Neural Network Prediction Analysis: The Financial Distress Case

Abstract

This study aims to to test internal financial factor company to financial distress in the year 2018 using data historical 2012-2018. The independent variable taken of a set the initial of du Jardin variables in 2010. Secondary data was used in the study with 32 mining firms that listed on the Indonesia Stock Exchange in 2012-2018 with purposive the sampling method of. A prediction done by means of a utensil artificial analysis the skill of artificial neural network. The application of a method of neural network used backpropagation algorithm. Architecture neural network used 3 layers (1 for input layer, 1 for hidden layer, 1 for output layer). The activation function used a logarithm sigmoid (logsig). The value of mean square error (MSE) training a network is 0,001. The results of forecasting neural network that there is no financial distress in 2018 with accuracy of prediction reaches 84,375%. For the next researcher, it is expected to use several models and several sectors as a comparison so that it can be proven which model is the most accurate in predicting symptoms of financial distress.