PERAMALAN JUMLAH PENUMPANG KERETA API DENGAN JARINGAN SARAF TIRUANMETODE PERAMBATAN BALIK (BACK PROPAGATION)

Train becomes one of the most popular mass transportation media in Indonesia, especially for medium and long distance journey. When holiday comes, PT. KAI (persero) need to make much additional departures of train in order to occupy the passanggers need. From this statement, risen the ideas to design a forecasting system to forecast the number of train’s passanger using artificial neural network and to analyze the neural network which gives the best performance in this system. This final project was designed a forecasting system to predict the number of train’s passanger using Back Propagation Neural Network. First step, all data need to be computerized and divided them into 2 groups, 60% for training data and 40% for testing data. When training the neural network, training data is used as the input and it process will stops when the values of required training parameters is reached. Testing the neural network can be done using training data and testing data. This process is aimed for knowing network’s performance by comparing simulation results with real target. Neural network with best performance can be used for forecasting process. Based on training data test results, the best performance neural network given by the network with 10 nodes of hidden layer for variation increasing the number of node of hidden layer and the network with its learning rate was 0,01 for variation decreasing the value of learning rate. Based on testing data test results, the best performance neural network given by the network with 5 nodes of hidden layer for variation increasing the number of node of hidden layer and the network with its learning rate was 0,001 for variation decreasing the value of learning rate Keywords : forecasting, train’s passanger, back propagation