Backpropagation for Predicting the Share of Share Index Industry in Asean
Agnes Novita · International Journal of Progressive Sciences and Technologies (Medical University Varna) · 2018
This study aims to predict the index of stock exchanges in several ASEAN countries represented by Indonesia, Malaysia, Singapore, Thailand, the Philippinesfor the period of 2014. This research usesa Backpropagation Neural Network method. Data used for Input in this research is the daily data fromtheindex of the stock exchange, and the data used is the stock closing price. The amount of data used is 213, this data is divided into 2 (two) parts, which arethe training dataas much as 200 data and the data testing as much as 13 data. The data training itself is divided into 2 (two) groups, the first group uses 1 hidden layer, epoch = 300, learning rate = 0.5, momenteum = 0.6, the function used is tansig and purelin, the neuronsused are1 2 3 5 7 10 12 15.The second group uses 1 hidden layer, epoch = 500, learning rate = 0.3, momenteum = 0.5, the functions used are tansig and purelin, the neurons used for the two groups are 1, 3, 5, 7, 10, 12, 15. At the end of the research, the smallest RMSE (Root Mean Square Error) value from the second group traning is 0.079388,while the smallest RMSE value for testing from the first group is0.392098.