Prediksi Data Time-series menggunakan Jaringan Syaraf Tiruan Algoritma Backpropagation Pada Kasus Prediksi Permintaan Beras

Abstrak

Recently, Indonesia, as a country where the majority of the population chooses rice as the primary food source, gets a decline in the rice consumption patterns, which resulted in reduced demand for rice that should have been stable. The decrease of rice purchasing power impacts several rice suppliers, commonly referred to as rice agents, to buy rice from rice production companies. Therefore, prediction of rice stock is essential to do. This paper aims to apply the backpropagation neural network method to forecast the amount of rice demand. The data used in the study is time-series data in the form of the number of requests for rice as much as 609 data from two types of rice. The modeling scenario in this study applies one to five hidden layers with a different number of hidden neurons in each experiment. The elastic net regularization method was applied after the data denormalization process to improve the quality of the resulting model. Based on the experiments, obtained the best model on architecture 7-50-200-300-250-300-1 with MSE = 0.001278, RMSE = 0.301950 in the training process and MSE results = 0.002391, RMSE = 0.204972 in the testing process.