一种基于用户偏好的移动计算卸载决策算法  

A Decision Making Algorithm of Mobile Computational Offloading Based on User Personalized Requirements

作　　者：蒋青苗[1] IANG Qing-miao(Information Engineering school,Communication University of China,Beijing 100024)

机构地区：[1]中国传媒大学信息与通信工程学院

出　　处：《中国传媒大学学报（自然科学版）》2019年第5期70-77,共8页Journal of Communication University of China：Science and Technology

摘　　要：移动计算卸载可以通过互联网将智能手机端中的计算密集型应用程序传输到服务器端运行并返回结果,有助于提升智能手机的性能。移动计算卸载决策算法往往只重视客观指标,而不考虑用户的个性化需求。本文提出了一种基于用户偏好的计算卸载算法。首先,结合机器学习算法设计和训练了一个用户模型对用户的个性化卸载需求进行预测。然后,通过系数调整,将影响移动计算卸载的用户主观因素与客观指标相结合,构建了系统运行时的动态网络流图。最后,结合最小割算法对移动端应用程序进行划分。实验结果表明,本文提出的卸载决策算法不仅比贪婪卸载算法更能满足用户个性化需求,而且在大数据量的情况下,算法的执行时间甚至优于直接在云服务器端运行。Mobile computational offloading can improve the performance of smart phone by transferring the compute-intensive applications on the smart phone to the server side and returning results via the network.Mobile computional offloading decision making algorithms mostly focus on objective indicators,regardless of the user's individual needs.This paper proposes a computational offloading algorithm based on user preferences.First,a user model is designed and trained in conjunction with machine learning algorithms to predict the user's personalized offloading needs.Then,through the coefficient adjustment,the subjective factors affecting the mobile computational offloading are combined with the objective indicators to construct a dynamic run-time network flow graph.Finally,the mobile application is partitioned by the Min-cut algorithm.The experimental results show that the proposed offloading decision making algorithm not only satisfies the user's personalized requirements than the greedy offloading algorithm but also the execution time is even better than that of the cloud server in the case of large data amount.

关 键 词：移动计算卸载 用户个性化需求 机器学习 最大流最小割 

分 类 号：TP3393.0[自动化与计算机技术—计算机系统结构]