0
$\begingroup$

I have historical transaction information of customers for the last 2 years and other information about the customers like what type of card (gold/platinum) they used for transactions etc. is also there in the dataset. Using this dataset, I will have to forecast the likelihood of each customers transacting next month. What are the different approaches that I can analyze before I choose one?

EDIT

More info on the problem:

I have customer credit information and transaction history of the cards for an online ticket booking portal. From this portal customers can book flights, cruises and cars. I have to predict the customers who are most likely to book something (flights, cruises and cars) in the next month. Now what possible approaches can I take for this?

$\endgroup$
0
$\begingroup$

Forecasting is about predicting a variable as a function of time, such as the number of sales during coming months). Your problem is a regression problem, where the output is between 0 (no chance) and 1 (certain).

Any method to solve a regression problem is valid, but I would recommend starting with the simplest thing you think may work, to establish a baseline performance that you can try to improve.

This problem is often phrased as churn prediction, which is just the opposite chance from what you need (although definitions of churn vary), so you can google and see how people do that.

One easy approach that allows you to calculate this is to use the RFM model, where R, F and M stand for recency (time since last purchase), frequency (of purchases) and monetary (total money spent). You could use these features and a logistic regression as your baseline model.

$\endgroup$
  • $\begingroup$ Edited my question for more info. Could you please have look now? $\endgroup$ – Mrinal Aug 27 '19 at 10:42
  • $\begingroup$ I'm not sure how the information you added changes my answer. As I suggested, I would ignore the type of card they used for now, construct the R, F and M features and run a logistic regression on those features to start. If that works well, you're done. If it doesn't, you need to understand why not, and move from there, but that would be a new question, probably, because it can't be answered in general. $\endgroup$ – Paul Aug 27 '19 at 12:47

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.