.png){kind=icon}
Return
ADOMC-NPR Automatic Decision-Making Offloading Framework for Mobile Computation Using Nonlinear Polynomial Regression Model
.png){kind=icon}
PRE
AI
DOI：10.4018/IJWSR.2019100104.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Nowadays, mobile computation applications attract major interest of researchers. Limited processing power and short battery lifetime is an obstacle in executing computationally-intensive applications. This article presents a mobile computation automatic decision-making offloading framework. The proposed framework consists of two phases: adaptive learning, and modeling and runtime computation offloading. In the adaptive phase, curve-fitting (CF) technique based on non-linear polynomial regression (NPR) methodology is used to build an approximate time-predicting model that can estimate the execution time for spending the processing of the detected-intensive applications. The runtime computation phase uses the time predicting model for computing the predicted execution time to decide whether to run the application remotely and perform the offloading process or to run the application locally. Eventually, the RESTful web service is applied to carry out the offloading task in the case of a positive offloading decision. The proposed framework experimentally outperforms a competitive state-of-the-art technique by 73% concerning the time factor. The proposed time-predicting model records minimal deviation of the originally obtained values as it is applied 0.4997, 8.9636, 0.0020, and 0.6797 on the mean squared error metric for matrix-determinant, image-sharpening, matrix-multiplication, and n-queens problems, respectively.
Keywords:
Computation Offloading
Mobile Cloud Computing
Polynomial Regression
RESTful
Journal
I
IF:
0.7
Papers:
6
Citations:
93
Organization
Cited Papers
ShareSave
ShareSaveShareSaveShareSaveFluctuation-Aware and Predictive Workflow Scheduling in Cost-Effective Infrastructure-as-a-Service Clouds.png)
IEEE ACCESS
OAAI
IEEE ACCESSIF3.6
OAAIShareSave
View more