Modelling corporate bank accounts

We discuss the modelling of corporate bank accounts using a proprietary dataset. thus offer principled treatment genuine industrial problem. The in our study constitute spare, irregularly-spaced time series that may take both positive and negative values. builds on previous models where underlying is real-valued. describe an intra-monthly effect identified by practitioners whereby account uncertainty typically lowest at beginning end each month highest middle. However, theory also allows for opposite to occur. In-sample applications demonstrate statistical significance hypothesised monthly effect. Out-of-sample forecasting 9% improvement compared standard SARIMA approach. • under-explored academically. solution uncover behavioural effects explicitly tied management company accounts. use SDE model develop methodology. Results compare favourably with machine-learning