Students Use Machine Learning to Make COVID Predictions

LMU Master of Science in Business Analytics (MSBA) students are using their newfound knowledge of machine learning and analytics to solve timely issues related to COVID-19. Machine learning is the science of getting computers to discover patterns in data that human brains cannot intuitively recognize. By feeding data and information in the form of observations and real-world interactions, computers can be trained to learn like human brains and improve their learning over time in autonomous fashion.

In Professor Arin Brahma’s “Introduction to Machine Learning” class, four student teams came up with their own research topics and questions around COVID-19, and used a variety of data sources to develop complex data methods and algorithms to find solutions to real issues. Here is a summary of each project:

Team 1: What is the severity level of a COVID-19 case based on a patient’s symptoms?

The goal of this project is to help hospitals and society more quickly determine how severe of a case a person has, which can be used to treat them accordingly to help save healthcare resources. The team harvested data from Kaggle.com, cleaned the data and transformed it into a machine learning format. The team was able to identify characteristics that contribute to the most critical and deceased cases: background diseases, pneumonia, breathing difficulty, respiratory failure and ages 60-69. In the end, the team successfully developed a small machine learning model with a high accuracy rate that could be used to predict future COVID cases in a controlled environment.

Team 2: When will the number of hospital beds will be overrun by COVID-19 patients?

The goal of this project is to predict when the number of hospital beds in a city will be overrun by patients. Using data from the World Health Organization (WHO) and American Hospital Association (AHA), the team aimed to create a model that could help with pandemic preparation. They ran into several challenges: as an academic project, they were permitted to use the AHA dataset for only three states: California, New York and Washington; the WHO did not have data for recovered patients; and the model they came up with used several assumptions that would be overly simplistic in the real world. In the end, the team determined that while the model demonstrates an approach to solve the problem, more data would be needed for the model to be effective in a real world situation.

Team 3: Where will COVID-19 have the most significant lasting impact on society?

The goal of this project is to determine which countries will experience the most economic turmoil post COVID-19. Using data integrated from six different sources, the team concluded that using the two variables “Quality of Life” and “Serious Patients” can determine which countries COVID-19 will have the most impact on. The team found that countries in South America need to pay more healthcare attention to patients who are in serious condition. The data also showed that developed countries are underestimating the threats and developing countries are not able to test/collect stats on all patients.

Team 4: Is China underreporting their COVID-19 numbers?

The goal of this project is to determine if the COVID figures being reported out of China are accurate. After reviewing available visualizations and exploratory analysis pertaining to COVID-19 around the world, it became apparent that the figures being reported by China are overwhelmingly low in comparison to other highly affected countries such as Italy, Spain and the U.S. The team compared Italy to China and found that Italy results were conclusive while China results were inconclusive. In the end, the team could not make a deterministic conclusion if China is indeed underreporting their cases. However, their solution approach can possibly lead to such analysis.

A common theme throughout all projects was lack or limitations of data. Since coronavirus is so new, there are still a lot of unknown factors to make a solid assessment.

“Overall, I was so impressed with all the teams and how quickly they turned around these complex data projects” said Professor Brahma. “In less than a year, our students have developed the skills to interpret machine learning-based predictive models to support business decision making.”

“These projects generated some really cool ideas and insights about the COVID-19 crisis, and are a great example of how we are integrating our CBA mission to use business as a force for good,” added Dayle Smith, dean of LMU College of Business Administration.