Prediction of pellet quality through machine learning techniques and near-infrared spectroscopy