Mobile applications usage has considerably increased since the last decade. Successful apps need to make the users feel comfortable while using them, thus demanding high-quality design and implementation. One of the most influencing factors for user experience is battery consumption, which should have the minimum possible impact on the battery. The current body of knowledge on energy consumption measurement only reports approaches relying on complex instrumentation or stressing the application with many test scenarios, thus making it hard to measure energy consumption in practice. In this work, we explore the performance of machine learning to predict the energy consumption level of JAVA classes in Android apps, leveraging only a set of structural properties extracted via source code analysis, without requiring any hardware measurements tools or executing the app at all. The preliminary results show the poor performance of learning-based estimation models, likely caused by (1) an insufficient amount of training data, (2) a limited feature set, and (3) an inappropriate way to label the dependent variable. The paper concludes by presenting the limitations of the experimented models and the possible strategies to address them.

Predicting The Energy Consumption Level of Java Classes in Android Apps: An Exploratory Analysis

Iannone E.;De Stefano M.;Pecorelli F.;De Lucia A.
2022-01-01

Abstract

Mobile applications usage has considerably increased since the last decade. Successful apps need to make the users feel comfortable while using them, thus demanding high-quality design and implementation. One of the most influencing factors for user experience is battery consumption, which should have the minimum possible impact on the battery. The current body of knowledge on energy consumption measurement only reports approaches relying on complex instrumentation or stressing the application with many test scenarios, thus making it hard to measure energy consumption in practice. In this work, we explore the performance of machine learning to predict the energy consumption level of JAVA classes in Android apps, leveraging only a set of structural properties extracted via source code analysis, without requiring any hardware measurements tools or executing the app at all. The preliminary results show the poor performance of learning-based estimation models, likely caused by (1) an insufficient amount of training data, (2) a limited feature set, and (3) an inappropriate way to label the dependent variable. The paper concludes by presenting the limitations of the experimented models and the possible strategies to address them.
2022
9781450393010
File in questo prodotto:
Non ci sono file associati a questo prodotto.

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11386/4808052
 Attenzione

Attenzione! I dati visualizzati non sono stati sottoposti a validazione da parte dell'ateneo

Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 0
  • ???jsp.display-item.citation.isi??? 1
social impact