New Research Agenda: Utilizing Machine Learning to Build Public Health Intelligence
Kemal N. Siregar, Retnowati
Faculty of Public Health Universitas Indonesia
Can Indonesia predict the emergence of COVID-19 earlier? It seems difficult to answer this question. In theory, the identification of viral infections in humans is recognized by the symptoms of emerging diseases. But the next question is the extent to which Indonesia’s existing infrastructure can identify, say, symptoms of fever that develop in the community, while the scope of Indonesia’s territory is vast. Issues such as this are the topic of Public Health Intelligence, where the application of these remains a major challenge in Indonesia.
Indeed, Public Health Intelligence is important to intelligence at the stage where we can draw evidence-based decisions for action to improve public health. This requires the application of a blend of analytics to generate meaningful information. Here machine learning is proposed as important data analytics for building reliable Public Health Intelligence.
In the era of disruptive technology, for example in surveillance that traditionally uses structured data, needs to be supplemented by textual data to explain why infectious diseases that can be prevented by immunization do not go down. With Machine Learning that makes use sentiment analysis or social network analysis for social media data, we can assess public opinion of certain communities that are negative about vaccination.
So, Machine Learning is important as it can be used to automate a lot of different tasks that where thought of as tasks that only humans can do like pattern recognition, image recognition or text generation. It gives computers the ability to learn without being programmed explicitly, and the steps systematically consist of gathering data, preparing data, choosing a model, training, evaluation, hyper parameter tuning, and prediction. Machine learning algorithms have been successfully employed for complex health-related problems such as disease diagnosis or risk prediction, but its potential in public health remains underexplored.
At present researches conducted under the auspices of the Health Informatics Research Cluster at the Faculty of Public Health, University of Indonesia are exploring the potential uses of Machine Learning for Public Health Intelligence, including the following. Analyzing textual data patterns for real-time surveillance to recognize public opinion about various health issues, for example about drugs, vaccinations. Predict personal risk profiles and behavior patterns, which are then followed by offering targeted and personalized health advice.