Accurate premium estimation is fundamental in insurance pricing, ensuring both fairness for policyholders and financial stability for insurers. Generalized Linear Models (GLMs) have been widely adopted due to their interpretability and compliance with regulatory standards. However, recent advancements in Machine Learning (ML) offer improved predictive accuracy by capturing non-linear relationships and complex risk structures. This study evaluates the performance of GLMs against ML models such as Decision Trees, Random Forest, Gradient Boosting (GBM), and XGBoost in predicting claim frequency and severity. Results demonstrate that GBM and XGBoost outperform GLMs in predictive accuracy, yet challenges remain in ensuring model interpretability and regulatory transparency. While GLMs provide an established actuarial framework, ML techniques offer optimization potential, leading to enhanced risk assessment and fairer pricing. The findings highlight the trade-off between predictive accuracy and interpretability in insurance pricing, emphasizing the need for hybrid approaches that integrate the strengths of both GLMs and ML models. This study contributes to the ongoing evolution of actuarial science by demonstrating how data-driven techniques can complement traditional methodologies while maintaining compliance with industry regulations.