Comparison of ARIMA and Random Forest time series models for prediction of avian influenza H5N1 outbreaks

Abstract

BackgroundTime series models can play an important role in disease prediction. Incidence data can be used to predict the future occurrence of disease events. Developments in modeling approaches provide an opportunity to compare different time series models for predictive power.ResultsWe applied ARIMA and Random Forest time series models to incidence data of outbreaks of highly pathogenic avian influenza (H5N1) in Egypt, available through the online EMPRES-I system. We found that the Random Forest model outperformed the ARIMA model in predictive ability. Furthermore, we found that the Random Forest model is effective for predicting outbreaks of H5N1 in Egypt.ConclusionsRandom Forest time series modeling provides enhanced predictive ability over existing time series models for the prediction of infectious disease outbreaks. This result, along with those showing the concordance between bird and human outbreaks (Rabinowitz et al. 2012), provides a new approach to predicting these dangerous outbreaks in bird populations based on existing, freely available data. Our analysis uncovers the time-series structure of outbreak severity for highly pathogenic avain influenza (H5N1) in Egypt.

Variable Selection in Time Series Forecasting Using Random Forests

“…Most of the studies focusing on time series forecasting aim at proposing a forecasting method based on its performance (compared to other methods, usually a few), when applied within a small number of case studies (e.g., [8,14,26,28,30]). Recognizing this specific fact and aiming at providing a tangible contribution in time series forecasting using RF, we have conducted an extensive set of computational experiments.…”

“…Nevertheless, other forecasting algorithms may exhibit better performance in some cases. This does not apply to every experiment conducted, neither does it imply that RF should not be used in one-step ahead time series forecasting in favour of these other algorithms as shown in [24][25][26]30]. In fact, in [12] the interested reader can find a small-scale comparison on 50 geophysical time series, which illustrates the fact that an algorithm can perform better or worse depending on the time series examined.…”

“…Regression using RF can be implemented for time series forecasting purposes. Representative applications can be found in many scientific fields including these of engineering [23,24], environmental and geophysical sciences [25][26][27], financial studies [28,29], and medicine [30], with varying performance. Furthermore, small datasets are used in these applications; therefore, the results cannot be generalized.…”

A comparative study on predicting influenza outbreaks

“…The third type of model is a combination of the first and second types. It uses the numbers of flu patients in the past as features (as in the first type) and regression models (as in the second type) (10). In this study, we adopted the third model type and tried 6 different models with hyperparameter adjustments, including: Auto-Regressive Integrated Moving Average (ARIMA), Support Vector Regression (SVR), Random Forest (RF), Gradient Boosting (GB), Artificial Neural Network (ANN), and Long Short Term Memory (LSTM).…”

Data-Driven Decision-Making and Its Application to the Corporate Cash Management Problem

“…Particularly, random forests (RF) are ensembles of randomly trained decision trees (Ho, 1995;Ho, 1998;Criminisi and Shotton, 2013). Recent examples of time-series forecasting using random forests can be found in Kumar and Thenmozhi (2006), Kane et al (2014), Mei et al (2014), andZagorecki (2015).…”