| dc.description.abstract | El Nino and La Nina cause changes to period of the rainy season in Indonesia. The condition causes negative effect directly on agriculture and indirectly on economy. In agriculture, these symptoms lead to crop failure resulting in increased food prices so that people's purchasing power decreases. One solution to reduce the negative impact of El Nino and La Nina is by predicting the monsoon season every year so that the government can determine the right types of plants according to seasonal conditions in a given year. This study aims to predict the beginning of the rainy season using feedforward neural networks. In general, feedforward neural network is used in the forecasting process. SOI data is used as a predictor and backpropagation is used as a neural network learning algorithm. We tested four groups of neural network architectures called Group 5, Group 10, Group 15, and Group 20 based on the number of neurons in the hidden layer. This study uses a leave-one-out cross-validation to validate forecasting models and simulation to see the results of the forecasting model. At the leave-one-out cross-validation, the best result is achived by Group 20 with an RMSE of 0.22 dasarian and a correlation coefficient of 0.99. In the simulation of forecasting the best result is obtained by Group 10 with an RMSE of 1.4 dasarian and a correlation coefficient of 0.8. We conclude that the best neural network architecture for predicting the beginning of the rainy season is Group 10. In addition, SOI month, suitable for predicting the beginning of the rainy season is June, July, and August. | en |
