Machine Learning‐Based Prediction of Supercapacitor Capacitance for MgCo2O4 Electrodes

Abstract

AbstractElectrode materials are essential in the electrochemical process of storing charge in supercapacitors and have a significant impact on the cost and capacitive performance of the final product. Hence, it is imperative to make precise predictions regarding the capacitance of electrode materials in order to further the development of supercapacitors. MgCo2O4, with a theoretical capacitance of up to 3122 F g−1, holds immense research value as an electrode material. The objective of this study is to predict the capacitance of MgCo2O4 with high accuracy. This will be achieved by extracting numerous data from published papers and using some parameters as input features. The Recursive Feature Elimination (RFE) method was employed, using Random Forest (RF), Extreme Gradient Boosting (XGBoost) and Regression Tree (RT) as selectors to identify the optimal feature subset. Then, combining them with these three regression models to construct nine machine learning (ML) models. After performance evaluation and outlier analysis, the XGB‐RFE‐XGB model achieved R‐squared (R2), root mean squared error (RMSE), and mean absolute error (MAE) of 0.95, 111.83 F g−1 and 68.25 F g−1, respectively, demonstrating its stability and reliability. Therefore, the XGB‐RFE‐XGB model can be used as a reliable predictive tool in subsequent experimental designs.