|M.Sc Student||Busany Nimrod|
|Subject||Incorporating Uncertainy into Event-Based Monitoring Systems|
|Department||Department of Industrial Engineering and Management||Supervisor||Professor Avigdor Gal|
|Full Thesis text|
Answering queries that correlate multiple events over time becomes increasingly important in applications from
urban transportation to security to office automation. This work proposes methods for answering such queries over data streams with timestamps that may be erroneously recorded or corrupted at the source and data items that may be dropped from a stream due to network failures or load shedding policies. We focus on a novel approach for probabilistically answering queries using constraints over data orderings in a stream. To that end, we introduce an abstraction of fragmented intervals, extend Allen's algebra to reason with such intervals, and show three methods, each using a different assumption regarding the underlying data, for probabilistically answering queries over fragmented intervals. Our empirical analysis, using both real-world and synthetic data, shows that the proposed methods allow for efficient and effective reasoning over data streams, with a graceful reduction of accuracy as uncertainty increases.