The accuracy of the new crown prediction model created by Chinese American little brother is far higher than that of famous universities and American official institutions


The accuracy of the new crown prediction model created by Chinese American little brother is far higher than that of famous universities and American official institutions

In 2020, with the outbreak of COVID-19, more and more people paid attention to the prediction of the number of new crown deaths. Two of the most striking models were established by Imperial College London and IHME in Seattle, USA.

However, the results predicted by the two models are quite different. Imperial College London said that until the summer, the number of deaths due to the new coronavirus in the United States may rise to 2 million, while IHME predicted that it was more conservative, predicting that the death toll in August last year may be 60000. But facts have proved that their conjectures are far from the facts. The data show that the death toll in the United States in early August last year was about 160000.

Self built model of Chinese brother

The huge difference in the predicted numbers attracted the attention of Youyang Gu, a 26 year old Chinese data scientist at the time.

This young man majored in electronic engineering and computer science and mathematics at MIT. He also received a master’s degree from MIT and finished his thesis in the natural language processing group of MIT’s computer science and artificial intelligence laboratory. After graduating from MIT, Gu Youyang entered the financial industry to write algorithms for high-frequency trading system. The most important task of his work is to improve the accuracy of model prediction.

He thinks that his background in dealing with data models can help predict COVID-19, but he has never learned knowledge in medicine or epidemiology, so when he started the project in April 2020, the first step was to search for epidemiology by Google. In the process of creating the new crown prediction model, Gu Youyang gave full play to his working experience in the financial industry. He constantly compared his prediction with the final report of the total number of deaths, and constantly adjusted his machine learning algorithm to make it more and more accurate.

A week later, Gu Youyang’s COVID-19 prediction model and a website showing information were released. Three weeks later, the Centers for Disease Control and prevention listed the model as one of the six major prediction models. With the development of the epidemic, Gu Youyang was invited to attend regular meetings of CDC, professional modelers and epidemiologist teams. The traffic of Gu Youyang’s website also experienced explosive growth, with millions of people checking the overall situation of their state and the United States every day. Over the next few years, his models were more accurate than those made by institutions with hundreds of millions of dollars and decades of experience.

Gu Youyang’s model performed well from the beginning. At the end of April, he predicted that 80000 people would die in the United States by May 9. The actual death toll was 79926. On May 18, 90000 deaths were predicted, and on May 27, 100000 deaths were predicted, which again matched the figures. In November last year, Gu Youyang decided to end the forecast, because he saw that other models had become better and better, and he thought his work had been completed. One month before the project was stopped, Gu Youyang predicted that 231000 deaths would be recorded in the United States on November 1. When November 1 arrived, the death toll reported in the United States was 230995.

The “acid” of anti ridicule organization

From March to April last year, IHME faced a lot of criticism because its predictions were seriously inconsistent with the actual situation. Nevertheless, this prestigious center, located at the University of Washington and funded by the bill and Melinda Gates Foundation with more than $500 million, is cited almost daily in the briefing by former US President trump and members of the US government.

Murray of IHME has his own views on Gu Youyang’s withdrawal. Gu’s model won’t find the seasonal nature of the coronavirus, and will miss the surge in winter cases and deaths, he said. He allowed the epidemic to disappear in the winter, and we found the seasonality as early as may. Gu Youyang’s machine learning method works well in short-term forecasting, but he is not good at understanding what is happening in the overall situation.

Murray also said the algorithms were based on the past and could not explain virus variants and the fact that vaccines might not work on them. On its own, IHME correctly invokes the early peaks of the virus and then makes mistakes in predicting a sharp drop in the number of deaths until it adjusts its model to better reflect reality.

Gu Youyang refused to respond to Murray’s remarks about his model and made a backhand taunt: “I am very grateful to Dr. Chris Murray and his team for their work. Without them, I would not have been where I am today.”

Now Gu Youyang is back in the process of creating the model. This time, he is creating the data about how many people in the United States are infected with the new vaccine, the speed of vaccine launch and when it is possible to achieve mass immunization. Before the outbreak, Gu Youyang was planning to start a business. At that time, he chose the field of sports analysis, but now he is considering doing public health. He said he wanted to do a job that could have a huge impact, while avoiding politics, prejudice and sometimes the burden of large institutions.

He also gave some advice to young people: age is just a number. You don’t need decades of experience to think critically and adapt. In this digital age, where information is very easy to obtain, don’t let the lack of professional knowledge prevent you from pursuing interesting things. Is the current system perfect? Of course not. I hope I can inspire others to pursue the projects they believe in like me, and don’t be afraid of not being recognized.

Gu Youyang’s personal blog:

Gu Youyang’s GitHub:

The accuracy of the new crown prediction model created by Chinese American little brother is far higher than that of famous universities and American official institutions