Abstract This study introduces ADCO, a hybrid model combining adaptive boosting regression (ADAR) with electric charged particles optimization (ECPO), designed to enhance predictive accuracy in modeling the environmental impact of high-performance computing (HPC) manufacturing. There is an environmental cost associated with the drive to develop HPC systems that are quicker and more potent, including increasing CO 2 emissions and energy use. Understanding and reducing HPC’s environmental effect is essential as its demand rises. This work uses ADAR and light gradient boosting regression (LGBR) to examine energy usage and CO 2 emissions in HPC manufacturing. smell agent optimization (SAO) and electric charged particles optimization (ECPO) are employed for hyperparameter optimization to achieve high prediction accuracy and model convergence. The results show that the improved models work better and provide insightful information for cutting emissions and energy use in the HPC sector. These results support the creation of more ecologically friendly production methods in a rapidly evolving technological environment. ADCO excelled in predicting energy trends with Test RMSE 50.999 and R 2 0.986, while ADA struggled, showing higher Test RMSE 79.882 and lower R 2 0.964. The optimal input for forecasting energy consumption is derived from visualizations that focus on both energy consumption and CO 2 emissions, with cement identified as a key feature in each.
Yiming Chen (Wed,) studied this question.