Monitoring & Analysing Big Data Streams

Traditionally, the aim of data mining has been to discover, monitor and improve processes by extracting knowledge from event logs that are readily available in today's information systems. In the last few years, process monitoring and analysis has enjoyed a tremendous growth and a rapid development at both conceptual and algorithmic levels. In particular, there have been successful implementation of process monitoring systems in many application areas, including manufacturing, e-health and e-government.

Today, the current trend toward large-scale collaborative processes featuring thousands of elementary activities per minute is generating a number of new research issues. When large-scale processes are executed on cloud-based service-oriented environments or even on the global Net, elementary activities can be mapped to fine or coarse-grained protocol events and process logs increasingly come to show all typical properties of "big data": wide physical distribution, diversity of formats, non-standard data models, heterogeneous semantics. Computing metrics over such "big logs" also requires to handle security and privacy concerns of many participants, and even to deal with non-uniform trustworthiness of log entries. New semantics-aware techniques are therefore required for designing, validating and deploying process metrics on big data streams, achieving (i) rich dash-boarding of
processes’ non-functional indicators (ii) online insight and improvement capabilities.

The seminar will discuss some emerging semantics-aware analysis techniques in the framework of large-scale processes.