A machine learning access network selection in a heterogeneous wireless environment

Abstract

In the context of Heterogeneous Wireless Networks (HWN), achieving the best connectivity is crucial for users. With the availability of multiple Radio Access Technologies (RATs), users are constantly looking to be “Always Best Connected” (ABC) by connecting to the best possible RAT. This challenge is also important in Next‐Generation Networks (NGN), where researchers are proposing models and algorithms to make the ABC paradigm a reality. One such challenge is the network selection problem, which involves the automatic and transparent selection of the best RAT at a given time. Many papers in this field have used Multi‐Attribute Decision‐Making (MADM) methods, which have been implemented with various objectives such as increasing quality of service (QoS), reducing energy consumption and costs etc. This paper proposes a modified machine learning algorithm, specifically an adapted K‐means algorithm, to address this challenge. The results show that this proposal outperforms legacy MADM methods in terms of latency, jitter, packet loss, and data rate. So, the paper presents a new approach to network selection in HWN and NGN, and the results suggest that it could be a promising solution to this important challenge.