USING MACHINE LEARNING TO PREDICT PROCESSING TIMES IN AN ENGINEER TO ORDER ENVIRONMENT: A CASE STUDY

The Engineer-to-Order (ETO) context is characterized by high demand variability and uncertainty, making it particularly challenging to predict processing times accurately at the proposal stage. This study developed predictive models to estimate processing times using real historical data from an ETO company specialized in the production of highly complex equipment. A complete data science pipeline was implemented, encompassing data preprocessing, exploratory analysis, and predictive algorithm modeling. Sixty models were generated by combining data encoding strategies with machine learning algorithms, supported by statistical analyses of data distribution and goodness-of-fit tests. Although predictive performance was limited, the study yielded valuable insights into challenges specific to the ETO context—such as the distinction between lead time and processing time, limitations of product groupings, data variability, and the difficulty of developing generalized models—thereby contributing to the advancement of ML applications in project-driven production environments, which are still underexplored.