Fertilizer price forecasting using machine learning: a case study of rice fertilizer in Thailand

This study forecasts fertilizer prices using Decision Tree (DT), Artificial Neural Network (ANN), and Ensemble Models. Data collected from 2020 to 2023 was used to forecast and compare the accuracy between the predicted results and the actual values and each model. The forecasts from 2020 to 2023 serve as a basis for decision-making regarding the selection of appropriate data for future fertilizer price prediction. The experimental results reveal that the Ensemble Model improves forecasting accuracy. Furthermore, the findings indicate that ANN outperforms the DT model in terms of prediction accuracy. This is due to ANN's superior ability to capture complex and nonlinear patterns in data. While DT models rely on simple rule-based splits and often suffer from overfitting, ANN models can generalize better, particularly when working with continuous and time series data. The layered structure of ANN allows it to learn intricate relationships and update its parameters through gradient-based optimization, making it more suitable for dynamic and uncertain environments such as fertilizer price forecasting. However, certain abnormal or unexpected conditions may still lead to forecasting inaccuracies.