Scholarlink

<p>This paper aims to develop machine learning (ML) models to predict bio-based supercapacitor energy density from electrode physical and elemental composition, as well as potential window (PW), to accelerate materials discovery and reduce the resources and time required for experimental optimization. Extreme gradient boosting (XGB), random forest (RF), extremely randomized tree (ERT) and categorical boosting (CB) ML algorithms were employed to develop predictive ML models. The training datasets for ML models are extracted from experimental data from existing published articles on bio-based supercapacitors. XGB, RF, ERT and RF models’ performance was evaluated with coefficient of determination (<i>R</i><sup>2</sup>), root-mean-square error (RMSE), mean absolute percentage error (MAPE) and mean absolute error (MAE). Among the models, XGB had the best performance with <i>R</i><sup>2</sup> value of 0.885, followed by CB, RF and ERT at 0.841, 0.781 and 0.637, respectively, when tested with new experimental data. With testing data, XGB recorded the lowest MAE of 4.81&#xa0;Wh/kg, ERT, CB and RF recorded MAE of 9.23&#xa0;Wh/kg, 6.51&#xa0;Wh/kg and 8.14&#xa0;Wh/kg, respectively. Shapley additive explanations (SHAP) analysis indicated that PW exhibited the highest impact in all models used to predict energy density as compared to other independent variables, such as specific surface area (SSA), intensity ratio (IR) and average pore size (PS), emphasizing its critical influence in developing predictive models and enhancing supercapacitor energy density. The impact of PW in predicting energy density of bio-based supercapacitors can be attributed to its quadratic relationship with energy density. Consequently, the models accurately predict bio-based supercapacitor energy density, demonstrating the role of PW in designing and optimizing bio-based materials for the development of sustainable supercapacitors.</p>

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

Machine learning-based energy density prediction of bio-based supercapacitor with Shapley additive model interpretation

  • 期刊论文

  • 发表日期：

    2026年4月18日

查看全文
  • Damilare Samuel Oyebamiji,
  • Davannendran Chandran

摘要

This paper aims to develop machine learning (ML) models to predict bio-based supercapacitor energy density from electrode physical and elemental composition, as well as potential window (PW), to accelerate materials discovery and reduce the resources and time required for experimental optimization. Extreme gradient boosting (XGB), random forest (RF), extremely randomized tree (ERT) and categorical boosting (CB) ML algorithms were employed to develop predictive ML models. The training datasets for ML models are extracted from experimental data from existing published articles on bio-based supercapacitors. XGB, RF, ERT and RF models’ performance was evaluated with coefficient of determination (R2), root-mean-square error (RMSE), mean absolute percentage error (MAPE) and mean absolute error (MAE). Among the models, XGB had the best performance with R2 value of 0.885, followed by CB, RF and ERT at 0.841, 0.781 and 0.637, respectively, when tested with new experimental data. With testing data, XGB recorded the lowest MAE of 4.81 Wh/kg, ERT, CB and RF recorded MAE of 9.23 Wh/kg, 6.51 Wh/kg and 8.14 Wh/kg, respectively. Shapley additive explanations (SHAP) analysis indicated that PW exhibited the highest impact in all models used to predict energy density as compared to other independent variables, such as specific surface area (SSA), intensity ratio (IR) and average pore size (PS), emphasizing its critical influence in developing predictive models and enhancing supercapacitor energy density. The impact of PW in predicting energy density of bio-based supercapacitors can be attributed to its quadratic relationship with energy density. Consequently, the models accurately predict bio-based supercapacitor energy density, demonstrating the role of PW in designing and optimizing bio-based materials for the development of sustainable supercapacitors.