Optimization of Genetic Algorithms on Backpropagation Neural Network to Predict National Rice Production Levels

Abstract

Rice is the staple food for Indonesian people that cannot be separated from anything. Every community needs rice for their survival. Annual Indonesia rice production is fluctuating. It needs rice production information to anticipate the rice quota which meets the demands. For this reason, this study purposes to predict rice production in Indonesia. The method used in this study is a Backpropagation Neural Network (BPNN) with genetic algorithm (GA) optimization for time series forecasting. Using Genetic Algorithms, BPNN was optimized then used as weighting input for better predictions, then the prediction results will be used as a benchmark for the next rice production so Indonesian rice demands will be fulfilled. We use the data of rice production published from 1970 until 2017. The results demonstrate that the BPNN model optimized using Genetic Algorithms has a very high performance with Mean Absolute Percentage Error (MAPE) values 1.4084% for training data and 2.3394% for testing data. It is more accurate than the BPNN model without GA optimization, it yields MAPE values 1.4797% for training data and 2.6944% for testing data. Based on the best model, we deliver the prediction of the rice production in the next four years after 2017.