Skip to content
All library documents

Using Multinomial Logit and Probit to Predict Bank Client Activity

Article Quant Q&A · Author: Ana S. H.
Read the original

Summary

The document asks how to predict whether bank customers will remain active, using their account movements, and points toward customer activity or churn analysis as the problem area. It recommends multinomial logit and probit models as statistical methods to investigate, with an econometrics textbook offered as a learning resource.

The response identifies feature selection and data collection as the main practical challenges, while describing model computation as comparatively straightforward. It advises building a model suited to the specific task rather than expecting to find an existing bank model that exactly fits. No dataset, validation results, model specification, or forecasting performance is presented, so the suggestions are an introductory direction rather than an evaluated approach.

Key ideas

  • Customer inactivity can be framed as a prediction problem using customers’ banking movements.
  • Multinomial logit and probit are proposed as candidate econometric methods.
  • Choosing explanatory variables and gathering suitable data are central modeling challenges.
  • The document provides no empirical results or specific model specification.

Tags

StatisticsMachine learning
Full text
# Methods or models to predict activity of clients of a bank


# Methods or models to predict activity of clients of a bank












I'm a Physicist but I'd like to know if there are some methods or models to predict the activity of the clients of a bank. I heard that banks are interested in this sort of analysis so I got curious about it. The thing is I don't know where to find information about this. Is there a name for such type of analysis? Can you recommend any good book to learn such methods?

EDIT (19/11/15)

In particular I'm interested in methods to predict the activity of the clients, say we know their banking movements and we want to know who of them will no longer be active the next month/year, for example.

## Answer by owner (score 1)

https://quant.stackexchange.com/a/22012

I would recommend you the following econometrics textbook Basics Econometrics, with a particular focus on `multinomial logit / probit` models. I guess the challenging part in your case will consist of specifying the `exogenous` variables, collecting data, before doing the computations. The latter being quick to perform. As far as I am concerned it's better to design your own model based on the suggested statistical techniques rather than hoping to find on Internet existing bank models which deals `stricto sensu` with this specific task. Hope it helps

Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.