The integration of the world economy, the removal of boundaries in financial markets, and the spread of financial crises have led to a more pronounced impact. In order to mitigate the negative effects of crises and implement policies promoting economic stability, an Artificial Neural Network-based Early Warning System has been developed using artificial intelligence techniques. In this study, monthly data for the period from January 1992 to December 2022 was used for 14 variables commonly employed in leading indicators literature. Crisis periods were identified using the Financial Stress Index, and crisis periods overlapping with financial crises in Turkey were determined. The Early Warning System, created with the Multilayer Perceptron model from Artificial Neural Networks, was tested based on training and validation performances. Among the models created, the highest performance was achieved by a model with the trainlm backpropagation algorithm, tanh and softmax activation functions in the hidden layers, and a purelin activation function in the output layer, accurately identifying crisis and normal periods with 100% accuracy.
Key words: Artifccal Neural Network, Early Warning Systems, financial crises, leading ındicators, Financial Stress Index. JEL Codes: G01, C40, C45. Article Language: EnglishTurkish
|
