However, the prediction results of these models are quite different. Imperial College London in London warned that as many as 2 million people in the United States may have died in COVID-19 by last summer, while IHME's forecast is much more conservative, with 60,000 people expected to die in August. As it turns out, these two predictions are not very accurate. By the beginning of August, the death toll in the United States finally reached 6.5438+0.6 million.
The huge difference in forecast data last spring attracted the attention of Gu Youyang, a 26-year-old data scientist at that time. The young man holds a master's degree in electronic engineering and computer science from MIT and a degree in mathematics, but he has no formal training in epidemic-related fields such as medicine or epidemiology. Nevertheless, he thinks his experience in dealing with data models may come in handy during the epidemic.
In mid-April, Gu Youyang lived with his parents in Santa Clara, California. He spent a week building his own death prediction model and a website displaying case information in COVID-19. Not long after, the results predicted by his model began to be more accurate than those predicted by institutions with hundreds of millions of dollars of funds and decades of experience.
Extended data:
The novel and complicated changes in Gu Youyang's model come from his application of machine learning algorithm and polishing of data.
After graduating from MIT, Gu Lei worked in the financial industry for several years, writing algorithms for high-frequency trading systems. If he wants to keep this job, his prediction must be accurate.
When talking about the epidemic situation in COVID-19, Gu Youyang constantly compared his prediction results with the final reported death toll, and constantly adjusted his algorithm to get a more accurate prediction. Although this job takes as much time as a full-time job that consumes energy, Gu Youyang still takes time to do it voluntarily and lives on his own savings. He hopes that there is no conflict of interest or political bias in his data.
Although not perfect, Gu Youyang's model performed well from the beginning. At the end of April, he predicted that 80 thousand people would die in the United States by May 9. The actual death toll is 79926.
A similar forecast by IHME at the end of April said that the death toll in the United States will not exceed 80,000 in 2020. Gu Youyang predicted that the death toll in May 18 was 90,000, and the death toll on May 27th was 654.38+10,000, which was consistent with the actual figures again.
Boys born in China after 1990s set up their own death toll model in COVID-19, which beat the authorities in accuracy.