Hybrid Machine Learning and Reinforcement Learning for IoT Device Classification and Bandwidth Optimization

Increased traffic in the Internet of Things (IoT) network results in a large amount of network data that should be processed efficiently in order to manage IoT network resources properly. Hence, bandwidth management and resource allocation are key challenges in managing large-scale IoT networks. In this study, a novel hybrid framework that is composed of machine learning (ML) and reinforcement learning (RL) methods has been proposed for IoT device classification as well as for dynamic bandwidth allocation in large-scale IoT networks. The proposed framework includes four ML classification methods, including Support Vector Machine (SVM), Random Forest (RF), K-Nearest Neighbor (KNN), and Neural Network (NN), for classifying various types of IoT devices by using their traffic and communication behaviors. After providing classification results to the bandwidth management framework, improved bandwidth allocation is achieved by using the proposed framework. In addition, the RL method is also incorporated into the framework for bandwidth allocation, in which the learning agent continuously improves allocating bandwidth by increasing the cumulative reward during the learning process. Experiment results revealed that the proposed framework significantly improved performance metrics of IoT networks, such as increased network throughput by approximately 50% and decreased average delay by approximately 40%, as well as reduced packet loss by approximately 30% compared with traditional bandwidth allocation methods. In terms of classification accuracy, the highest accuracy was achieved by the Neural Network model. The proposed framework enables to efficiently and adaptively manage large-scale IoT networks.

@article{N2026,
  title     = {Hybrid Machine Learning and Reinforcement Learning for IoT Device Classification and Bandwidth Optimization},
  journal   = {International Journal of Advanced Computer Science and Applications},
  volume    = {17},
  number    = {7},
  year      = {2026},
  publisher = {The Science and Information Organization},
  author    = {Manjunatha T N and Vidyalakshmi K and Dankan Gowda V},
  doi       = {10.14569/IJACSA.2026.0170712},
  url       = {https://doi.org/10.14569/IJACSA.2026.0170712}
}