This paper presents a comparative study of machine learning (ML) models for predicting the remaining useful life (RUL) and capacity fade of lithium-ion (Li-ion) batteries. The study evaluates the performance of multiple ML models, including Random Forest (RF), XGBoost, LightGBM, MLP, LSTM, and Attention-LSTM, using two distinct datasets. The first two cases use a synthetic dataset with linear capacity fade, while the third case uses a real dataset with nonlinear battery behavior. The study introduces a multi-feature multi-target (MFMT) feature mapping to enhance the accuracy of RUL prediction. The results show that the XGBoost-HT model, which incorporates hyperparameter tuning, outperforms other models in terms of root-mean-squared error (RMSE) and mean absolute percentage error (MAPE). For the third case, the XGBoost-HT model achieves an RMSE of 69 cycles and a MAPE of 6.5%. The model also performs well in handling the MFMT feature mapping, demonstrating its effectiveness in predicting battery capacity fade and RUL across different operating conditions. The study highlights the importance of using appropriate ML models and feature mappings to accurately predict the RUL of Li-ion batteries. The XGBoost-HT model is identified as a promising tool for battery management systems due to its high accuracy and robustness. The findings contribute to the advancement of battery prognostics and support the development of more efficient and reliable battery management systems.This paper presents a comparative study of machine learning (ML) models for predicting the remaining useful life (RUL) and capacity fade of lithium-ion (Li-ion) batteries. The study evaluates the performance of multiple ML models, including Random Forest (RF), XGBoost, LightGBM, MLP, LSTM, and Attention-LSTM, using two distinct datasets. The first two cases use a synthetic dataset with linear capacity fade, while the third case uses a real dataset with nonlinear battery behavior. The study introduces a multi-feature multi-target (MFMT) feature mapping to enhance the accuracy of RUL prediction. The results show that the XGBoost-HT model, which incorporates hyperparameter tuning, outperforms other models in terms of root-mean-squared error (RMSE) and mean absolute percentage error (MAPE). For the third case, the XGBoost-HT model achieves an RMSE of 69 cycles and a MAPE of 6.5%. The model also performs well in handling the MFMT feature mapping, demonstrating its effectiveness in predicting battery capacity fade and RUL across different operating conditions. The study highlights the importance of using appropriate ML models and feature mappings to accurately predict the RUL of Li-ion batteries. The XGBoost-HT model is identified as a promising tool for battery management systems due to its high accuracy and robustness. The findings contribute to the advancement of battery prognostics and support the development of more efficient and reliable battery management systems.