Predicting the Price of Crude Palm Oil: A Deep Learning Approach

Background

The literature on deep learning and on statistical approaches used in forecasting is relatively broad. This study limits the present review to a stream of research that is most relevant to the problem being investigated by the authors.

In recent studies, several authors use a comparative approach in their forecasting models to assess the accuracy of various univariate response variables with a goal of producing more reliable forecasts. Some studies, for example, compared the performance of the ARIMA model to its extensions, while others compared the performance of traditional forecasting methods (e.g., ARIMA or Holt-Winters) to artificial neural networks. Khalid et al. (2018), for example, investigate the performance of many econometric forecasting models, including the ARDL (Autoregressive Distributed Lag), the ARIMA (Autoregressive Integrated Moving Average), and the ARIMA with exogenous inputs (ARIMAX). Their study employs a monthly time series data set from 2008 to 2017 and found that the ARIMAX model outperformed the other techniques (Khalid et al. 2018).

Karia et al. (2012) compare the ANN, ANFIS, and AFRIMA models in order to come out with the most appropriate model for forecasting the CPO prices. Their analysis demonstrates that the ANN model is superior to the ANFIS and ARFIMA models in predicting the CPO prices (Karia et al. 2013).