Ensemble Learning for Mobility-Aware of Edge User

The mobility of edge users in cellular networks critically affects throughput and signal reception performance. The mobile manner of users can deteriorate network connection, causing a significant delay in data delivery. In addition, a limited number of cellular base stations also increases the signal blocking for far-sight users, making it difficult to achieve optimum signal reception for ubiquitous locations. To maintain throughput performance, the network coordinator needs to identify the mobility type; thus, it can provide a priority resource for the high-mobility user. Accordingly, this paper proposes an ensemble learning to classify the mobility type of edge users by utilizing throughput, distance, and delay as features. The proposed ensemble learning combines multiple weak models, i.e., Random Forest and XGBoost, to construct more robust and established models. The verification results show that ensemble learning predicts a model with more than 97% accuracy. Moreover, the evaluation results confirm that the proposed algorithm outperforms conventional algorithms in precision, recall, and F1-score for every feature.