Predicting Company Bankruptcy
=============================
Summary
-------
This project is about predicting the bankruptcy of companies
given a set of financial information about the company. The
project uses data collected from the Taiwan Economic Journal
within the 10-year period starting in 1999. The team seeks to
use the results to develop a wider understanding of financial
statement analysis.
Things I liked
--------------
1. There were definitely a lot of really good visualizations
used in the report! They made understanding the data significantly
easier.
2. The model methods were explained very well, as were the
diagrams which corresponded to them.
3. In the model explanations, the upsides and downsides insofar
as the use of the model was concerned (e.g., false positives for
higher class weights) were well-explained.
Areas for improvement
---------------------
1. The correlation heatmap for all 96 features was pretty hard
to read; the one with only 9 features would probably have been
sufficient (and a better use of space). Also, it would have made
sense to not include very many figures detailing statistics about
the data, as those could have been included in the project proposal
rather than the final report.
2. Although fairness was discussed, it was not done through the
explicit lens of whether or not the model constituted a "Weapon
of Math Destruction".
3. In the Conclusion and Future Work section, there aren't really
any specific recommendations given regarding what businesses should
do to prevent themselves from going bankrupt, or any other
recommendations (e.g., to someone looking to short a company's stock)
to more broadly contextualize the work done in the report in light
of the results.
Overall great job!