A Model for Hour-Wise Prediction of Mobile Device Energy Availability

LONGO, M.; MATEOS, C.; ZUNINO, A.

Conferencia; 15th International Conference on Information Technology: New Generations, ITNG; 2018

Mobile devices have become so ubiquitous and their computational capabilities have increased so much that they have been deemed as first-class resource providers in modern computational paradigms. Particularly, novel Mobile Cloud Computing paradigms such as Dew Computing promote offloading heavy computations to nearby mobile devices. Not only this requires to produce resource allocators to take advantage of device resources, but also mechanisms to quantify current and future energy availability in target devices. We propose a model to produce hour-wise estimations of battery availability by inspecting past device owner?s activity and relevant device state variables. The model includes a feature extraction approach to obtain representative features/variables, and a prediction approach, based on regression models and machine learning classifiers. Comparisons against a relevant related work in terms of the Mean Squared Error metric shows that our method provides more accurate battery availability predictions in the order of several hours ahead.