Efficiency of Direct Labor

This thesis examines discrepancies between planned and actual direct labor hours in the production of automotive seating systems at Lear Corporation and proposes a data-driven forecasting framework to improve labor planning accuracy. The current Excel-based estimation approach relies on static average values and does not reflect changes in production mix, production calendar constraints, or operational conditions. To overcome these limitations, a predictive model integrating actual production mix, time-series dynamics, and seasonality effects is developed. Using cycle-time data from Palantir, production records from the LJS system, and HR attendance data, three machine learning algorithms—Linear Regression, Random Forest, and XGBoost—were evaluated using an expanding-window time-series validation approach in accordance with the CRISP-DM methodology. The results show that the Random Forest model achieves the highest predictive accuracy (R² = 0.809) and demonstrates strong robustness across all evaluated periods. The thesis concludes with the design of an analytical dashboard that supports Controlling in identifying inefficiencies and increasing transparency in direct labor planning within the context of Industry 4.0.