daily net cash flow analysis and forecasting : transition from microscopic to macroscopic stochastic equations

The aim of this study is to enhance our understanding of behavior and improve net cash flow forecasting. the research data comprises daily trial balances spanning one year, gathered from 48 bank branches. to achieve this goal, we assessed four distinct models: geometric brownian, arithmetic brownian, vasicek, and modified square root, across three levels: microscopic, mesoscopic, and macroscopic. subsequently, the geometric brownian model emerged as the most suitable model at the microscopic level. the findings reveal that the geometric brownian motion model excels in accurately simulating net cash flow, meeting the criteria for mean absolute percentage error. additionally, net cash flow forecasting for each time series under investigation was conducted over various forecasting horizons, including 7, 14, 21, 30, 60, 90, and 180-day periods, in line with mean absolute percentage error criteria. another noteworthy outcome of this study is that, based on eight distinct prediction accuracy criteria, the ability of the gbm model to simulate and forecast net cash flow diminishes as the forecasting horizon extends.