Crude Palm Oil Price Forecasting Based on ECoS-MARS: in Data Science Models

The crude palm oil (CPO) industry is highly competitive due to fluctuating and unpredictable prices. To accurately predict future CPO prices, a forecasting technique is necessary. This can be achieved through various regression techniques and data mining approaches. This paper proposes a Multivariate Adaptive Regression Splines (MARS) model, which utilizes the ECoS machine learning algorithm to obtain optimal forecasting values. This research mainly focuses on forecasting CPO prices using the ECoS-MARS approach. The functional basis (BF) validation has yielded an accuracy of 0.0532%. Furthermore, validation has been conducted to achieve accurate forecasting values, resulting in an accuracy of 0.11714%. The best parameter combination for this forecasting model consists of a learning rate of 0.6 for both learning rates 1 and 2, a sensitivity threshold of 0.3, and an error threshold of 0.05.