Machine learning for disease surveillance or outbreak monitoring: A review

Amidst the global pandemic, the methodical acquisition, analysis, and evaluation of health-related data is crucial to learn from experience and strengthen preparedness for future challenges of similar nature. The goal of this paper is to survey the recent approaches to disease surveillance or outbreak monitoring in the context of artificial intelligence or machine learning. Utilizing Elsevier's Scopus database, the keywords, “Disease Surveillance” or “Outbreak Monitoring” yielded 12,648 document results (with year duration starting 2016). Then, the documents were reduced to 367 after conjunction with the terms, namely, “Artificial Intelligence” or “Machine Learning (ML)” or “Data Science” and limiting the document type to article and review only. The documents were examined one-by-one leaving only the most recent and those papers which applied machine learning methods. The survey showed four major ML tasks in the recent literature, namely, topic modeling, time series forecasting & regression, classification, and clustering. It was also observed that many disease surveillance systems are still utilizing traditional (shallow) machine learning techniques. However, deep learning-based techniques was also demonstrated, in the literature, to perform better than the traditional models.