Predictive Model to Optimize Material Stock Management Using Machine Learning in a Construction Company

Chamorro B.M.; Ubaldo R.A.; Alarcon J.S.D. · International Conference on Artificial Intelligence, Computer, Data Sciences, and Applications, ACDSA 2026 (2026) · DOI: 10.1109/acdsa67686.2026.11467796

Resumen

In the precast concrete industry, fluctuations in material demand lead to cost overruns and production delays. Although technologies such as RFID, BIM, and computer vision have been incorporated in recent years, their implementation is often affected by the need for operational adjustment and adaptation, resulting in data inconsistencies and delays in inventory control. Within this context, the lack of integration between control systems and production planning prevents accurate forecasting of actual material consumption. Therefore, a predictive model based on Machine Learning algorithms is proposed to optimize stock management and reduce uncertainty in material planning. The model is structured into four stages: (i) dataset collection and preprocessing, (ii) training using Random Forest, (iii) validation using the coefficient of determination (R2), MAE, and RMSE, and (iv) result analysis for deployment at the plant level. The results of the first experiment showed a high R2 value, along with a notable reduction in both MAE and RMSE. This performance indicates that the model achieves a closer alignment between estimated and actual values, resulting in more accurate and stable predictions despite demand variability. Accordingly, the metrics demonstrate a coherent improvement aligned with the study's purpose, enabling more efficient and anticipatory material management within the production process. Allows optimizing stock levels and reducing uncertainty in planning.

Instituciones: Universidad Peruana de Ciencias Aplicadas