| 引用本文: | 史殿习,李寒,杨若松,等.用户日常频繁行为模式挖掘.[J].国防科技大学学报,2017,39(1):74-80.[点击复制] |
| SHI Dianxi,LI Han,YANG Ruosong,et al.Mining user frequent behavior patterns in daily life[J].Journal of National University of Defense Technology,2017,39(1):74-80[点击复制] |
|
| |
|
|
| 本文已被：浏览 9918次 下载 9071次 |
| 用户日常频繁行为模式挖掘 |
| 史殿习, 李寒, 杨若松, 莫晓赟, 魏菁 |
|
(国防科技大学 计算机学院， 湖南 长沙 410073)
|
| 摘要: |
| 针对如何在智能手机上高效准确地进行用户日常频繁行为模式挖掘问题展开研究。提出一个基于智能手机的用户日常频繁行为模式挖掘框架；为了减少用来挖掘的上下文篮子的数量、提高挖掘效率，提出一个动态的滑动窗口算法，进而提出一个将上下文出现的频率和持续时间有机地结合起来的加权模式挖掘算法；在此基础上，基于为期6周21个用户的上下文数据，对所提出的挖掘框架和算法进行实验验证，结果表明，所提出的框架和频繁模式挖掘算法可以高效地在资源有限的智能手机上运行，而且能够挖掘出反映用户生活方式的日常频繁行为模式；从两个纬度对用户日常频繁行为模式进行可视化，以可视化方式对用户在不同地方和不同时段的行为模式进行展现，从而方便用户随时了解其日常行为模式。 |
| 关键词: 移动数据挖掘 移动感知 行为模式 |
| DOI：10.11887/j.cn.201701012 |
| 投稿日期：2015-09-01 |
| 基金项目:国家自然科学基金资助项目（61202117，91118008） |
|
| Mining user frequent behavior patterns in daily life |
| SHI Dianxi, LI Han, YANG Ruosong, MO Xiaoyun, WEI Jing |
|
(College of Computer, National University of Defense Technology, Changsha 410073, China)
|
| Abstract: |
| Research focused on how to mine frequent daily behavior patterns of users on smartphones feasibly and efficiently was started. Firstly, a frequent behavior patterns mining framework based on smartphones was proposed. Secondly, a dynamic sliding window algorithm DSW(dynamic slide window) to decrease the context baskets in quantity and improve mining efficiency was proposed. Furthermore, a frequent patterns mining algorithm WePM(weighted pattern mining) which takes both frequency and duration of context occurrence into consideration was developed. On the basis of the above preparation, the mining framework and algorithm were verified experimentally with the context data from 21 users over 6 weeks. Results indicate that the proposed framework and frequent patterns mining algorithm can feasibly and efficiently run on resources limited smartphones to mine daily behavior patterns, and then to reflect users’ lifestyles. Finally, the patterns from two perspectives, namely behavior patterns in different locations and time periods are visualized, which benefits the users to realize their daily behavior patterns at any time. |
| Keywords: mobile data mining mobile sensing behavior patterns |
|
|
|
|
|
