Integrating multi-criteria methodology with symbolic regression on loan modelling in banking sector

Forecasting loans accurately is essential for the banking sector as it underpins effective risk management, capital allocation, and portfolio optimization. This study aims to model loans in the Turkish banking sector by integrating symbolic regression with multi-criteria decision-making methodologies. Monthly data from January 2004 to September 2024, derived from banks? financial statements, are utilized for the analysis. The optimal parameter configuration for symbolic regression is determined using the TODIM (an acronym in Portuguese for Interative Multi-criteria Decision Making) methodology. The forecasting performance of symbolic regression is evaluated against established models, including Autoregressive Integrated Moving Average (ARIMA), Gaussian Process Regression (GPR), Support Vector Machines (SVM), Neural Networks (NN), Regression Trees (RT), and Long Short-Term Memory (LSTM) network models. The proposed approach is applied across private, public, and foreign banks, as well as the overall banking sector. A significant finding of this study is the identification of a robust relationship between loans and two critical variables: assets and deposits. These results underscore the importance of strengthening deposit mobilization strategies and enhancing asset utilization to effectively grow banks? loan portfolios.