An Ensemble Machine Learning Approach For Time Series Forecasting of COVID-19 Cases

Renato R. Maaliw, Melvin A. Ballera, Zoren P. Mabunga, Aubee T. Mahusay, Dhenalyn A. Dejelo, Mariebeth P. Seño

20212021 IEEE 12th Annual Information Technology, Electronics and Mobile Communication Conference (IEMCON)20 citationsDOI

Abstract

Forecasting assists governments, epidemiologists, and policymakers make calculated decisions to mitigate the spread of the COVID-19 pandemic, thus saving lives. This paper presents an ensemble machine learning model by combining the distinctive strengths of autoregressive integrated moving averages (ARIMA) and stacked long short-term memory networks (S-LSTM) using extensive training procedures and model integration algorithms. We validated the model's generalization capabilities by analyzing time series data of four countries, such as the Philippines, United States, India, and Brazil spanning 467 days. The quantitative results show that our ensemble model outperforms stand-alone models of ARIMA and S-LSTM for a 15-day forecast accuracy of 93.50% (infected cases) and 87.97% (death cases).

Topics & Concepts

Autoregressive integrated moving averageGeneralizationTime seriesComputer scienceCoronavirus disease 2019 (COVID-19)Artificial intelligenceMachine learningSeries (stratigraphy)Ensemble learningEnsemble forecastingAutoregressive modelLong short term memoryEconometricsData miningArtificial neural networkRecurrent neural networkMathematicsPathologyPaleontologyMathematical analysisBiologyDiseaseMedicineInfectious disease (medical specialty)COVID-19 epidemiological studiesCOVID-19 diagnosis using AIAnomaly Detection Techniques and Applications