MODEL PREDIKSI KEPAILITAN BANK UMUM DI INDONESIA MENGGUNAKAN ALGORITMA BACKPROPAGATION

Abstract

Abstract: The early warning model was built using 13 CAMEL ratios. The method used is the neural network with backpropagation. Networks are built using a hidden layer and bipolar sigmoid activation function. In this study, the training phase trials performed 12 times by combining the rate of learning rate and iteration to find the best network model. The value of learning rate is 0.1, 0.3, 0.5 and 0.7 combined with 1000, 2000 and 5000 iterations. The result found that the combination of learning rate 0.7 and iteration 2000 as the best model with 100% accuracy and computational time 21 seconds. The resulting output value compared to the actual status of the bank. As a result, the network model is able to produce an accuracy of 86.11%.

Keywords: bankruptcy prediction, commercial Bank, CAMEL, neural network, and backpropagation.