Gum Rosin Price Forecasting Using A Hybrid ARIMA - LSTM Model

The agricultural sector is one of Indonesia's economic strengths. It can be seen from the various kinds of agricultural commodities. As the largest agrarian country in the world, Indonesia can produce agricultural products up to 14.7% - 15% of the total gross domestic product. One of the largest agricultural products in Indonesia is gondorukem (Gum Rosin). Rosin is the sap from this pine tree. Gum rosin exports for 2012 were in response to the increasing international market demand and prices tended to rise from 1,300 USD to 1,500 USD/ton. On January 1, 2019, the price of this commodity became 1,050 USD/ton. It shows that the price of agricultural commodities is very volatile. With this condition, there is concern from both buyer and producer community. There is a difficulty in buying and selling gum rosin. To overcome this, forecasting is applied to predict the estimated price of this commodity. The forecasting model of Autoregressive Integrated Moving Average (ARIMA) and Long-Short Term Memory (LSTM) are used. In terms of performance, the ARIMA in this case is better than the LSTM. While the Hybrid ARIMA-LSTM performs better than the single method if we allow more data training included.