Bipolar function in backpropagation algorithm in predicting Indonesia’s coal exports by major destination countries
Bayu Febriadi, Zamzami Zamzami, Yogi Yunefri, Anjar Wanto · IOP Conference Series Materials Science and Engineering · 2018
Coal is one of the most widely used energy sources in the world, and Indonesia is one of the coal exporting countries. Therefore, Indonesia's long-term availability of coal must be maintained, to support various industrial projects and the world economy. One way to maintain coal reserves is to predict the timing data of coal exports, to make it easier for the government to issue a coal export policy. In this study, the prediction method used is back propagation algorithm. The algorithm is able to solve many problems by building a well-trained model that shows good performance in some non-linear problems. The function used is bipolar, because this function is able to calculate data whose value is not stable. The data used in this study is the data of Coal Exports in Indonesia based on the main destination countries processed from customs documents of the Directorate General of Customs and Excise and Statistics Indonesia. This study uses 3 architectural models, namely: 4-5-1, 4-10-1 and 4-15-1. The best architectural model is 4-5-1, yielding 93% accuracy, Margin eror 7%, MSE 0,0117017098 with error rate 0.001 - 0.04. It is expected that these results can predict well.