Machine Learning Based Access Point Selection

The continuously growing number of users and increasing application demands, such as high-definition video streaming, augmented reality, and IoT connectivity, are placing strain on cellular networks. The primary cause of this strain is spectrum shortage, as the limited availability of licensed radio frequencies struggles to accommodate the growing data demand, with an estimated 68% of the world’s population using the internet [1] . This highlights the need for efficient network management to ease pressure on cellular infrastructure, leading to a significant rise in public WiFi access points (APs) and small cell deployments in recent years. As of 2025, 87% of service providers, equipment manufacturers, and enterprises worldwide have deployed or are in the process of deploying OpenRoaming technology for WiFi to enable seamless, secure, and automatic connectivity across networks [2] . Similarly, by 2027, cumulative small cell deployments are expected to reach nearly 36 million radio units [3] , driven by increasing demand for enhanced network capacity and 5G expansion. The widespread adoption of WiFi and small cells positions them as key solutions for offloading traffic and reducing congestion on cellular networks.