Two Different Points of View through Artificial Intelligence and Vector Autoregressive Models for Ex Post and Ex Ante...

Area of Science:

• Economics
• Computational Finance
• Machine Learning

Background:

• Macroeconomic variables significantly influence financial stability.
• Accurate forecasting models are crucial for predicting financial distress.
• Previous studies have utilized various econometric methods for financial forecasting.

Purpose of the Study:

• To apply Artificial Neural Network (ANN) and Vector Autoregressive (VAR) models for financial forecasting in Turkey.
• To compare the predictive performance of ANN and VAR methods using macroeconomic data.
• To identify potential financial distress in Turkey based on model predictions.

Main Methods:

• Utilized multilayered feedforward neural networks (MLFNs) with the ENCOG machine learning framework and JAVA.
• Employed resilient propagation for network training.
• Applied Vector Autoregressive (VAR) method for comparative analysis.

Main Results:

• ANN modeling indicated a potential financial crisis in Turkey starting October 2017.
• VAR method results supported the ANN findings regarding financial distress.
• ANN demonstrated superior prediction performance compared to the VAR method.

Conclusions:

• ANN models offer a more accurate approach to forecasting financial distress than traditional econometric methods.
• Macroeconomic indicators, including USD/TRY exchange rate, gold prices, and BIST 100 index, are key predictors.
• The study highlights the utility of advanced machine learning techniques in economic forecasting and risk assessment.