org.processmining.nlplogutils.algorithms.analysis.feature

Class EntropyContextProvider

java.lang.Object

org.processmining.nlplogutils.algorithms.analysis.feature.EntropyContextProvider

All Implemented Interfaces:

ContextFeatureProvider

public class EntropyContextProvider
extends java.lang.Object
implements ContextFeatureProvider

Looks at the context in terms of entropy. Each sequence is a unique item.

Constructor Summary

Constructors

Constructor and Description
EntropyContextProvider()

Method Summary

Modifier and Type	Method and Description
double[]	computeFeature(org.apache.commons.collections4.MultiSet<Context> equalContexts) Extracts a feature from the context.
int	getNumberOfResultFeatures() If a feature provider computes 1 feature, it return 1, if 2, then 2.

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

EntropyContextProvider

public EntropyContextProvider()

Method Detail

computeFeature

public double[] computeFeature(org.apache.commons.collections4.MultiSet<Context> equalContexts)

Description copied from interface: ContextFeatureProvider

Extracts a feature from the context. A context is an int[] where individual event classes are encoded as integers. A typical context of an event would be a symmetric window of size 4. Let us consider the trace (with event ids in brackets): a(1), b(2), c(3), b(2), d(4), e(5) A symmetric window context of event c in this trace would be: {1,2,2,4} b has two occurrences. Therefore, it has two contexts: {-1,1,3,2}, and {2,3,4,5}. Contexts can be defined left, right, or symmetric. See ContextParameters.WindowType

Specified by:

computeFeature in interface ContextFeatureProvider

Parameters:

equalContexts - a preprocessed bag/multiset of contexts

Returns:

some feature of this context depending on the subclass.

getNumberOfResultFeatures

public int getNumberOfResultFeatures()

Description copied from interface: ContextFeatureProvider

If a feature provider computes 1 feature, it return 1, if 2, then 2. :) This is useful, as some feature providers need to to complex computation and can provide a number of results based on that See for example the HighDimClusteringProvider that returns the number of clusters and a cluster variance metric.

Specified by:

getNumberOfResultFeatures in interface ContextFeatureProvider

Returns: