Peramalan Harga Minyak Mentah Dunia (Crude Oil) Menggunakan Metode Radial Basis Function Neural Network (RBFNN)

Abstract

Forecasting is a technique to estimate a value in the future with past data and current data. One of the forecasting method that includes neural network is Radial Basis Function Neural Network (RBFNN). In this research, RBFNN method is used to get the best model and to forecast world crude oil price (US$) data. World crude oil prices forecasting is very important for many stakeholder, both from the government sector, business entities and investors so that all activities can go according to plan. In the RBFNN method, the network input and the number of hidden layers is very influential to get the best model from RBFNN and also the forecasting. To get the best model by using network input determination by identifying the Partial Autocorrelation Function (PACF) lag, and to determine the number of hidden layers by the K-Means cluster method. Results of the research showed that from the training data, the best model of RBFNN is using 2 network inputs X_t−1 and X_t−2 and 3 hidden layers with Mean Absolute Percentage Error (MAPE) accuracy level is 6,8150%. With the model, for the next period from June 2017 to December 2017 the world crude oil price (US $) shows a downward trend.