Machine Learning for Agricultural Price Prediction: A Case of Coffee Commodity in Vietnam Market

Predicting agricultural commodity prices is crucial for farmers, governments, and related stakeholders to make important decisions about crop planning and food security. This is especially important in economies that rely greatly on agriculture, such as Vietnam and other developing countries, where the instability and inadequate predictability of agricultural commodity prices may lead to overproduction, "good season - devaluation" phenomenon, and significant economic losses. Aiming to address this issue, this paper presents the use of machine learning models to predict the price of coffee, a major agricultural commodity in Vietnam. Various machine learning techniques, namely LSTM (Long Short-Term Memory), ARIMA (Autoregressive Integrated Moving Average), SARIMA (Seasonal ARIMA), GRU (Gated Recurrent Unit), SVM (Support Vector Machine), and RF (Random Forest), are trained and evaluated using different data, including historical prices and additional data, such as fuel prices and weather data. The results of numerical experiments demonstrated that using additional data can significantly improve prediction performance.