Scholarlink

Communities and municipalities struggle to handle the enormous volumes of waste they generate. Significant amounts of waste organic matter need to be processed using environmentally friendly methods that support circular bioeconomy and also maintain the cleanliness of the ecosystem. Several bioprocesses, including fermentation, anaerobic digestion (AD), microbial conversion, and biochemical processes, can convert waste into energy. Among these, AD has relatively few disadvantages when it comes to producing material on a big scale over a long time. The modeling of the AD process is challenging since it is a complex and nonlinear process. The present study employs data-driven prediction models for the biogas production process using three distinct machine learning techniques: random forest (RF), k-nearest neighbors (kNN), and linear regression (LR). By comparing the observed and expected biogas yield values, the performance of these models was evaluated. With a mean squared error of 2,077,796.93 (using training data) and 1,082,301.5 (using testing data), the RF model outperformed the LR and kNN models. The MAPE values for the RF model were 10.88% during testing but 8.98% during training. Taylor diagrams provide more proof that the RF model outperformed the other two approaches throughout the training and testing phases.

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

Random Forest Machine Learning-Based Prediction of Biogas Synthesis for Anaerobic Co-digestion of Organic Matters

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Huu Cuong Le,
  • Duc Chuan Nguyen,
  • Van Quy Nguyen,
  • Van Huong Dong,
  • Prabhu Paramasivam

摘要

Communities and municipalities struggle to handle the enormous volumes of waste they generate. Significant amounts of waste organic matter need to be processed using environmentally friendly methods that support circular bioeconomy and also maintain the cleanliness of the ecosystem. Several bioprocesses, including fermentation, anaerobic digestion (AD), microbial conversion, and biochemical processes, can convert waste into energy. Among these, AD has relatively few disadvantages when it comes to producing material on a big scale over a long time. The modeling of the AD process is challenging since it is a complex and nonlinear process. The present study employs data-driven prediction models for the biogas production process using three distinct machine learning techniques: random forest (RF), k-nearest neighbors (kNN), and linear regression (LR). By comparing the observed and expected biogas yield values, the performance of these models was evaluated. With a mean squared error of 2,077,796.93 (using training data) and 1,082,301.5 (using testing data), the RF model outperformed the LR and kNN models. The MAPE values for the RF model were 10.88% during testing but 8.98% during training. Taylor diagrams provide more proof that the RF model outperformed the other two approaches throughout the training and testing phases.