# Balance Prediction model for each customer

I need to build a prediction model for the amount balance (Bank account) of a customer.

The problem: It should be done for each customer where each customer will have a specific date where the prediction should be made for him.

My data should be like something like this:

cycle_date = the date where the cycle of income and outcomes will end for the customer for a specific date in a month.

For example: the credit card charges are in the 02/07 and the salary enters to the customer bank account in the 10/07. So I will say the cycle_Date for this customer will be 10/07. In general I can say: The max date between credit card charges and salary but I still needs to think about that.

customer cycle_date balance
10002       t1    1000
10002       t2    2000
10002       t3    4000
10002       t4     ?
10003       tt1   10
10003       tt2   200
10003       tt3    ?


I have 30,000 customers. I guess building an arima model for each customer seprately will not be a good idea or even doable. So I will be gratfule reading some of your advices on approching this problem. What will be the correct model to choose? There is a way to address the problem of a specific date prediction for each customer?

Thank you.

• Did you try fitting an ARIMA model with one customer's data? Moreover, could you provide us a sample of the data for one or two customers? – paf Jul 17 '18 at 16:36