<![CDATA[Agricultural commodity prices fluctuations often cause losses, especially for middle to lower-class people. Predicting future prices is essential for formulating policies for each entity related to agricultural commodities. This study aims to develop a short-term forecasting model for the prices of ten commodities using a machine learning-based method. The predictor variable used was the keyword search index from Google Trends, which represents the public interest in the related commodity. The determination of keywords was done using the web scraping method on various news channels in Indonesia. Before entering the modeling stage, all variables undergone pre-processing and optimal lag selection. The alternative models used were linear regression, random forest, gradient boosting, and regression trees, which were several models based on machine learning. Model implementation measured using the MAPE score has accurate performance with seven out of ten commodities having a value below 10%. Compared with previous studies, the proposed model performs better with lower MAPE values. The proposed model is expected to be a new input for policy-making by the government, business players, and consumers related to agricultural commodities to help reduce price fluctuations.]]>
Copyrights © 2023
