Enhancing forecast accuracy in the paper scrap sector: a comparative study of ensemble and machine learning models