Optimization of illuminance sensor placement and combinations for indoor illuminance distribution prediction using XGBoost

<p dir="ltr">This study evaluates the impact of feature selection methods on XGBoost-based indoor illuminance prediction. Over 335,000 measurements were collected from a real office space in Seoul, and three algorithms—Forward Selection, Backward Elimination, and Brute-force Combination—were applied. Backward Elimination achieved the best RMSE for E1 (484.81) and E3 (14.38), while Brute-force performed best for E2 (145.44). Results highlight that optimized variable subsets significantly improve model accuracy. Findings provide insights for data-driven lighting control in smart buildings.</p>