Graph streams represent data interactions in real applications. The mining of graph streams plays an important role in network security, social network analysis, and traffic control, among others. However, the sheer volume and high dynamics cause great challenges for efficient storage and subsequent query analysis on them. Current studies apply sketches to summarize graph streams. We propose LSketch that works for heterogeneous graph streams, which effectively preserves the label information carried by the streams in real scenes, thereby enriching the expressive ability of sketches. In addition, as graph streams continue to evolve over time, edges too old may lose their practical significance. Therefore, we introduce the sliding window model into LSketch to eliminate the expired edges automatically. LSketch uses sub-linear storage space and can support structure based queries and time-sensitive queries with high accuracy. We perform extensive experiments over four real datasets, demonstrating the superiority of the proposed method over state-of-the-art methods, in aspects of query accuracy and time efficiency.

翻译：图流表示现实应用中的数据交互。对图流的挖掘在网络安全性分析、社交网络分析以及交通控制等领域发挥着重要作用。然而，其海量数据与高度动态性为高效存储及后续查询分析带来了巨大挑战。现有研究采用Sketch（草图）对图流进行摘要。我们提出LSketch，该方案适用于异构图流，能够有效保留真实场景中图流携带的标签信息，从而增强Sketch的表达能力。此外，随着图流随时间持续演化，过旧的边可能失去实际意义。因此，我们将滑动窗口模型引入LSketch，以自动淘汰过期边。LSketch使用亚线性存储空间，能够高精度支持基于结构的查询与时间敏感查询。我们在四个真实数据集上进行了大量实验，证明了所提方法在查询精度与时间效率方面均优于现有最优方法。