Details

Autor(en) / Beteiligte

• Cheng, Yunyun
• Bai, Yanping
• Xu, Ting
• Hu, Hongping

Titel

Global COVID-19 development trend forecast based on machine learning

Ist Teil von

• 2022 International Conference on Big Data, Information and Computer Network (BDICN), 2022, p.128-131

Ort / Verlag

IEEE

Erscheinungsjahr

2022

Quelle

IEEE Electronic Library Online

Beschreibungen/Notizen

• The outbreak of COVID-19 not only affects people's health, but also hinders the pace of economic progress of various countries. Our goal was to develop a prediction model based on machine learning, which could be used to predict development trend of COVID-19 in the future. It can provide governments and health authorities with useful information conducive to decision-making. Considering that the propagation of COVID-19 is affected by many factors and a single prediction model lacks all-round monitoring of the data set, the ARIMA-SVM integration model was established by using the global cumulative number of confirmed cases. The individual models of ARIMA and SVM were used to predict the COVID-19 trend. Based on the prediction results of the above prediction model, a new integration forecast model was formed through a combination of weighted weights. Finally, the forecast results of the combined model and the individual model were compared. The prediction performance of models were compared according to Mean Absolute Percentage Error (MAPE). The prediction results showed that the MAPE values of ARIMA model, SVM model and ARIMA-SVM integration model were 15.843%, 1.251%, 1.132% respectively. Compared with the traditional machine learning models ARIMA and SVM, the combined model has reduced the average absolute error percentage by 92.103% and 9.51%, respectively, and can achieve more accurate and reliable COVID-19 trend prediction. It used two single models to complement each other, reduced the systematic error of the prediction model, and significantly improved the prediction effect.