Ensemble Approach for Zoonotic Disease Prediction Using Machine Learning Techniques

Abstract

More than two thirds of emerging infectious diseases in recent decades are zoonotic in origin. Timely prediction of these diseases which migrate from animals to humans and preventive measures to stop the loss in terms of morbidity and mortality is the requirement of health care industry. Avian Influenza is one of the zoonotic diseases that has created havoc in recent past especially in Asian subcontinent. In past, attempts have been made to predict influenza using traditional time-series techniques ( AR, MA, ARMA, ARIMA etc.) as well as machine learning techniques to capture the cyclicity and seasonality of these virus strains. In current research an effort has been made to utilize the Empirical Mode Decomposition (EMD) to extract the Intrinsic Mode function (IMF) and then apply state of art Machine Learning (ML) techniques to predict the series. Several machine learning techniques like Random Forest (RF) along with Gradient Boosting Machine (GBM) and Support Vector Regression (SVR) have been applied on the decomposed series. Exogenous meteorological variables like temperature, humidity and precipitation have been incorporated to improve upon the forecast. An ensemble approach of ML models showed significant improvement over the traditional models in terms of long term forecast accuracy.