A learning-based approach for network selection in WLAN/3G heterogeneous network

To meet the huge traffic growth, heterogeneous networks composed of wireless local area networks (WLAN) and 3G cellular networks are used to provide higher capacity and coverage. When the two networks are available, selecting the best network for downloading data with minimum device energy consumption and high quality of service (QoS) becomes a challenging issue especially that mobile devices have limited energy capacity. This paper proposes a learning based approach for performing network selection based on real-network implementations. The main contributions are first, presenting an approach for building training data as a basis for machine learning of network selection and then developing the classification model for network selection. The model considers the features that affect the selection decision known by the user: availability of the networks, signal strength reflecting the channel quality, data size, battery life, speed of the user, location, and type of application. The training data set is based on experimental measurements of WiFi and 3G links using a Samsung Galaxy SII device. The network class annotation chooses the network that provides the user either highest QoS, lowest energy consumption or highest energy efficiency based on its current features status and service requirements. For real-time network selection, the developed model uses decision tree classification. Testing the performance of the classifier using cross validation demonstrated high accuracy for selecting betweenWiFi and 3G networks.