Enhancing Price Classification of Chili Using Gradient Boosting Machines

Abstract

This study explores the application of Gradient Boosting Machines (GBM) to enhance the classification and prediction of chili prices. The research uses a comprehensive dataset collected from various sources, including local markets, online platforms, and agricultural databases, covering multiple attributes such as chili type, region, harvest season, weather conditions, and demand-supply dynamics. The GBM model outperforms traditional machine learning algorithms, achieving an accuracy of 87%, with a high area under the ROC curve (AUC) of 0.91. Feature importance analysis indicates that harvest season and region are the most significant factors influencing price variations. The findings suggest that the GBM model provides reliable price predictions and insights into price-driving factors, offering valuable tools for stakeholders in the agricultural market. The study emphasizes the need for broader data sources and advanced techniques, such as time-series forecasting and XGBoost, to further improve chili price prediction models. These insights can help optimize supply chain management, price forecasting, and decision-making for producers, traders, and policymakers.