2
$\begingroup$

I have two main options to calculate PD of a loan in a commercial bank; with and without machine learning.

On one hand, there are traditional methods such as Merton or KVM. On the other hand, I could use machine learning with random forests.

What are the pros/cons of using vs. not using machine learning? Are there any regulations inclined on one or the other?

$\endgroup$
1
+50
$\begingroup$

Merton model has been highly criticized in academic literature for its accuracy, though it provides good ranking of credit risk, it fails to quantify it. I'd say use machine learning or better yet deep learning. I used a recurrent neural network with time series inputs like amount due and changing monthly income among many more. It provided a good estimate but modelling credit risk is not just a quantitative task, black box models can only get you so far unless you understand the business you are investing inside-out

$\endgroup$
  • $\begingroup$ Can you explain how you calculated PD with neural networks? $\endgroup$ – ps0604 May 28 at 13:52
  • $\begingroup$ Hey, sorry for the late reply, I used a time series input of monthly credit data of various customers including dynamics variables like amount due, days since last payment, current CIBIL score and some business specific variables that i cannot disclose. I also used static variables like yearly income, number of dependents etc. Finally i fit a LSTM ( modified version of recurrent neural network) to predict a binary outcome : Default vs No-default and the probabilities associated with it for each customer in training and testing data with accuracy of 95%(better measure is AUC score which was 90% $\endgroup$ – Dhruv Mahajan Jun 24 at 14:14
  • $\begingroup$ You can find tutorials on LSTM on the internet as it is a very popular technique and there isn't much specific case restriction in case of default prediction so you should be able to generalize easily. $\endgroup$ – Dhruv Mahajan Jun 24 at 14:15
  • $\begingroup$ Thanks, good stuff. Have you worked on predicting liabilities balances (i.e. bank deposit balances) using ML/DL? $\endgroup$ – ps0604 Jun 24 at 20:32
  • $\begingroup$ No, I have not worked on those $\endgroup$ – Dhruv Mahajan Jun 24 at 23:13

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.