Website performance has become a critical determinant of user experience, service reliability, and digital competitiveness in contemporary online systems. Empirical evidence indicates that even marginal delays in page response or loading time can lead to substantial declines in user engagement, conversion rates, and overall trust in web-based platforms. Conventional performance monitoring solutions predominantly rely on static thresholds and predefined alerting rules, which are increasingly inadequate for capturing complex, non-linear, and evolving degradation patterns in modern, high-load web environments. This study proposes an intelligent web performance monitoring framework based on machine learning techniques that enables continuous performance assessment, predictive analysis, and real-time anomaly detection. The proposed system integrates multi-source performance data, advanced feature engineering, and ensemble learning models to identify subtle deviations from normal operational behavior before critical failures occur. Experimental evaluation demonstrates that the machine learning-driven approach significantly outperforms traditional threshold-based monitoring methods in terms of prediction accuracy, anomaly detection timeliness, and system reliability. The results confirm that intelligent monitoring systems provide a viable foundation for proactive performance optimization and resilient web service management in dynamic digital environments.