Forecasting natural rubber prices using commodity market indicators: a machine learning approach

Natural rubber is an essential raw material in sectors such as automotive, construction, healthcare and manufacturing. Volatility in natural rubber prices can have a long-run impact on rubber growers and rubber-based industries. This study develops and compares different forecasting models of natural rubber prices based on machine learning algorithms - support vector machine (SVM), artificial neural networks (ANNa), k-nearest neighbours (KNNs), random forest (RF) and decision tree - along with the traditional forecasting ARIMAX method. Natural rubber price forecasting models are developed using a set of explanatory commodity market indicators encompassing macroeconomic factors, demand and supply factors and price of related commodities. Based on our results, we propose a forecasting model of natural rubber prices employing random forest algorithm, which outperformed the other machine learning algorithms in its predictive capabilities. This paper makes substantial contributions to policymakers, businesses and rubber growers in making informed decisions and managing price risk in the natural rubber sector.