This paper proposes a Graph Neural Network (GNN)-based approach for movie genre classification. The method models movies as nodes in a fully connected weighted movie networks (FCWMN). Each node represents a movie and includes a feature vector combining statistical, embedding, and spectral network properties. The Network Portrait Divergence computes edge weights to quantify movie similarities. The study evaluates the approach on the Movie Galaxy dataset, which includes ten movie genres. The model applies a Graph Convolutional Network (GCN) to learn node representations and classify movies. To enhance learning, the study addresses class imbalance, applies five-fold cross-validation, and conducts an ablation study to assess the contribution of different feature types. The results show that combining statistical and spectral features achieves the highest accuracy of 87%. A comparison with traditional classifiers (Random Forest, Decision Trees, and SVMs) confirms the superiority of the GCN-based approach. The improvement stems from GCN’s ability to exploit relationships between movies, which traditional classifiers fail to capture.