빅데이터 분석 기반의 코로나 바이러스 사망자 수 예측 - 변화점 추적 방식과 예측조합법을 중심으로 -

As the spread of the corona virus slows down, this study predicts the corona virus deaths in Korea using various prediction methods. While the previous studies in this domain mainly focused on predicting the corona virus confirmed cases using autoregressive models and deep learning based time-series models, this study aims at the accurate prediction of the corona virus deaths using various machine learning methods with the above prediction methods. In particular, this study facilitated the prediction stability by adopting change-point detection and forecasting combination method. As explanatory variables, this study used various corona virus-related data, social distancing level and daily temperature, and it also showed the comparison of the prediction results of various prediction methods and two forecasting combination methods. The prediction performances of the various prediction methods differed depending on the prediction interval segmented by change-point detection. But it was found that the forecast combination methods mainly showed more accurate prediction results compared to other methods. In addition, the forecast combination methods showed the low volatility of the prediction error implying their stability of prediction. This study has significance in that it accurately predicted the corona virus deaths and achieved the accuracy and stability of the prediction, facilitating the practicality in the real application cases.