Peramalan Pasokan Tandan Buah Segar dan Penjualan Minyak Sawit Kasar Menggunakan Jaringan Syaraf Tiruan Propagasi Balik

Abstract

Crude palm oil sales forecasting and palm oil fresh fruit bunch stock forecasting are common problems for palm agro industry company. The low rate of sales due to wrong prediction may influence company’s expansion plan. Moreover, crude palm oil stock is strongly dependent on palm oil fresh fruit bunch stock in each fields. The objective of this study is to create a model for sales forecasting of crude palm oil by using backpropagation Neural Network. This study uses secondary data which contains historical data of crude palm oil sales dependent on TBS stocks from 3 fields (inti, plasma, luar). The data is obtained from PT. Perkebunan Nusantara (PTPN) XIII term of January 2005 to December 2007. Factors that influence backpropagation neural network model are the number of neurons in the hidden layer and the learning rate. The conclusion obtained based on the result test is that the level of accuracy of the Backpropagation Neural Network in predicting the supply of Palm Fresh Fruit Bunch and Crude Palm Oil sales is more competitive than the accuracy of ARIMA time series data model for Palm Fresh Fruit Bunch and Crude Palm Oil sales with 750.45 for RMSE and 25.67 MAPE.