This paper introduces a GraphSAGE-based framework for movie genre classification. Our method reformulates the problem as a node classification task, representing each movie as a node in a fully connected weighted graph. The edge weights are derived from the Network Laplacian Descriptor to quantify structural similarity between movie character networks, ensuring meaningful connections. Each node is characterized by a feature vector that combines statistical, embedding, and spectral information from its associated character network. Our models aggregate features and node relationships to improve classification accuracy. Evaluation on the Movie Galaxy dataset demonstrates that the proposed approach substantially outperforms traditional classifiers, with the GraphSAGE achieving 96.17%. Ablation analysis further indicates that combining statistical and spectral features yields the highest classification accuracy, underscoring the importance of feature selection.