Scholarlink

The steel sector is a prime part of the world infrastructure and economic development. Since the steel industry contributes 8% to global energy consumption and 7% of the energy-related carbon dioxide, there is a practical and ethical need to estimate the energy consumption in the current sustainability-oriented society. The current methods involved the use of the imbalanced data and machine learning models such as Stacking, Voting, Categorical Boosting (CatBoost), CatBoost Bagging, eXtreme Gradient Boosting (XGBoost), Random Forest, Decision Tree, Light Gradient Boosting Machine (LightGBM), and Extra Tree algorithms. The proposed approaches employ machine and deep learning algorithms such as Support Vector Machine (SVM), CatBoost, XGBoost, Stacking, Voting, Bagging, Multilayer Perceptron (MLP), TabNet and Temporal Fusion Transformer (TFT) have the R2 score values of 0.6711, 0.7504, 0.7593, 0.9991, 0.9991, 0.9681, 0.4536, 0.8669 and 0.7168 on the imbalanced dataset. The stacking and voting have the best value of R2 at 0.9991. On the balanced dataset, the suggested models such as SVM, CatBoost, XGBoost, Stacking, Voting, Bagging, MLP, TabNet and TFT have the R2 score values of 0.7211, 0.9713, 0.9852, 0.9957, 0.9892, 0.9767, 0.5748, 0.8733, and 0.8853 respectively. The stacking has the maximum value of R2 Score of 0.9957.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

An Analytical Comparison of Energy Utilization in the Steel Sector Through the Application of Machine and Deep Learning Models

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Satwik Agrawal,
  • Rachit Joshi,
  • Srinivasa Gupta Nagarajan,
  • Valarmathi Balasubramanian

摘要

The steel sector is a prime part of the world infrastructure and economic development. Since the steel industry contributes 8% to global energy consumption and 7% of the energy-related carbon dioxide, there is a practical and ethical need to estimate the energy consumption in the current sustainability-oriented society. The current methods involved the use of the imbalanced data and machine learning models such as Stacking, Voting, Categorical Boosting (CatBoost), CatBoost Bagging, eXtreme Gradient Boosting (XGBoost), Random Forest, Decision Tree, Light Gradient Boosting Machine (LightGBM), and Extra Tree algorithms. The proposed approaches employ machine and deep learning algorithms such as Support Vector Machine (SVM), CatBoost, XGBoost, Stacking, Voting, Bagging, Multilayer Perceptron (MLP), TabNet and Temporal Fusion Transformer (TFT) have the R2 score values of 0.6711, 0.7504, 0.7593, 0.9991, 0.9991, 0.9681, 0.4536, 0.8669 and 0.7168 on the imbalanced dataset. The stacking and voting have the best value of R2 at 0.9991. On the balanced dataset, the suggested models such as SVM, CatBoost, XGBoost, Stacking, Voting, Bagging, MLP, TabNet and TFT have the R2 score values of 0.7211, 0.9713, 0.9852, 0.9957, 0.9892, 0.9767, 0.5748, 0.8733, and 0.8853 respectively. The stacking has the maximum value of R2 Score of 0.9957.