The network infrastructures have to keep growing exponentially with the ever-increasing data traffic; hence, accurate and efficient network traffic prediction models are imperative. Evaluations of GNN and KNN-GNN hybrids for network traffic prediction, as well as KNN model for protocol and service prediction with anomaly detection across various batches of parameters are presented in this paper. While Classic models struggle to capture such complex patterns in network data, the proposed framework combines the strengths of the spatial capabilities of GNNs with the feature of finding local patterns by KNNs to provide superior performance in the network traffic prediction. Concurrently, we extend this work in predicting the network protocols along with their associated services and anomaly detection using KNN classifier. The hybrid KNN-GNN model proves more effective for traffic prediction, enabling optimized network management, while KNN-based protocol and anomaly detection enhances the security by accurately identifying protocol-service combinations and anomalies. Furthermore, the predicted network traffic volume can be used to infer the corresponding protocol based on its expected characteristics. Experimental results on a real-time dataset demonstrated that KNN-GNN based model is more effective than traditional methods for traffic prediction with an MAE of 17.1664, in addition KNN performed more effectively in the prediction of protocol and anomaly with an accuracy of 90% and 99%, respectively.