I need to forecast non-maturity deposits in a bank. My intent is to use Recurrent Neural Networks (aka deep learning) to model time series.

The model will learn from past bank data and macroeconomic variables, and forecast deposits in the next three years based on projected variables.

Two questions:

  1. Do you see a problem with this approach?
  2. I believe that using this technique there's no need to calculate core deposits level to forecast total deposits, is this assumption correct?
  • 3
    $\begingroup$ I heard of a large insitution that had put many people with PhDs to work on trying to use AI/ML to predict NMD, deposit beta, and the like. Their model performed so badly that the project has been abandoned, lots of people got fired, and the institution is now more skeptical of fads. $\endgroup$ Commented Jan 9, 2021 at 14:38
  • $\begingroup$ Did you also hear what was the reason? $\endgroup$
    – ps0604
    Commented Jan 9, 2021 at 14:43
  • 1
    $\begingroup$ The model's predictions for out of sample periods were too different from reality. $\endgroup$ Commented Jan 9, 2021 at 14:59
  • 1
    $\begingroup$ Some very relevant insight from V. Piterbarg: ci.natwest.com/insights/articles/… $\endgroup$ Commented Jan 26, 2021 at 15:54
  • 1
    $\begingroup$ we used machine learning, neural networks with LSTM to model time series $\endgroup$
    – ps0604
    Commented Oct 13, 2023 at 16:20

2 Answers 2


This boils down to timeseries forecasting which is particularly challenging for financial data because finance exhibits all the things that makes forecasting hard: non stationary, low signal to noise, etc.

In the case of forecasting deposits, I very much doubt that you will be able to get satisfactory results even if you do manage to figure out all the relevant variables. And then you will have a problem called curse of dimensionality because you will never have millions or thousands (or even hundreds) of years of data to train your model.

One thing to be aware about machine learning algorithms in general is that they are (can be) very good at interpolating and pretty bad at extrapolating, which means your model will fail when you need it the most.

So my answer to your question "Do you see a problem with this approach?" is that you might get ok forecasts when is doesn't matter, ie, when deposits are not really changing, but forecasts may fail miserably when you are in unchartered territories.

And this goes for forecasting just a few months in advance... never mind three years.

Regarding the second question, I would try with and without core deposits to see if there is any explanatory power.

A useful starting reference would be tensorflow's tutorial on Time series forecasting which uses Convolutional and Recurrent Neural Networks to forecast weather. You could try to use it with your data to forecast out of sample periods.


  • 1
    $\begingroup$ i agree with answer and understand why modelling NMD with ML will either be very difficult or yield no results. I would rather custom build a model and try to fit it. $\endgroup$
    – Attack68
    Commented Jan 11, 2021 at 18:58
  • $\begingroup$ I read in a Moody’s paper that they use regression analysis to forecast bank deposits, how is that better/more accurate than RNN? $\endgroup$
    – ps0604
    Commented Feb 6, 2021 at 15:08

Responding to the first question and building on Dimitri's comments, it is probably worth your while to do some due-diligence on the some of the standard issues that come up in Financial ML and see to what extent they apply to your problem:

  1. Interpretability: How will you interpret the results of your Deep Learning model and share them with users/decision makers?
  2. Data requirements: Deep learning algorithms are data hungry. Do you have enough data to sufficiently calibrate the model? Since historical data sets are rarely adequate, do you have a robust simulation process to generate synthetic data and thus get a sense for how the model performs across various economic scenarios?
  3. Signal-to-noise ratio: Financial data sets are notorious for having low signal-to-noise ratios. Is your ML algorithm appropriately accounting for macro changes in industry structure, regulatory actions etc.
  4. Over/under fitting: How do you control for time series effects in constructing your training, validation and testing data sets?

Of course, these issues are not unique to ML models. You may find it useful to construct a baseline "adversary" for your model using some standard econometric approach and then compare and contrast how well it performs with respect to the above considerations (among others). You also don't have to get locked into an "either-or" approach, model ensembles often give better results that any single individual model although we get dragged back into issues of interpretability again.

Although its specific focus is on asset pricing, you may find it useful to peruse de Lopez de Prado's book Advances in Financial Machine Learning which, apart from referencing these issues and discussing how to deal with them, also talks about how to structure a Financial ML project.

  • $\begingroup$ thanks for putting together your answer, however it applies to machine learning models in general. The question is specific: can deposits be modeled with Recurrent Neural Networks? By the way, thanks for pointing out the book, it's very interesting. $\endgroup$
    – ps0604
    Commented Jan 10, 2021 at 8:39
  • $\begingroup$ I am not familiar with any specific study of this topic but the reason I tried to answer this question in a general way is that, based on my experience, in a business context there is unlikely to be a simple "yes-no" answer. Instead, one typically has to consider a number of tradeoffs, some of which are referenced above. In retrospect, I read question #1 more generally than you intended it to be. $\endgroup$
    – Sharad
    Commented Jan 10, 2021 at 22:02

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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