Predicting accounting liquidity at Dutch SMEs using Support Vector Regression and Multilayer Perceptrons

Abstract

Small to medium-sized enterprises (SMS) are critical for the global economy, but face challenges in liquidity management due to limited access to external financing and market uncertainties. This study systematically evaluates the effectiveness of multilayer perceptrons and support vector regression models in predicting liquidity for SMEs. Through multiple experiments conducted on a dataset of 496 companies, the study reveals potential for both models, although they exhibit a large number of limitations and sensitivity to data quality and size. Notably, the MLP model demonstrates a closer alignment to target values compared to SVR. While the models are not suitable for practical use, the findings highlight the importance of refining both models to improve liquidity forecasting. This research contribute to more effective decision making processes, benefitting long-term success and sustainability of SMEs as contributors of the economy.