weka.filters.unsupervised.timeseries

Class FlatToRelational

java.lang.Object

weka.filters.Filter

weka.filters.SimpleFilter

weka.filters.SimpleStreamFilter

weka.filters.unsupervised.timeseries.FlatToRelational

All Implemented Interfaces:

java.io.Serializable, weka.core.CapabilitiesHandler, weka.core.CapabilitiesIgnorer, weka.core.CommandlineRunnable, weka.core.OptionHandler, weka.core.RevisionHandler, weka.filters.StreamableFilter

public class FlatToRelational
extends weka.filters.SimpleStreamFilter

A filter that converts a flat representation of multivariate time series to a relational representation. This relational representation can then be used for time series classification with other filters and classifiers in WEKA. The flat representation is popular in applications in which each time series is of the same length. However, time series filters and classifiers in WEKA use relational attributes to store time series instead because they support time series of varying length. The data in each time series is assumed to be given as the values of the selected attributes (the class, if set, is ignored). By default, each time series is expected to be univariate (the number of variables is set to 1). In general, assuming each time series is multivariate with M attributes, the first multivariate observation in the time series is assumed to be given by the first M selected attributes, the second observation is assumed to be given by attributes M + 1 to 2M, and so on. Hence, the number of selected attributes (minus the class) must be a multiple of M.

Author:

Steven Lang

See Also:

Serialized Form

Constructor Summary

Constructors

Constructor and Description
FlatToRelational()

Method Summary

Modifier and Type	Method and Description
weka.core.Capabilities	getCapabilities() Returns the Capabilities of this filter.
boolean	getDoNotKeepOtherAttributes()
int	getNumVariables()
java.lang.String	getRange()
java.lang.String	globalInfo()
void	setDoNotKeepOtherAttributes(boolean keepOtherAttributes)
boolean	setInputFormat(weka.core.Instances instanceInfo) This method needs to be called before the filter is used to transform instances.
void	setNumVariables(int numVariables)
void	setRange(java.lang.String range)

Methods inherited from class weka.filters.SimpleStreamFilter

batchFinished, input

Methods inherited from class weka.filters.Filter

batchFilterFile, debugTipText, doNotCheckCapabilitiesTipText, filterFile, getCapabilities, getCopyOfInputFormat, getDebug, getDoNotCheckCapabilities, getOptions, getOutputFormat, getRevision, isFirstBatchDone, isNewBatch, isOutputFormatDefined, listOptions, main, makeCopies, makeCopy, mayRemoveInstanceAfterFirstBatchDone, numPendingOutput, output, outputPeek, postExecution, preExecution, run, runFilter, setDebug, setDoNotCheckCapabilities, setOptions, toString, useFilter, wekaStaticWrapper

Methods inherited from class java.lang.Object

equals, getClass, hashCode, notify, notifyAll, wait, wait, wait

Methods inherited from interface weka.core.OptionHandler

makeCopy

Constructor Detail

FlatToRelational

public FlatToRelational()

Method Detail

globalInfo

public java.lang.String globalInfo()

Specified by:

globalInfo in class weka.filters.SimpleFilter

setInputFormat

public boolean setInputFormat(weka.core.Instances instanceInfo)
                       throws java.lang.Exception

This method needs to be called before the filter is used to transform instances. Will return true for this filter because the output format can be collected immediately. This specialised version of setInputFormat() calls the setInputFormat() method in SimpleFilter and then sets the output locators for relational and string attributes correctly. This is necessary because the setInputFormat() method in SimpleFilter calls setOutputFormat(), which initialises the output locators incorrectly in the case of this filter because the filter adds a relational attribute to the output. We also set the input locators correctly here, making sure that only attributes that are kept but not placed into the relation-valued attribute are covered by the locators.

Overrides:

setInputFormat in class weka.filters.SimpleFilter

Throws:

java.lang.Exception

getCapabilities

public weka.core.Capabilities getCapabilities()

Returns the Capabilities of this filter.

Specified by:

getCapabilities in interface weka.core.CapabilitiesHandler

Overrides:

getCapabilities in class weka.filters.Filter

getNumVariables

@OptionMetadata(displayName="number of variables",
                description="The number of variables in the timeseries (default = 1, univariate time series)",
                commandLineParamName="N",
                commandLineParamSynopsis="-N <int>",
                displayOrder=0)
public int getNumVariables()

setNumVariables

public void setNumVariables(int numVariables)

getRange

@OptionMetadata(displayName="attribute range",
                description="The attributes to transform into a multivariate timeseries, skipping the class if set (default = \"first-last\")",
                commandLineParamName="R",
                commandLineParamSynopsis="-R <string>",
                displayOrder=1)
public java.lang.String getRange()

setRange

public void setRange(java.lang.String range)

getDoNotKeepOtherAttributes

@OptionMetadata(displayName="do not keep other selected attributes",
                description="Whether not to keep (non-class) attributes that are outside the range.",
                commandLineParamName="K",
                commandLineParamSynopsis="-K",
                commandLineParamIsFlag=true,
                displayOrder=2)
public boolean getDoNotKeepOtherAttributes()

setDoNotKeepOtherAttributes

public void setDoNotKeepOtherAttributes(boolean keepOtherAttributes)