Accurate State of Charge (SOC) estimation in lithium-ion batteries is critical for ensuring the safety, reliability, and efficiency of modern energy storage systems, particularly under dynamic operating conditions and temperature variations. Traditional SOC estimation methods often fall short due to their inability to fully account for non-linear thermal effects, leading to reduced accuracy and potential safety risks. This study introduces a data-driven approach leveraging advanced machine learning (ML) models, including Random Forest and Long Short-Term Memory (LSTM) networks, to enhance SOC prediction accuracy by incorporating temperature as a core variable. High-fidelity datasets simulating real-world conditions were generated using MATLAB to train and evaluate these models. The Random Forest algorithm exhibited superior accuracy and robustness, while LSTM excelled in capturing temporal dependencies under dynamic loading conditions. The findings underscore the potential of integrating ML-driven SOC estimation into battery management systems to mitigate temperature-induced errors, paving the way for smarter and more reliable energy storage solutions in electric vehicles and renewable energy applications.