Use of automated machine learning for an outbreak risk prediction tool

Abstract

The frequency of serious epidemics has continued to increase in the last decade. The ability to predict the risk of outbreaks can improve prevention and control. There are few prediction models available, and of these most are manually constructed by human experts. These manual models are affected by the lack of automation and have limitations in data processing. They can be enhanced with modern machine-learning techniques. Machine learning (ML) based prediction models, however, have higher requirement for professional knowledge and are not broadly accessible to researchers who do not have ML expertise. We proposed automated machine learning (AutoML) as an advanced solution. It automates the entire ML design process for users without requiring ML knowledge, therefore allows non-ML experts to individually build ML models. To demonstrate the functionality of AutoML to develop reliable systems, we expanded an existing manually developed risk analysis model, EPIRISK, that uses economic, social and medical risk factors to predict epidemic risk. Using the AutoML platform BrewAI, we obtained an automatically generated ML model to predict epidemic risk. This was compared with six traditionally built machine learning models. The AutoML tool generated a model of 77% accuracy in predicting risk. It had similar accuracy to the six traditional built ML models. Such tools are easy to use and could make ML models more accessible to non-ML experts.

Keywords

• Automated machine learning
• Outbreak risk prediction
• Public health
• Epidemics