The SSC special COVID-19 Student Case Study Competition
In October 2020, the SSC Biostatistics Section launched a special COVID-19 Student Case Study Competition. The goal of the competition was to take any COVID-19 data publicly available and create an analytic tool or develop a model that can be useful for decision makers. A total of 20 teams submitted reports from 12 universities across Canada. Three winning teams have been selected and eight additional reports were selected to be featured on the SSC web site.
We are pleased to share with you the three winners of the competition. We would also like to thank a generous donor, member of the SSC, who contributed an amount of $1,000 to the first prize.
1st prize: (University McGill)
Incorporating mobility data into COVID-19 forecasting
Dirk Douwes-Schultz and Mila Sun
Supervisors: Alexandra Schmidt and Erica Moodie
Forecasting daily COVID-19 cases is critical for short-term planning of hospital and other public resources. One potentially important piece of information for forecasting COVID-19 cases is cellphone data, which measures the amount of time individuals spend at home. Endemic-epidemic time series models are recently proposed auto-regressive models where the current mean case count is modeled as a weighted average of past case counts multiplied by the reproductive number (i.e., the number of secondary infections produced per infectious individual), plus an endemic component. We extend endemic-epidemic models to include a distributed lag model for the effect of mobility on the reproductive number of COVID-19. Further, we introduce a shifted negative binomial weighting scheme for the past counts which is more flexible than previously proposed weighting schemes and perform inference within a Bayesian framework to incorporate uncertainty into the forecasts. Our methods are illustrated in two U.S. counties: King and New York.
2nd prize: (Wilfried Laurier University)
COVID-19 fake news detector
Youjia Zhang, Mohsen Bahremani, Rini Perencsik and Daniel Berezovski
Supervisor: Sunny Wang
The increasing number of Novel Coronavirus (COVID-19) cases has given rise to a proliferation of misinformation related to COVID-19. This misinformation makes it difficult for individuals to find reliable news sources, resulting in protest against government measures to control the virus, social turmoil, and even death. To help alleviate the spread of misinformation, we built and evaluated a variety of machine learning models to help predict the reliability of COVID-19 related news. We combined data from two sources, which include news articles and website posts from official institutions. As our final method, we presented an ensemble of methods that achieves an Area Under Curve (AUC) of 0.97 and an F1-score of 0.92. Additionally, we created a website for readers to interact with the model at: www.modellingcomp.com.
3rd prize (HEC Montréal)
Semantic classification tools in scientific publications
Gabriel Boulanger-Theberge and Simon Tye-Giguère
Supervisor: Laurent Charlin
The current pandemic context has resulted in the production of a colossal volume of work by the scientific community in a wide range of fields of knowledge. Equally important is the resulting quantity of publication, and their consultation is a challenge for public decision-makers. Our ambition in this project was to build a research tool using semantics in order to allow a user to identify, among a large body of science, the articles corresponding to a specific theme. In this context, we have developed a semantic-based research engine using two topic modeling algorithms, namely Latent Dirichlet Allocation (LDA) and K-Means. From a group of words entered by a user, our tool returns scientific articles with the semantics closest to that group of words. A parameter makes it possible to make more general or more specific the topics on the basis of which the research is carried out. Finally, the tool makes it possible to generate recommendations to users for new articles, considering their preferences.