Adversarial Training for Predicting the Trend of the COVID-19 Pandemic

It is significant to accurately predict the epidemic trend of COVID-19 due to its detrimental impact on the global health and economy. Although machine learning-based approaches have been applied to predict epidemic trend, standard models have shown low accuracy for long-term prediction due to a high level of uncertainty and lack of essential training data. This paper proposes an improved machine learning framework employing generative adversarial network (GAN) and long short-term memory (LSTM) for adversarial training to forecast the potential threat of COVID-19 in countries where COVID-19 is rapidly spreading. It also investigates the most updated COVID-19 epidemiological data before October 18, 2020 and models the epidemic trend as time series that can be fed into the proposed model for data augmentation and trend prediction of the epidemic. The model is trained to predict daily numbers of cumulative confirmed cases of COVID-19 in Italy, USA, China, Germany, UK, and across the world. The paper further analyzes and suggests which populations are at risk of contracting COVID-19.

Predicting the Epidemics Trend of COVID-19 Using Epidemiological-Based Generative Adversarial Networks

The Coronavirus disease 2019 (COVID-19) is a respiratory illness that can spread from person to person. Since the COVID-19 pandemic is spreading rapidly over the world and its outbreak has affected different people in different ways, it is significant to study or predict the evolution of its epidemic trend. However, most of the studies focused solely on either classical epidemiological models or machine learning models for COVID-19 pandemic forecasting, which either suffer from the limitation of the generalization ability and scalability or the lack of surveillance data. In this work, we propose T-SIRGAN that integrates the strengths of the epidemiological theories and deep learning models to be able to represent complex epidemic processes and model the non-linear relationship for more accurate prediction of the growth of COVID-19. T-SIRGAN first adopts the Susceptible-Infectious-Recovered (SIR) model to generate epidemiological-based simulation data, which are then fed into a generative adversarial network (GAN) as adversarial examples for data augmentation. Then, Transformers are used to predict the future trends of COVID-19 based on the generated synthetic data. Extensive experiments on real-world datasets demonstrate the superiority of our method. We also discuss the effectiveness of vaccine based on the difference between the predicted and the reported number of COVID-19 cases.

[Prediction of severe outcomes of patients with COVID-19].

Objective: To establish a new model for the prediction of severe outcomes of COVID-19 patients and provide more comprehensive, accurate and timely indicators for the early identification of severe COVID-19 patients. Methods: Based on the patients' admission detection indicators, mild or severe status of COVID-19, and dynamic changes in admission indicators (the differences between indicators of two measurements) and other input variables, XGBoost method was applied to establish a prediction model to evaluate the risk of severe outcomes of the COVID-19 patients after admission. Follow up was done for the selected patients from admission to discharge, and their outcomes were observed to evaluate the predicted results of this model. Results: In the training set of 100 COVID-19 patients, six predictors with higher scores were screened and a prediction model was established. The high-risk range of the predictor variables was calculated as: blood oxygen saturation <94%, peripheral white blood cells count >8.0×10(9), change in systolic blood pressure <-2.5 mmHg, heart rate >90 beats/min, multiple small patchy shadows, age >30 years, and change in heart rate <12.5 beats/min. The prediction sensitivity of the model based on the training set was 61.7%, and the missed diagnosis rate was 38.3%. The prediction sensitivity of the model based on the test set was 75.0%, and the missed diagnosis rate was 25.0%. Conclusions: Compared with the traditional prediction (i.e. using indicators from the first test at admission and the critical admission conditions to assess whether patients are in mild or severe status), the new model's prediction additionally takes into account of the baseline physiological indicators and dynamic changes of COVID-19 patients, so it can predict the risk of severe outcomes in COVID-19 patients more comprehensively and accurately to reduce the missed diagnosis of severe COVID-19.