Abstract
In the context of HWN (Heterogeneous Wireless Networks), the claim for the best connectivity is a mandatory challenge for the users. As all of us know, there is a panoply of RATs (Radio Access Technologies) that users can connect to due to the multi-interface property of the actual user-devices. Indeed, these multi-interfaces users are always seeking for reaching the ABC concept (Always Best Connected). This challenge is also in the heart of the NGN Next-Generation Networks, where scholars are proposing models and algorithms to make in practical the ABC paradigm. So, the idea is to make user-devices capable to select the best RAT instantaneously each time the users need to use the network. This process is known as the network selection problem. Simply, the network selection consists on the automatic and transparent selection of the best RAT at a given time; many papers exist in this field using mainly MADM (Multi-Attribute Decision-Making) methods, different implementation of the MADM methods are proposed with various objectives, increasing the QoS, reducing the energy consumption, reducing the cost. The objective of this paper is to address this challenge by proposing a modified machine learning algorithm (an adapted K-means) to make the network selection, the results show that the proposal is more performer compared to the legacy MADM methods.