Machine Learning Regression-Based Approach for Dynamic Wireless Network Interface Selection

Battery consumption is a general problem in any portable wireless device and it depends directly on the transmission technology (cellular, Wi-Fi or short-range wireless networks) that is used to send and receive data. When various networks are available, mobile devices should be able to choose which network interface to use based on a variety of factors, such as required bandwidth or energy efficiency. This work proposes a dynamic wireless network interface-selection mechanism focused on minimizing the energy consumption of the mobile device, allowing an increase in battery life. In doing so, Machine Learning (ML) regression-based algorithms are used to predict the energy cost per transferred byte for each type of available network interface using field data. A comparison of the energy consumptions for both the proposed mechanism and the Android native method is performed. Numerical results show that our proposal helps save energy.