Steel is the most widely used material globally, with the Basic Oxygen Furnace (BOF) process accounting for over 70% of its production. In the realm of industrial optimization, machine learning has recently emerged as a crucial tool, particularly for leveraging the extensive data generated in industrial processes. This study taps into this potential by employing substantial datasets from BOF steelmaking, ranging from 500 to 20,000 heats. Various machine learning models were developed to predict the end-point temperature of the BOF process, utilizing production line features that closely mimic real-world production scenarios. These features were carefully selected during the feature engineering phase. 

The study presents a thorough comparison of these models. Furthermore, it delves into the effects of dataset size and the integration of realistic production line features on the overall performance of the machine learning models.