Home // eKNOW 2019, The Eleventh International Conference on Information, Process, and Knowledge Management // View article
Authors:
Weizhu Qian
Franck Gechter
Fabrice Lauri
Keywords: smartphone data; user behavior; energy modeling; regression model.
Abstract:
Nowadays, billions of smartphones are used worldwide. The energy consumption is a critical issue when using such devices. In this context, smartphone power modeling is a mandatory step to better understand energy drain. On that way, the most widespread methods are based on specific hardware/software level analysis. As opposed to these classical approaches, we propose, in this paper, an alternative method aimed at constructing smartphones power models based on user context data provided by Device Analyzer, an Android application developed by the University of Cambridge. From a very large-scale smartphone usage data, we extract the energy-related events. Then, the energy-related context is formulated as the input features of the energy models. So as to predict the energy consumption of a smartphone, we compare four different machine learning models: a Linear Regression model, an AdaBoosted Decision Trees model, a Gradient Boosted Regression Tree model and a Random Forest model. The proposed energy models are then validated on a real user dataset in different usage scenarios.
Pages: 7 to 12
Copyright: Copyright (c) IARIA, 2019
Publication date: February 24, 2019
Published in: conference
ISSN: 2308-4375
ISBN: 978-1-61208-690-3
Location: Athens, Greece
Dates: from February 24, 2019 to February 28, 2019