Trying to think of ways you could use exisiting machine learning techniques on a system-wide scale or individual bank level for the banking industry. Any ideas?
Are you talking about risk of fraud, risk of credit default or anything else? Either way there are already quite a few companies doing this at very large scale, which shouldn't discourage you - just means that you can find some good resources if you dig deep enough.
I meant starting moreso with what they have on their books. Predicting problematic positions, only a small scale or on large scale. Or some sort of learning using this clean and well-defined data.
No idea how many data points you'd need in order to get a reasonably good model but the task at hand would be pretty daunting. After all you need to consider individual risk posed by employees all the way up to systemic and political. On the other hand if successful you'd hit a home run.
Yeah I guess I would focus on the base case of using it on the options, fixed income, etc that the bank has position information on (x100000000 positions). Just because all that data is clean.
