Pattern Detection with StreamInsight

Pattern Detection with StreamInsight

  • Comments 0

We have added a new sample solution to our samples package on CodePlex. This sample uses the StreamInsight extensibility framework to implement a pattern detection application. The solution can be found in Applications\PatternDetector contains - apart from the used adapters - the following projects:

  • AugmentedFiniteAutomaton (refers to a project in UserExtensions), which implements a pattern matching user-defined operator.
  • PatternDetector, which uses the user-defined operator to look for a V-pattern in a stock ticker stream

Our pattern matching user-defined operator is based on an augmented finite automaton (AFA) - which is basically a non-deterministic finite automaton (NFA) with additional information, called a register, that can be created and maintained as part of the automaton during runtime.

The PatternDetector example uses a sample input stream of stock prices as point events (read from a file with the SimpleTextFileReader input adapter), and defines an AFA that detects a sequence of downticks followed by an equal number of upticks. We detect the pattern over a tumbling window of one hour. The corresponding AFA is shown in the file AFAexample.pptx (included in the solution), and is implemented in AFA_EqualDownticksUpticks.cs. The output is simply dumped on the console.

For more information on pattern matching using AFA, see:

Badrish Chandramouli, Jonathan Goldstein, and David Maier. High-Performance Dynamic Pattern Matching over Disordered Streams. In VLDB 2010.

Regards,
Badrish

Leave a Comment
  • Please add 2 and 7 and type the answer here:
  • Post