mining.extraction.KmeansSymbolDistribution Class Reference

Create clusters over the distribution of symbols in a set of input streams. More...

Static Public Member Functions
static void	main (String[] args) throws FunctionException

Detailed Description

Create clusters over the distribution of symbols in a set of input streams.

In this example, input traces are made of symbols a and b. A set of seven such traces is located in the file strings-1.csv. For each of these traces, the pattern processor β computes a feature vector made of two numbers, corresponding to the fraction of a's and b's in the trace. This is done by

• slicing the input trace according to the current symbol
• counting the events in each slice
• taking the values of the resulting map into a list, and normalizing them (so that their sum is equal to 1)

For example, on the input sequence

a, b, a, a, b, b

the resulting feature vector would be (0.4, 0.6).

We then use the K-means clustering algorithm to find the centroids of two clusters based on those feature vectors.

The processor mining function is therefore parameterized as follows:

Parameter	Value
	(K-means with K=2)

The traces in the input CSV file either have an approximate 30%-70% distribution of a's and b's, or the reverse. The feature vectors can be plotted as follows, with each dot representing the a-b distribution of a single trace.

Applying the k-means algorithm, with k=2, will compute two cluster centers, represented by crosses in the above plot.

Author

Sylvain Hallé

Definition at line 96 of file KmeansSymbolDistribution.java.

---

The documentation for this class was generated from the following file:

• Source/src/mining/extraction/KmeansSymbolDistribution.java