ATM cash prediction (withdrawal and replenishment)
Budget: $30 – $250 USD
Banks lose quite a lot of money just lying in ATMs. Instead, the money could be successfully invested and earn a return.Thus, the formal statement of the problem looks like this.
At the entrance:
there is a history of withdrawing / replenishing cash at an ATM (in our case, these were data from 2019);
Holiday file where, all features regarding to the date (holidays, weekend, national holidays and etc)
The output is expected:
a system of recommendations on the amount of money and the time for which they need to be pledged;
Quality metric: MAE and MAPE
Based on the forecast or historical data of cash withdrawals/deposits, the amount to be pledged is calculated.
Forecast step one month ahead (determined empirically).
As a result, two models (withdrawal model and replenishment model)
At the entrance:
there is a history of withdrawing / replenishing cash at an ATM (in our case, these were data from 2019);
Holiday file where, all features regarding to the date (holidays, weekend, national holidays and etc)
The output is expected:
a system of recommendations on the amount of money and the time for which they need to be pledged;
Quality metric: MAE and MAPE
Based on the forecast or historical data of cash withdrawals/deposits, the amount to be pledged is calculated.
Forecast step one month ahead (determined empirically).
As a result, two models (withdrawal model and replenishment model)