A Comparison of Infectious Disease Forecasting Methods across Locations, Diseases, and Time

Abstract

Accurate infectious disease forecasting can inform efforts to prevent outbreaks and mitigate adverse impacts. This study compares the performance of statistical, machine learning (ML), and deep learning (DL) approaches in forecasting infectious disease incidences across different countries and time intervals. We forecasted three diverse diseases: campylobacteriosis, typhoid, and Q-fever, using a wide variety of features (n = 46) from public datasets, e.g., landscape, climate, and socioeconomic factors. We compared autoregressive statistical models to two tree-based ML models (extreme gradient boosted trees [XGB] and random forest [RF]) and two DL models (multi-layer perceptron and encoder–decoder model). The disease models were trained on data from seven different countries at the region-level between 2009–2017. Forecasting performance of all models was assessed using mean absolute error, root mean square error, and Poisson deviance across Australia, Israel, and the United States for the months of January through August of 2018. The overall model results were compared across diseases as well as various data splits, including country, regions with highest and lowest cases, and the forecasted months out (i.e., nowcasting, short-term, and long-term forecasting). Overall, the XGB models performed the best for all diseases and, in general, tree-based ML models performed the best when looking at data splits. There were a few instances where the statistical or DL models had minutely smaller error metrics for specific subsets of typhoid, which is a disease with very low case counts. Feature importance per disease was measured by using four tree-based ML models (i.e., XGB and RF with and without region name as a feature). The most important feature groups included previous case counts, region name, population counts and density, mortality causes of neonatal to under 5 years of age, sanitation factors, and elevation. This study demonstrates the power of ML approaches to incorporate a wide range of factors to forecast various diseases, regardless of location, more accurately than traditional statistical approaches.

Predicting emerging zoonotic disease under resource-limited settings: Case study of Kyasanur Forest Disease using event-based surveillance data and transfer learning

“…This KFD study is the rst to demonstrate the merits of combining open-source uno cial case count data from local news and internet searches with traditional climate data to strengthen the predictive capabilities of a disease facing all of the above challenges. Forecasting models rely heavily on historic data represented as time lags to improve predictive accuracies 48,49 . Our study showed that uno cial case counts reported in news media could be used as an alternative to historic data obtained from o cial sources to produce timely and reliable nowcasting and short-term forecasting estimates.…”

Predicting infectious disease for biopreparedness and response: A systematic review of machine learning and deep learning approaches

“…Within ML techniques, tree-based methods were popular among all prediction categories. Tree-based methods are often among the best performing types of prediction models [ 19 , 42 , 43 ]. For instance, XGB and, RF outperformed other traditional modeling approaches in predicting diseases such as brucellosis, avian influenza, and influenza-like illnesses across different regions of the world [a107, a167, a180].…”

“…The ML and DL methods are gaining popularity and are widely being used for a variety of disease intelligence tasks, including temporal, spatial, and risk factor predictions [ 18 ]. ML models have been shown to outperform traditional statistical techniques to give more accurate and reliable predictions [ 19 , 20 ]. The popular ML techniques most widely used in the field of ID prediction include tree-based approaches [ [20] , [21] , [22] ] and Support Vector Machines (SVM) [ [23] , [24] , [25] ] due to their ease of implementation and interpretability.…”

Predicting infectious disease for biopreparedness and response: A systematic review of machine learning and deep learning approaches

“…Within ML techniques, tree-based methods were popular among all prediction categories. Tree-based methods such as RF, BTR, and XGB are often among the best performing types of prediction models 9,28,29 . These models are also easy to implement, fast to compute, highly performant, and provide a form of interpretability through input feature importance, which could be the main reasons for their popularity in ID modeling 30,31 .…”

“…Machine Learning (ML) and Deep Learning (DL) methods are widely used for a variety of disease intelligence tasks, including temporal, spatial, and risk factor predictions 8 . ML models have been shown to outperform traditional statistical techniques to give more accurate and reliable predictions 9,10 . The popular ML techniques most widely used in the field of ID prediction include tree-based approaches [10][11][12] and Support Vector Machines (SVM) [13][14][15] due to their ease of implementation and interpretability.…”