Addressing the paradox of data-driven rate of penetration prediction: Challenges and solutions