Stream Mining

The course is devoted to processing and mining data streams in which unbounded, high-volume data streams are ingested, analyzed and optionally stored. Typically, data is processed with one pass by algorithms which take into account that data may evolve over time. The course content is mainly focused on algorithms applicable in streaming contexts, ranging from simple counting and frequency analysis, right up to change and anomaly detection approaches.

Apart from the analysis of the most important algorithms, the course also equips students with theoretical knowledge about the key challenges which streaming systems need to solve.

Main topics:

• Sketches and standing queries;
• Clustering data streams;
• Data stream classification;
• Frequent pattern mining in data streams;
• Time-series analysis and anomaly detection in data streams;
• Change detection in stream mining;
• Streaming systems: windows, watermarks, triggers and correctness;