Researchers at WeBank applied machine learning to satellite imagery of Chinese manufacturing sites.

The AI Moonshot Team at China’s WeBank, a private Chinese online banking company founded by TenCent, combined big data and AI to check on the impact of the coronavirus in Chinese manufacturing. (IEEE Spectrum)

They innovatively processed satellite, GPS and social media data through AI. They found that the Chinese economy may be getting back on its feet after the coronavirus disaster.

Satellite and manufacturing

The researchers customized a deep learning system that was originally devised to map terrestrial solar panel installations from satellite imagery. The AI-powered system then analyzed satellite images from various satellites including the Sentinel-2 satellite.

Some images of steel manufacturing that appeared as hotspots in infrared pictures showed that the Chinese steel sector had bounced to 76% of capacity by February 9. This compared to a low of 29% of capacity seen at the time the outbreak started. Satellite imagery of corporate parking lots revealed that by February 10 Tesla (NASDAQ: TSLA)  had resumed full production at its Shanghai car plant.

Meanwhile, according to a Reuters report, calls at Chinese ports have zoomed up since the second week after the lunar New Year holiday. According to data from Clarkson’s Research, a maritime consultancy, these had surpassed 2090 levels by the end of February.

GPS data from mobile phones

The AI Moonshot Team also used AI to analyze daily anonymized GPS data from several million mobile phone users in 2019 and 2020. AI could identify those mobile phone users who were commuters. It could, therefore, provide a comparative analysis of the number of commuters between the two periods. The researchers found that approximately 75% of the Chinese workforce covered in the analysis had returned to work by March 10.

Further, a Nomura research report quoted by Reuters said about 62% of the Chinese firms most affected by the virus had resumed work by March 8. Over 74% had returned to their jobs in the broader economy.

Natural language processing of social media services

In an innovative application of natural language processing techniques applied to social media platforms, the research team looked for trends in telecommuting, gaming, online education, social networking, express delivery, streaming video, and e-commerce services.

They found a solid uptick in the usage of these services ranging from +47% in social networking to +537% in remote working. However, express delivery was an exception, being a shade less than 1% in the negative, while e-commerce was flat.

The exercise also found that the consensus on Chinese social media was that the country’s economy would be near-normal by the end of March.

Related Story:   Venture Capital: Chinese Startups On The Rocks As Funds Dry Up Following Virus