APPLICATION OF ARTIFICIAL NEURAL NETWORK METHOD IN CURRENCY CRISIS DETECTION IN INDONESIA BASED ON MACROECONOMIC INDICATORS

Abstract

The global crises of 1997 and 2008 that affected various countries had significant impacts on the economies of developing countries in the world, including Indonesia. If not addressed, these impacts could have adverse effects on Indonesia. Therefore, it is necessary to have an early warning system for currency crises to anticipate the negative effects of such crises. This study employs the Financial Pressure Index (FPI) crisis threshold approach with perfect signal value as the dependent variable and 13 macroeconomic indicators as independent variables to develop an early warning model for currency crises in Indonesia using Artificial Neural Network method with Multilayer Perceptron Backpropagation algorithm and adding Adaptive Moment Estimation (Adam) optimization in weight modification process. The testing results on validation and test data show that Adam optimization produces high accuracy, sensitivity, and specificity. Based on the best model, it is found that the period from July 2021 to June 2022 has a perfect signal value of 0, meaning that there will be no crisis in Indonesia from July 2022 to June 2023. In conclusion, this study shows that the Artificial Neural Network method with Adam optimization can effectively detect currency crises in Indonesia