Predicting and optimizing the liquidity required by branch ATMs using artificial intelligence

Purpose of this corpus is too optimize and predict required cash for atm devices among KESHAVARZI bank branches with in QAZVIN Province, with artificial intelligence algorithm depending on risk appetite. To achieve this purpose 2 years of daily transaction belonging to KESHAVARZI BANK branches data has been used. Which data were indicating balance, financial daily book, treasury fund and transactions and ATM transactions. The procedure of cleaning, standardization was done before the data sent to linear regression algorithms, the algorithms used in this approach are linear regression, lasso regression and ridge regression. After calculating MSE error, algorithm chose the best model with less MSE error using python programming language Also Jupiter notebook has been used as IDE This research indicates that there are a reverse meaningful relationship between cash balance and risk appetite and opportunity cost. Also a straight meaningful relationship between daily cash flow and debt-credit (which has been provided by the bank and costumers). Features and data that has been used for training an AI model revaluate again with a different values that has been also generated by AI in next step.Results shows that using artificial intelligence can predict daily cash needs with 95% accuracy with an acceptable error.