Exploring cement Production's role in GDP using explainable AI and sustainability analysis in Nepal

Abstract

Due to rising demand, the worldwide cement market is expected to increase from $340.61 billion in 2022 to $481.73 billion by 2029. Quarrying, raw material processing, and calcination are steps in cement production. The societies in India and Nepal have to deal with environmental issues such as air pollution, resource depletion, and the effects of climate change. A case study of Nepal's Udayapur Cement Industry Limited (UCIL) exposed antiquated production methods that reduce energy efficiency. Utilizing regression models like Extra Trees (Extremely Randomized Trees) Regressor, CatBoost (Categorial Boosting) Regressor, and XGBoost (eXtreme Gradient Boosting) Regressor, Random Forest and Ensemble of Sparse Embedded Trees (SET) machine learning is used to examine the demand, supply, and Gross Domestic Product (GDP) performance of cement manufacturing in India which shares a common cement related infrastructure to Nepal. Since businesses understand how important sustainability is to attract new customers and minimizing environmental effects, our study emphasizes the necessity of sustainable practices in the cement production industry. On evaluation, the Extra Trees Regressor showed strong performance, along with the SET (Stacking) model, which was further validated using a nested cross-validation technique. Random Forest, on the other hand, had trouble; it displayed the greatest RMSE (15617.85) and the lowest testing (0.8117), suggesting poorer generalization. The SET (Stacking) Ensemble model gained a testing R² score (0.9372) and a testing RMSE (9019.76). In cross-validation, the Extra Trees model with a mean cross-validation R² score of 0.93 and a low standard deviation of 0.04 proved to be the best-performing model, as evidenced by lower differences in R² score across folds compared to other models, demonstrating its high predictive performance. The SHAP (SHapley Additive exPlanations) interpretability analysis indicates that population is the primary factor influencing GDP estimates. A Tkinter-based application was also developed to forecast GDP using the training model. To attain sustainability and lessen the effects of climate change on the cement sector, these findings highlight the adoption of cutting-edge technologies and energy-efficient procedures.