PlanetData Network of Excellence FP 7 – 257641 D 1 . 5 Trend and anomaly detection in non - structured data Coordinator : Alexandra Moraru

for dissemination) Non-structured or unstructured data is data that doesn’t conform to an explicit and well-defined formal data model. This deliverable focuses on textual and network data. We discuss several statistical properties by which these types of data differ from more structured data. Trend and anomaly detection is the process of discovering patterns in the data that do not conform to normal or expected behaviour; it has many applications and draws upon techniques from several related disciplines. We present the state of the art and directions for future work in several areas relevant to trend and anomaly detection in textual and network data: text processing of informal documents, online learning, adaptive data summarization, event processing, social media management, network sampling, and network evolution.