Data Open Global Championship 2021
As the third place winner of Europe Regional Data Open, we were invited to attend this Gloabl Championship in November 2021.
Plastic revolutionised modern life in many ways. Despite the benefits it brings to human life as well as scientific development, increasing production, consumption and disposal of plastic products is becoming one of the major concerns of global environmental health.
In a group of four, we investigated the historical and forecasted end-of-life processing of international plastic trade through a gravity model. In particular, we verified that current trading flows exacerbate the emission of end-to-life plastics, especially in countries that do not possess and/or enforce sufficiently responsible waste processing protocols.
Our study observes countries with great management being the exporters of plastics whereas the distribution (of management level) of importing countries is more varied. However, it is evident that there has been a trend of importing countries with lesser waste management capabilities possessing a larger share of the overall trade imports. This dynamic is even more evident when specifically looking at plastic waste trading.
We then estimate how changes in the growth for countries will transform the current configuration and made suggestions on improving this situation pragmatically and realistically.
Data Science for All Women's Summit 2021
Xinyi was selected to participant in a rigorous 7-week data science training program focused on building skills in data science, culminating in a capstone presentation.
Banks and bank performance are closely tied to the overall economic health of a country. Bank failures can lead to lower income and compensation growth, higher poverty rates, and lower employment. It is therefore important to understand the factors that may cause a bank to fail. Our project investigates whether it is possible to predict bank failure, and how early we can start seeing signs of failure.
Collaborated with a team of seven, we were able to predict bank failure (on average) five quarters in advance using a Random Forest model.
The success and failure of each US commercial bank in each quarter over the last 25 years were classified using publicly available data on US banks mergers and acquisitions and commercial banks' quarterly call reports.
Datafolio
Presentation
CodeIT Suisse Singapore 2021
In a team of two, we wrote Python programs and built server using Heroku App to perform certain tasks and in a 24-hour time frame.
See code sample for one problem solely written by Xinyi (more samples written collaboratively as a team, including Git Repo, Heroku App is available upon request).
Code Sample
Terminal - Europe Regional Competition 2021
In a group of three, we developed an AI algorithm to play the Terminal game and defeated more than half of pre-selected teams, where the selection process was already competitive.
Our algorithm was also ranked top 50 in Global competition Season 8, as of July 2021.
Asked by the company, we decide (between different strategies) how to minimise the fuel cost if two lorries were sent to deliver vaccines to n cities, while only one of the drivers is certified to distribute vaccines.
(I.e. the other driver can only drive and deliver.)
This is a 4-day virtual workshop bringing PhD students from different universities together.
Organiser: ICMS and MAC-MIGS. See the event page here.
Presentation
Europe Regional Data Open 2020 - Third Place Winner
The Europe Regional Data Open 2020 is an one-week online competition in October 2020. Under the topic of gentrification in the New York Metropolitan area, we focused on identifying what made some tracts being able to develop and absorb outside investments without replacing original residents.
With a logistic regression model using Python, we studied the relationship between dominating industry in a tract and its behaviour when outside investment came in. Our study advises a list of tracts that will absorb investment with the least population displacement, which suggests investors on possible ethical and sustainable investments in the city, in the spirit of developing without destroying culture diversity.
We won the third place and our datafolio was referred as the best datafolio amongst all participants.
Teammates: Pei Dong, Joel Perez Ferrer, and Sean Nassimiha.
Organiser: Citadel and Correlation One. Data: Census Bureau’s American Community Survey (2009-2018).
Datafolio
Certificate