Accurate prediction of lithium-ion batteries’ state of health (SOH) is crucial for preventing catastrophic system failures. This study investigates the application of ensemble modeling to characterize capacity degradation and forecast remaining charge–discharge cycles. Leveraging NASA's battery charge/discharge dataset, we developed and compared feed-forward neural network (FNN) and random forest (RF) regression models. To enhance predictive accuracy, we constructed an ensemble model that combines the strengths of both individual models. A key aspect of our methodology was the accurate evaluation of model performance across different battery datasets. Rather than using a single dataset for training and testing, we adopted a cross-validation approach to assess model generalization capabilities. This strategy allowed us to identify the robustness of the models for predicting SOH and estimating remaining battery life. Our findings indicate comparable performance among the FNN, RF, and ensemble models. While all models demonstrated effective capacity degradation prediction, the ensemble model exhibited slightly superior performance in a few scenarios. These findings emphasize the advantages of ensemble modeling in enhancing the accuracy and reliability of lithium-ion battery prognostics.