Cash flow forecasting: a comparison of classical and machine learning models in the DocFinance case study

Cash flow forecasting plays a crucial role in financial planning, liquidity management, and strategic decision-making. The aim of this thesis is to compare classical and machine learning methodologies for cash flow forecasting through an empirical analysis conducted on real corporate data provided by DocFinance, an Italian fintech company specializing in treasury management solutions. The study evaluates the forecasting performance of DocFinance’s software, which relies on the TBATS model, against selected machine learning algorithms trained on historical time series of customer receipts. The analysis examines differences, strengths, and potential limitations of each approach.