[Paper Review] Artificial Intelligence Forecasting of Covid-19 in China

Zixin Hu, Qiyang Ge+3 more|arXiv (Cornell University)|Feb 17, 2020

COVID-19 epidemiological studiesMathematics7 references118 citations

TL;DR

The paper proposes an AI-inspired approach using a modified stacked auto-encoder to forecast COVID-19 trajectories across China, achieving multi-step forecast errors in the low single digits and clustering provinces into transmission-based groups.

ABSTRACT

BACKGROUND An alternative to epidemiological models for transmission dynamics of Covid-19 in China, we propose the artificial intelligence (AI)-inspired methods for real-time forecasting of Covid-19 to estimate the size, lengths and ending time of Covid-19 across China. METHODS We developed a modified stacked auto-encoder for modeling the transmission dynamics of the epidemics. We applied this model to real-time forecasting the confirmed cases of Covid-19 across China. The data were collected from January 11 to February 27, 2020 by WHO. We used the latent variables in the auto-encoder and clustering algorithms to group the provinces/cities for investigating the transmission structure. RESULTS We forecasted curves of cumulative confirmed cases of Covid-19 across China from Jan 20, 2020 to April 20, 2020. Using the multiple-step forecasting, the estimated average errors of 6-step, 7-step, 8-step, 9-step and 10-step forecasting were 1.64%, 2.27%, 2.14%, 2.08%, 0.73%, respectively. We predicted that the time points of the provinces/cities entering the plateau of the forecasted transmission dynamic curves varied, ranging from Jan 21 to April 19, 2020. The 34 provinces/cities were grouped into 9 clusters. CONCLUSIONS The accuracy of the AI-based methods for forecasting the trajectory of Covid-19 was high. We predicted that the epidemics of Covid-19 will be over by the middle of April. If the data are reliable and there are no second transmissions, we can accurately forecast the transmission dynamics of the Covid-19 across the provinces/cities in China. The AI-inspired methods are a powerful tool for helping public health planning and policymaking.

Motivation & Objective

• Motivate an AI-driven alternative to traditional epidemiological models for real-time COVID-19 forecasting in China.
• Develop a modified stacked auto-encoder to model transmission dynamics from early 2020 data.
• Use latent representations to cluster provinces/cities and uncover transmission structure.
• Provide multi-step forecasts of cumulative confirmed cases and identify plateau timings across regions.

Proposed method

• Develop a modified stacked auto-encoder to model COVID-19 transmission dynamics.
• Apply the model to real-time Chinese provincial data from January 11 to February 27, 2020.
• Use latent variables from the auto-encoder for clustering to group provinces/cities into transmission-based clusters.
• Perform multiple-step forecasting to predict cumulative cases from Jan 20, 2020 to Apr 20, 2020.
• Estimate forecast accuracy with stepwise errors for 6- to 10-step horizons.

Experimental results

Research questions

• RQ1 Can AI-inspired models, specifically a modified stacked auto-encoder, accurately forecast COVID-19 trajectories across Chinese provinces and cities?
• RQ2 How do latent representations capture regional transmission structures and support clustering of provinces/cities?
• RQ3 What are the expected accuracies of multi-step forecasts (6–10 steps) for cumulative confirmed cases in China?
• RQ4 When do provinces/cities enter the plateau phase of the forecasted transmission dynamics under reliable data assumptions?

Key findings

• The AI-based method achieved low average forecast errors for multi-step horizons (6-step to 10-step) with reported values 1.64%, 2.27%, 2.14%, 2.08%, and 0.73% respectively.
• 34 provinces/cities were grouped into 9 clusters based on latent representations.
• The model forecasted the trajectory of cumulative confirmed cases across China from Jan 20, 2020 to Apr 20, 2020.
• Predicted that the epidemics could be over by the middle of April under reliable data and no second transmissions.
• Latent-variable clustering revealed transmission structure across provinces/cities, supporting targeted public health planning.