak prediction. The first layer is an Adaptive Boosting Support Vector Machine (AdaBoost SVM) for sequence modelling. The second layer is K-Nearest Neighbours (KNN) for better categorisation, and the third layer is an optimised Artificial Neural Network (O-ANN) for precise prediction with an accuracy 97.73%, which is better than the standard model.