透過隱馬可夫模型預測使用者行為進行在智慧型手機上的耗電量優化

Energy Optimization on Smartphone by Predicting User Behavior Using Hidden Markov Model

羅尹聰

指導教授 ： 金仲達

國立清華大學/電機資訊學院/資訊工程學系/碩士(2012年)

摘要

行動裝置上的耗電量優化一直是熱門的研究領域，然而這部份的相關研究大多是專注於系統層面，且幾乎沒有將使用者與裝置間互動的行為，習慣與使用經驗上納入考量。背景程式的管理是一個值得探討的方向，因為持續運行於背景的應用程式將會造成電量的消耗，如此將會縮短裝置的使用時間並且導致不良的使用經驗。然而若僅是簡單的將所有使用者跳出的應用程式關閉，這樣造成的結果將與智慧型行動裝置的高互動體驗的特色相互違背。因此，我們必須透過了解使用者的行為才能適當的解決此問題。

我們需要一套良好的機制來確保被保留的程式都是使用者很可能會使用的應用程式。這需要讓機器能夠學習並準確的預測裝置使用者接下來的使用行為，我們相信透過使用者的使用行為樣式以及其他相關的使用者資訊能夠達成此目標。

在這篇論文中，我們提出一套透過使用者行為建模方式的機制，並且透過此機制來做背景程式的管理並達到智慧型裝置耗電量上的優化。

關鍵字

隱藏馬可夫模型 ； 耗電優化 ； 智慧型裝置 ； 使用者經驗 ； 使用者行為 ； 機器學習 ； Google Android ； 行動應用程式

並列摘要

Energy optimization has been a popular area of research in mobile computing device. However, most of the previous researches on energy optimization focus on the device itself and seldom consider the interaction of the user and the device, or take the user behavior pattern as well as user experience into account. One area where energy optimization can be exercised on mobile device is background applications. Applications running in background will consume energy, leading to shorter usage time and worse user experience. However, if we kill every application whenever it is put into background, the relaunch time will be very long once the user wants to switch back to that application. This also leads to bad user experience.

Apparently, we need a good mechanism that keeps only those background applications that will be needed by user. This requires that user behavior in the next period time be modeled and predicted accurately. We believe accurate prediction can be made depends on the behavior patterns of the usage of the apps and other observable context variables.

In this thesis, we introduce such a mechanism that manages background application for energy optimization by predicting user behavior through usage modeling.

並列關鍵字

Hidden Markov Model ； Energy Optimization ； Smartphone ； Machine Learning ； User Context ； User Behavior ； User Experience ； Google Android ； Mobile Application

參考文獻

[2] L.E. Baum, T. Petrie, G. Soules, and N. Weiss. A maximization technique oc- curring in the statistical analysis of probabilistic functions of markov chains. The annals of mathematical statistics, 41(1):164–171, 1970.

[4] Huanhuan Cao, Daxin Jiang, Jian Pei, Enhong Chen, and Hang Li. Towards context-aware search by learning a very large variable length hidden markov model from search logs. In Proceedings of the 18th international conference on World wide web, WWW ’09, pages 191–200, New York, NY, USA, 2009. ACM.

[5] Chen-Ling Chou and Radu Marculescu. Designing heterogeneous embedded network-on-chip platforms with users in mind. Trans. Comp.-Aided Des. Integ. Cir. Sys., 29(9):1301–1314, September 2010.

[6] G. Contreras and M. Martonosi. Power prediction for intel xscaleR processors using performance monitoring unit events. In Low Power Electronics and Design, 2005. ISLPED’05. Proceedings of the 2005 International Symposium on, pages 221–226. IEEE, 2005.

[11] Jason Flinn and M. Satyanarayanan. Powerscope: A tool for profiling the energy usage of mobile applications. In Proceedings of the Second IEEE Workshop on Mobile Computer Systems and Applications, WMCSA ’99, pages 2–, Washington, DC, USA, 1999. IEEE Computer Society.