Their project used data from the Australian Financial Security Authority (AFSA) and the Australian Tax Office (ATO) to make predictions about whether Australians who are bankrupt will comply with their obligations.
The Insolvit platform had 3 main features. The compliance dashboard provides an overview of the state of non-compliance in Australia. Their heat map visualises past non-compliance based on locations. They can then overlay other data sets to find correlations and determine why certain areas have high non-compliance. In addition, their cutting edge machine learning model allowed them to find the key attributes that lead to non-compliance.
The machine learning model used was a state of the art random forest that used categorical feature engineering, bagging, out-of-bag scoring, and testing out hyperparameters through grid search. The model was able to predict issues of non-compliance with 97% accuracy.
Apart from Aaron and Denver Stove who were the technical part of the team, the other half was primarily filled with lawyers who have a passion in technology. One included the Legal Director of Data61, Australia’s leading data innovation group, part of CSIRO. “An unexpected mix of people” but it worked out really well because the legal part of the team understood the problem extremely well.
The team won several prizes including from AFSA which subsequently invited (and paid for) Aaron and Denver to come to come to Canberra to take the project further for them in order to implement it in production for Australians everywhere. Aaron says he was really impressed that a government agency was genuinely trying to use the very latest technology to predict non-compliance and help average Australians.
The data used by the team at GovHack was just a snippet of the full set that they could access when working inside AFSA.
Aaron has always been interested in space, and NASA was a natural extension of that curiosity. Not surprisingly, when we spoke to Aaron, he was working as a data scientist intern at an artificial intelligence company, a portfolio company of NASA & California Institute of Technology’s Jet Propulsion Laboratory. This company was allowed to commercialize technology used and developed through NASA’s Deep Space program. Because of their unique heritage, the company uses this technology to solve industrial problems on Earth instead.
The world’s current supercomputers are still not powerful or robust enough to simulate massive problems such as climate change, or high resolution computational fluid dynamics for aerospace. One potential solution is to use hardware, optimized for machine learning to approximate these simulations instead. Theoretically, this will not only be faster, but allow higher resolution approximations.
Aaron was tasked with building a deep learning model that could approximate general nonlinear partial differential equations (PDEs – which are mathematical expressions of physical phenomena). Experts believe that the future of scientific computing will be through these type of deep learning models that are built with physics in mind.
One aspect that differentiates this work from others is the fact that case-based reasoning or knowledge-based AI was used as an additional layer rather than purely numeric AI.
The GovHack experience
Aaron enjoyed his GovHack experience. The group of lawyers were extremely bright and required Aaron to give a lot of concise communication and to delve deep into the problem. By doing so, Aaron learned a lot about the importance of communication, as he needed to explain to the team was technically possible, while they had to explain to him the legal definitions of the objective.
Despite being competitive, everyone was very open and friendly to each other. Teams would often discuss problems they were having and discuss solutions.
“It was a weird amalgamation of engineers, scientists, lawyers, and business people, all focused on the mission of solving their representative problem”
Aaron definitely recommends competing in GovHack and says that one should be open to being on a team where the skills are incredibly diverse, especially in an area that you know nothing about.
“If you told me you were you going to pair me with a group of lawyers over engineers, I would have been heavily reluctant. However, in hindsight, it was exactly this mix of law and tech that allowed us to win”
Aaron is returning to Melbourne and plans to become a machine learning engineer locally.