Biomass-derived carbon electrodes have emerged as a sustainable and cost-effective alternative to conventional electrodes in lithium-ion and sodium-ion batteries (LIBs and SIBs), offering a promising pathway toward more efficient battery technologies with enhanced electrochemical performance. However, determining the optimal preparation method for biomass-derived carbon electrodes remains challenging due to the complex interplay of various processing parameters and their impact on electrochemical performance. In this context, machine learning offers powerful support by analyzing data to uncover underlying relationships and optimize preparation strategies. In this study, nine prediction models were selected to predict the electrochemical performance of carbonized cellulose and lignin electrodes in LIBs and SIBs through machine learning-assisted training of 288 data sets from the literature, considering different doping methods, carbonization conditions and processing parameters. After model optimization, the results indicate that extreme gradient boosting (XGB) and gradient boosting regression (GBR) models are well-suited to predict the specific capacity after the first and multiple charge/discharge cycle(s), respectively, achieving predicted R 2 values of 0.88 and 0.94. As a result, the carbonization temperature, current density and doping (such as the nitrogen doping, inorganic non-metallic compounds) are identified as the primary factors influencing the capacity after the first cycle. The specific surface area (SSA) plays a significant role in influencing the capacity after multiple cycles. Additionally, the SHapley Additive exPlanations analysis reveals that during the carbonization process, a lower temperature and longer duration are beneficial for the capacity of the carbon electrode. And the larger SSA and more doped materials further enhance the capacity. Furthermore, experimental validation confirms that the XGB and GBR models can accurately predict the specific capacity of the electrodes, demonstrating their reliability in electrochemical performance forecasting. In conclusion, this study provides a useful guideline to assist in the optimized design of high-performance biomass-derived carbon electrodes for LIBs and SIBs.