Towards optimising Wi-Fi energy consumption in mobile phones: A data driven approach

Abstract

Contemporary mobile devices are equipped with multiple network interfaces with diverse characteristics. Although the Wi-Fi interface bestows commendable throughput and data transfer efficiency, it is least power efficient in the idle state and causes highest energy overhead when scanning for networks. In this paper we present a data driven approach to alleviate this issue, focusing on Wi-Fi usage by the user's perspective. We model the Wi-Fi usage of mobile users based on their past usage to predict usage requirements. This allows intelligently switching on the Wi-Fi interface only if the user context demands. Thus, it reduces long periods of time being in the idle state and significantly lessens the number of futile network scans. Based on the trace data collected from Rice-Livelab study, we extract temporal, application usage, operational state and location context data to build our prediction model. This study includes a systematic feature engineering process followed by the deployment of machine learning algorithms on the target dataset. We used Sampling, Ensemble and Hybrid techniques to mitigate the class imbalance problem of our prediction model. Evaluated metrics indicate that decision tree based classification algorithms perform well with the dataset and suit for working with mobile usage data, which are mostly conflated with noise, data imbalance.