Web traffic forecasting plays a crucial role in optimizing network resources, enhancing user experience, and ensuring efficient server load management. Traditional approaches, such as ARIMA, SARIMA, and machine learning methods like support vector machines and decision trees, focus primarily on temporal data. These models, however, often fail to capture the intricate relationships within web traffic data, such as user interactions or page-to-page connections, resulting in sub-optimal predictive performance. Graph Convolutional Networks (GCNs) address these limitations by modeling web traffic as a graph, where web pages or users are represented as nodes, and their interactions form edges. GCNs aggregate node information through graph structures, enabling the model to learn both spatial and temporal dependencies inherent in web traffic. This ability to exploit complex data relationships makes GCNs well-suited for more accurate and dynamic web traffic predictions. In this work, we propose a GCN-based framework for web traffic forecasting, incorporating multiple optimizers like Adam, RMSProp, and SGD to identify the model’s fair performance. By optimizing training through these methods, the GCN model efficiently captures both short-term fluctuations and long-term trends in web traffic patterns. Our study highlights the potential of GCNs in elevating the accuracy and reliability of web traffic forecasting. The integration of advanced optimizers further enhances convergence and prediction efficiency, offering a more scalable solution to meet the demands of rapidly growing and complex web systems.