摘要:签到日志记录了用户对于某类设施的使用情况，从中发现用户行为模式，在很多领域如精确广告投放、犯罪团伙发现等方面都具有非常广泛的应用价值.但是，发现过程却较为困难，主要因为：（1）日志数据体现为长时间序列且含有噪声，导致数据在高维空间分布较为稀疏，影响模式提取的准确性；（2）行为模式往往与不同的时间尺度相关；（3）多样的参数选择空间以及数据处理方式使得传统的机器学习方法很难获得可信且易于理解的行为分析结果.提出一种面向签到日志的用户行为模式交互探索的方法，该过程采用动态子空间策略，动态改变用于分析相似行为模式的时间片，从而减少人为设定参数对于分析结果的影响.方法集成了一个可视分析工具以支持该过程，利用该工具，分析人员可以实时了解方法每一步发现的模式，及时调整分析过程、直观理解和验证分析结论.包含了一个基于真实数据集的案例分析和一个来自不同领域专家的评审，其结果验证了方法的有效性.



Abstract:Check-in logs record how users access certain facilities. Discovering users' behavior patterns via logs has a wide range of applications, such as targeted advertising, criminal activity detection, etc. However, the discovery process is complex and challenging, due to the following reasons. (1) Log data is usually of long-term and contains noise, with sparse distribution of data in high-dimensional space. (2) Behavior patterns always relate to different time scales. (3) The variety of parameter selections and methods of data processing make traditional machine learning approaches difficult to obtain credible and understandable behavior analysis results. This study proposes an interactive approach to exploring behavior patterns from check-in logs. The process uses a dynamic subspace strategy which changes the time slices to analyze similar behavior patterns dynamically. The strategy reduces the effect of setting parameters artificially on the analytical results. The proposed approach integrates a visual analytical tool to support the process. Through visualization, analysts could understand the patterns found in each step-in real time, adjust the analysis process, comprehend and verify the results intuitively. The paper also presents a case study based on a real data set and a review of experts from different fields. The results confirm the effectiveness of the approach.



PDF全文下载地址:

http://jos.org.cn/jos/article/pdf/5824