TimeSeriesClassificationData Class Reference

#include <TimeSeriesClassificationData.h>

Public Member Functions
	TimeSeriesClassificationData (UINT numDimensions=0, std::string datasetName="NOT_SET", std::string infoText="")
	TimeSeriesClassificationData (const TimeSeriesClassificationData &rhs)
virtual	~TimeSeriesClassificationData ()
TimeSeriesClassificationData &	operator= (const TimeSeriesClassificationData &rhs)
TimeSeriesClassificationSample &	operator[] (const UINT &i)
const TimeSeriesClassificationSample &	operator[] (const UINT &i) const
void	clear ()
bool	setNumDimensions (const UINT numDimensions)
bool	setDatasetName (const std::string datasetName)
bool	setInfoText (const std::string infoText)
bool	setClassNameForCorrespondingClassLabel (const std::string className, const UINT classLabel)
bool	setAllowNullGestureClass (const bool allowNullGestureClass)
bool	addSample (const UINT classLabel, const MatrixFloat &trainingSample)
bool	removeLastSample ()
UINT	eraseAllSamplesWithClassLabel (const UINT classLabel)
bool	relabelAllSamplesWithClassLabel (const UINT oldClassLabel, const UINT newClassLabel)
bool	setExternalRanges (const Vector< MinMax > &externalRanges, const bool useExternalRanges=false)
bool	enableExternalRangeScaling (const bool useExternalRanges)
bool	scale (const Float minTarget, const Float maxTarget)
bool	scale (const Vector< MinMax > &ranges, const Float minTarget, const Float maxTarget)
bool	save (const std::string &filename) const
bool	load (const std::string &filename)
bool	saveDatasetToFile (const std::string filename) const
bool	loadDatasetFromFile (const std::string filename)
bool	saveDatasetToCSVFile (const std::string &filename) const
bool	loadDatasetFromCSVFile (const std::string &filename)
bool	printStats () const
std::string	getStatsAsString () const
TimeSeriesClassificationData	partition (const UINT partitionPercentage, const bool useStratifiedSampling=false)
bool	merge (const TimeSeriesClassificationData &labelledData)
bool	spiltDataIntoKFolds (const UINT K, const bool useStratifiedSampling=false)
TimeSeriesClassificationData	getTrainingFoldData (const UINT foldIndex) const
TimeSeriesClassificationData	getTestFoldData (const UINT foldIndex) const
TimeSeriesClassificationData	getClassData (const UINT classLabel) const
UnlabelledData	reformatAsUnlabelledData () const
std::string	getDatasetName () const
std::string	getInfoText () const
UINT	getNumDimensions () const
UINT	getNumSamples () const
UINT	getNumClasses () const
UINT	getMinimumClassLabel () const
UINT	getMaximumClassLabel () const
UINT	getClassLabelIndexValue (const UINT classLabel) const
std::string	getClassNameForCorrespondingClassLabel (const UINT classLabel) const
Vector< MinMax >	getRanges () const
Vector< ClassTracker >	getClassTracker () const
Vector< TimeSeriesClassificationSample >	getClassificationData () const
MatrixFloat	getDataAsMatrixFloat () const

Protected Attributes
std::string	datasetName
	The name of the dataset.
std::string	infoText
	Some infoText about the dataset.
UINT	numDimensions
	The number of dimensions in the dataset.
UINT	totalNumSamples
	The total number of samples in the dataset.
UINT	kFoldValue
	The number of folds the dataset has been spilt into for cross valiation.
bool	crossValidationSetup
	A flag to show if the dataset is ready for cross validation.
bool	useExternalRanges
	A flag to show if the dataset should be scaled using the externalRanges values.
bool	allowNullGestureClass
	A flag that enables/disables a user from adding new samples with a class label matching the default null gesture label.
Vector< MinMax >	externalRanges
	A vector containing a set of externalRanges set by the user.
Vector< ClassTracker >	classTracker
	A vector of ClassTracker, which keeps track of the number of samples of each class.
Vector< TimeSeriesClassificationSample >	data
	The labelled time series classification data.
Vector< Vector< UINT > >	crossValidationIndexs
	A vector to hold the indexs of the dataset for the cross validation.
DebugLog	debugLog
	Default debugging log.
ErrorLog	errorLog
	Default error log.
WarningLog	warningLog
	Default warning log.

Detailed Description

GRT MIT License Copyright (c) <2012> <Nicholas Gillian, Media Lab, MIT>

Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.

Definition at line 42 of file TimeSeriesClassificationData.h.

Constructor & Destructor Documentation

GRT_BEGIN_NAMESPACE TimeSeriesClassificationData::TimeSeriesClassificationData	(	UINT	numDimensions = 0,
		std::string	datasetName = "NOT_SET",
		std::string	infoText = ""
	)

Constructor, sets the name of the dataset and the number of dimensions of the training data. The name of the dataset should not contain any spaces.

Parameters

numDimensions	the number of dimensions of the training data, should be an unsigned integer greater than 0
datasetName	the name of the dataset, should not contain any spaces
infoText	some info about the data in this dataset, this can contain spaces

Definition at line 25 of file TimeSeriesClassificationData.cpp.

TimeSeriesClassificationData::TimeSeriesClassificationData

(

const TimeSeriesClassificationData &

rhs

)

Copy Constructor, copies the TimeSeriesClassificationData from the rhs instance to this instance

Parameters

rhs	another instance of the TimeSeriesClassificationData class from which the data will be copied to this instance

Definition at line 42 of file TimeSeriesClassificationData.cpp.

TimeSeriesClassificationData::~TimeSeriesClassificationData ( )

virtual

Default Destructor

Definition at line 51 of file TimeSeriesClassificationData.cpp.

Member Function Documentation

bool TimeSeriesClassificationData::addSample	(	const UINT	classLabel,
		const MatrixFloat &	trainingSample
	)

Adds a new labelled timeseries sample to the dataset. The dimensionality of the sample should match the number of dimensions in the dataset. The class label should be greater than zero (as zero is used as the default null rejection class label).

Parameters

classLabel	the class label of the corresponding sample
trainingSample	the new sample you want to add to the dataset. The dimensionality of this sample (i.e. Matrix columns) should match the number of dimensions in the dataset, the rows of the Matrix represent time and do not have to be any specific length

Returns

true if the sample was correctly added to the dataset, false otherwise

Definition at line 131 of file TimeSeriesClassificationData.cpp.

void TimeSeriesClassificationData::clear

(

)

Clears any previous training data and counters

Definition at line 73 of file TimeSeriesClassificationData.cpp.

bool TimeSeriesClassificationData::enableExternalRangeScaling

(

const bool

useExternalRanges

)

Sets if the dataset should be scaled using an external range (if useExternalRanges == true) or the ranges of the dataset (if false). The external ranges need to be set FIRST before calling this function, otherwise it will return false.

Parameters

useExternalRanges

sets if these ranges should be used to scale the dataset

Returns

returns true if the useExternalRanges variable was set, false otherwise

Definition at line 278 of file TimeSeriesClassificationData.cpp.

UINT TimeSeriesClassificationData::eraseAllSamplesWithClassLabel

(

const UINT

classLabel

)

Deletes from the dataset all the samples with a specific class label.

Parameters

classLabel

the class label of the samples you wish to delete from the dataset

Returns

the number of samples deleted from the dataset

Definition at line 168 of file TimeSeriesClassificationData.cpp.

TimeSeriesClassificationData TimeSeriesClassificationData::getClassData

(

const UINT

classLabel

)

const

Returns the all the data with the class label set by classLabel. The classLabel should be a valid classLabel, otherwise the dataset returned will be empty.

Parameters

const

UINT classLabel: the class label of the class you want the data for

Returns

returns a dataset containing all the data with the matching classLabel

Definition at line 967 of file TimeSeriesClassificationData.cpp.

Vector< TimeSeriesClassificationSample > TimeSeriesClassificationData::getClassificationData ( ) const

inline

Gets the classification data.

Returns

a vector of TimeSeriesClassificationSample

Definition at line 445 of file TimeSeriesClassificationData.h.

UINT TimeSeriesClassificationData::getClassLabelIndexValue

(

const UINT

classLabel

)

const

Gets the index of the class label from the class tracker.

Returns

an unsigned int representing the index of the class label in the class tracker

Definition at line 1021 of file TimeSeriesClassificationData.cpp.

std::string TimeSeriesClassificationData::getClassNameForCorrespondingClassLabel

(

const UINT

classLabel

)

const

Gets the name of the class with a given class label. If the class label does not exist then the std::string "CLASS_LABEL_NOT_FOUND" will be returned.

Returns

a std::string containing the name of the given class label or the std::string "CLASS_LABEL_NOT_FOUND" if the class label does not exist

Definition at line 1031 of file TimeSeriesClassificationData.cpp.

Vector< ClassTracker > TimeSeriesClassificationData::getClassTracker ( ) const

inline

Gets the class tracker for each class in the dataset.

Returns

a vector of ClassTracker, one for each class in the dataset

Definition at line 438 of file TimeSeriesClassificationData.h.

MatrixFloat TimeSeriesClassificationData::getDataAsMatrixFloat

(

)

const

Gets the data as a MatrixFloat. This returns just the data, not the labels. This will be an M by N MatrixFloat, where M is the number of samples and N is the number of dimensions.

Returns

a MatrixFloat containing the data from the current dataset.

Definition at line 1062 of file TimeSeriesClassificationData.cpp.

std::string TimeSeriesClassificationData::getDatasetName ( ) const

inline

Gets the name of the dataset.

Returns

returns the name of the dataset

Definition at line 368 of file TimeSeriesClassificationData.h.

std::string TimeSeriesClassificationData::getInfoText ( ) const

inline

Gets the infotext for the dataset

Returns

returns the infotext of the dataset

Definition at line 375 of file TimeSeriesClassificationData.h.

UINT TimeSeriesClassificationData::getMaximumClassLabel

(

)

const

Gets the maximum class label in the dataset. If there are no values in the dataset then the value 0 will be returned.

Returns

an unsigned int representing the maximum class label in the dataset

Definition at line 1009 of file TimeSeriesClassificationData.cpp.

UINT TimeSeriesClassificationData::getMinimumClassLabel

(

)

const

Gets the minimum class label in the dataset. If there are no values in the dataset then the value 99999 will be returned.

Returns

an unsigned int representing the minimum class label in the dataset

Definition at line 996 of file TimeSeriesClassificationData.cpp.

UINT TimeSeriesClassificationData::getNumClasses ( ) const

inline

Gets the number of classes.

Returns

an unsigned int representing the number of classes

Definition at line 396 of file TimeSeriesClassificationData.h.

UINT TimeSeriesClassificationData::getNumDimensions ( ) const

inline

Gets the number of dimensions of the labelled classification data.

Returns

an unsigned int representing the number of dimensions in the classification data

Definition at line 382 of file TimeSeriesClassificationData.h.

UINT TimeSeriesClassificationData::getNumSamples ( ) const

inline

Gets the number of samples in the classification data across all the classes.

Returns

an unsigned int representing the total number of samples in the classification data

Definition at line 389 of file TimeSeriesClassificationData.h.

Vector< MinMax > TimeSeriesClassificationData::getRanges

(

)

const

Gets the ranges of the classification data.

Returns

a vector of minimum and maximum values for each dimension of the data

Definition at line 1041 of file TimeSeriesClassificationData.cpp.

std::string TimeSeriesClassificationData::getStatsAsString

(

)

const

Gets the dataset info (such as its name and infoText) and the stats (such as the number of examples, number of dimensions, number of classes, etc.) as a std::string.

Returns

returns a std::string containing the dataset stats

Definition at line 668 of file TimeSeriesClassificationData.cpp.

TimeSeriesClassificationData TimeSeriesClassificationData::getTestFoldData

(

const UINT

foldIndex

)

const

Returns the test dataset for the k-th fold for cross validation. The spiltDataIntoKFolds(UINT K) function should have been called once before using this function. The foldIndex should be in the range [0 K-1], where K is the number of folds the data was spilt into.

Parameters

const

UINT foldIndex: the index of the fold you want the test data for, this should be in the range [0 K-1], where K is the number of folds the data was spilt into

Returns

returns a test dataset

Definition at line 947 of file TimeSeriesClassificationData.cpp.

TimeSeriesClassificationData TimeSeriesClassificationData::getTrainingFoldData

(

const UINT

foldIndex

)

const

Returns the training dataset for the k-th fold for cross validation. The spiltDataIntoKFolds(UINT K) function should have been called once before using this function. The foldIndex should be in the range [0 K-1], where K is the number of folds the data was spilt into.

Parameters

const

UINT foldIndex: the index of the fold you want the training data for, this should be in the range [0 K-1], where K is the number of folds the data was spilt into

Returns

returns a training dataset

Definition at line 919 of file TimeSeriesClassificationData.cpp.

bool TimeSeriesClassificationData::load

(

const std::string &

filename

)

Load the data from a file. If the file format ends in '.csv' then the function will try and load the data from a csv format. If this fails then it will try and load the data as a custom GRT file.

Parameters

filename

the name of the file the data will be loaded from

Returns

true if the data was loaded successfully, false otherwise

Definition at line 317 of file TimeSeriesClassificationData.cpp.

bool TimeSeriesClassificationData::loadDatasetFromCSVFile

(

const std::string &

filename

)

Loads the classification data from a CSV file. This assumes the data is formatted with each row representing a sample. The first column should represent the timeseries counter. The class label should be the second column followed by the sample data as the following N columns, where N is the number of dimensions in the data.

Parameters

filename

the name of the file the data will be loaded from

Returns

true if the data was loaded successfully, false otherwise

Definition at line 584 of file TimeSeriesClassificationData.cpp.

bool TimeSeriesClassificationData::loadDatasetFromFile

(

const std::string

filename

)

Loads the labelled timeseries classification data from a custom file format.

Parameters

filename

the name of the file the data will be loaded from

Returns

true if the data was loaded successfully, false otherwise

Definition at line 377 of file TimeSeriesClassificationData.cpp.

bool TimeSeriesClassificationData::merge

(

const TimeSeriesClassificationData &

labelledData

)

Adds the data in the labelledData set to the current instance of the TimeSeriesClassificationData. The number of dimensions in both datasets must match. The names of the classes from the labelledData will be added to the current instance.

Parameters

const

TimeSeriesClassificationData &labelledData: the dataset to add to this dataset

Returns

returns true if the datasets were merged, false otherwise

Definition at line 789 of file TimeSeriesClassificationData.cpp.

TimeSeriesClassificationData & TimeSeriesClassificationData::operator=

(

const TimeSeriesClassificationData &

rhs

)

Sets the equals operator, copies the data from the rhs instance to this instance

Parameters

rhs	another instance of the TimeSeriesClassificationData class from which the data will be copied to this instance

Returns

a reference to this instance of TimeSeriesClassificationData

Definition at line 53 of file TimeSeriesClassificationData.cpp.

TimeSeriesClassificationSample& TimeSeriesClassificationData::operator[] ( const UINT &  i )

inline

Array Subscript Operator, returns the TimeSeriesClassificationSample at index i. It is up to the user to ensure that i is within the range of [0 totalNumSamples-1]

Parameters

i	the index of the training sample you want to access. Must be within the range of [0 totalNumSamples-1]

Returns

a reference to the i'th TimeSeriesClassificationSample

Definition at line 82 of file TimeSeriesClassificationData.h.

const TimeSeriesClassificationSample& TimeSeriesClassificationData::operator[] ( const UINT &  i ) const

inline

Const Array Subscript Operator, returns the TimeSeriesClassificationSample at index i. It is up to the user to ensure that i is within the range of [0 totalNumSamples-1]

Parameters

i	the index of the training sample you want to access. Must be within the range of [0 totalNumSamples-1]

Returns

a reference to the i'th TimeSeriesClassificationSample

Definition at line 93 of file TimeSeriesClassificationData.h.

TimeSeriesClassificationData TimeSeriesClassificationData::partition	(	const UINT	partitionPercentage,
		const bool	useStratifiedSampling = false
	)

Partitions the dataset into a training dataset (which is kept by this instance of the TimeSeriesClassificationData) and a testing/validation dataset (which is returned as a new instance of a TimeSeriesClassificationData).

Parameters

partitionPercentage	sets the percentage of data which remains in this instance, the remaining percentage of data is then returned as the testing/validation dataset
useStratifiedSampling	sets if the dataset should be broken into homogeneous groups first before randomly being spilt, default value is false

Returns

a new TimeSeriesClassificationData instance, containing the remaining data not kept but this instance

Definition at line 701 of file TimeSeriesClassificationData.cpp.

bool TimeSeriesClassificationData::printStats

(

)

const

Prints the dataset info (such as its name and infoText) and the stats (such as the number of examples, number of dimensions, number of classes, etc.) to the std out.

Returns

returns true if the dataset info and stats were printed successfully, false otherwise

Definition at line 661 of file TimeSeriesClassificationData.cpp.

UnlabelledData TimeSeriesClassificationData::reformatAsUnlabelledData

(

)

const

Reformats the TimeSeriesClassificationData as UnlabeledData so the data can be used to train unsupervised training algorithms such as K-Means Clustering and Gaussian Mixture Models.

Returns

a new UnlabelledData instance, containing the reformated timeseries classification data

Definition at line 977 of file TimeSeriesClassificationData.cpp.

bool TimeSeriesClassificationData::relabelAllSamplesWithClassLabel	(	const UINT	oldClassLabel,
		const UINT	newClassLabel
	)

Relabels all the samples with the class label A with the new class label B.

Parameters

oldClassLabel	the class label of the samples you want to relabel
newClassLabel	the class label the samples will be relabelled with

Returns

returns true if the samples were correctly relablled, false otherwise

Definition at line 223 of file TimeSeriesClassificationData.cpp.

bool TimeSeriesClassificationData::removeLastSample

(

)

Removes the last training sample added to the dataset.

Returns

true if the last sample was removed, false otherwise

Definition at line 197 of file TimeSeriesClassificationData.cpp.

bool TimeSeriesClassificationData::save

(

const std::string &

filename

)

const

Saves the data to a file. If the file format ends in '.csv' then the data will be saved as comma-seperated-values, otherwise it will be saved to a custom GRT file (which contains the csv data with an additional header).

Parameters

filename

the name of the file the data will be saved to

Returns

true if the data was saved successfully, false otherwise

Definition at line 306 of file TimeSeriesClassificationData.cpp.

bool TimeSeriesClassificationData::saveDatasetToCSVFile

(

const std::string &

filename

)

const

Saves the data to a CSV file. This will save the timeseries counter as the first column, the class label as the second column, and the sample data as the following N columns, where N is the number of dimensions in the data. Each row will represent a sample.

Parameters

filename

the name of the file the data will be saved to

Returns

true if the data was saved successfully, false otherwise

Definition at line 555 of file TimeSeriesClassificationData.cpp.

bool TimeSeriesClassificationData::saveDatasetToFile

(

const std::string

filename

)

const

Saves the labelled timeseries classification data to a custom file format.

Parameters

filename

the name of the file the data will be saved to

Returns

true if the data was saved successfully, false otherwise

Definition at line 328 of file TimeSeriesClassificationData.cpp.

bool TimeSeriesClassificationData::scale	(	const Float	minTarget,
		const Float	maxTarget
	)

Scales the dataset to the new target range.

Parameters

minTarget	the minimum range you want to scale the data to
maxTarget	the maximum range you want to scale the data to

Returns

true if the data was scaled correctly, false otherwise

Definition at line 286 of file TimeSeriesClassificationData.cpp.

bool TimeSeriesClassificationData::scale	(	const Vector< MinMax > &	ranges,
		const Float	minTarget,
		const Float	maxTarget
	)

Scales the dataset to the new target range, using the vector of ranges as the min and max source ranges.

Parameters

ranges	a vector of source ranges, should have the same dimensions as your data
minTarget	the minimum range you want to scale the data to
maxTarget	the maximum range you want to scale the data to

Returns

true if the data was scaled correctly, false otherwise

Definition at line 291 of file TimeSeriesClassificationData.cpp.

bool TimeSeriesClassificationData::setAllowNullGestureClass

(

const bool

allowNullGestureClass

)

Sets if the user can add samples to the dataset with the label matching the GRT_DEFAULT_NULL_CLASS_LABEL. If the allowNullGestureClass is set to true, then the user can add labels matching the default null class label (which is normally 0). If the allowNullGestureClass is set to false, then the user will not be able to add samples that have a class label matching the default null class label.

Parameters

allowNullGestureClass

true if you want to use the default null gesture as a label

Returns

returns true if the allowNullGestureClass was set, false otherwise

Definition at line 126 of file TimeSeriesClassificationData.cpp.

bool TimeSeriesClassificationData::setClassNameForCorrespondingClassLabel	(	const std::string	className,
		const UINT	classLabel
	)

Sets the name of the class with the given class label. There should not be any spaces in the className. Will return true if the name is set, or false if the class label does not exist.

Parameters

className	the new name for the class
classLabel	the label ID that you want to set the class name for

Returns

returns true if the name is set, or false if the class label does not exist

Definition at line 114 of file TimeSeriesClassificationData.cpp.

bool TimeSeriesClassificationData::setDatasetName

(

const std::string

datasetName

)

Sets the name of the dataset. There should not be any spaces in the name. Will return true if the name is set, or false otherwise.

Parameters

datasetName

the new name of the dataset

Returns

returns true if the name is set, or false otherwise

Definition at line 97 of file TimeSeriesClassificationData.cpp.

bool TimeSeriesClassificationData::setExternalRanges	(	const Vector< MinMax > &	externalRanges,
		const bool	useExternalRanges = false
	)

Sets the external ranges of the dataset, also sets if the dataset should be scaled using these values. The dimensionality of the externalRanges vector should match the number of dimensions of this dataset.

Parameters

externalRanges	an N dimensional vector containing the min and max values of the expected ranges of the dataset.
useExternalRanges	sets if these ranges should be used to scale the dataset, default value is false.

Returns

returns true if the external ranges were set, false otherwise

Definition at line 268 of file TimeSeriesClassificationData.cpp.

bool TimeSeriesClassificationData::setInfoText

(

const std::string

infoText

)

Sets the info std::string. This can be any std::string with information about how the training data was recorded for example.

Parameters

infoText

the infoText

Returns

true if the infoText was correctly updated, false otherwise

Definition at line 109 of file TimeSeriesClassificationData.cpp.

bool TimeSeriesClassificationData::setNumDimensions

(

const UINT

numDimensions

)

Sets the number of dimensions in the training data. This should be an unsigned integer greater than zero. This will clear any previous training data and counters. This function needs to be called before any new samples can be added to the dataset, unless the numDimensions variable was set in the constructor or some data was already loaded from a file

Parameters

numDimensions

the number of dimensions of the training data. Must be an unsigned integer greater than zero

Returns

true if the number of dimensions was correctly updated, false otherwise

Definition at line 79 of file TimeSeriesClassificationData.cpp.

bool TimeSeriesClassificationData::spiltDataIntoKFolds	(	const UINT	K,
		const bool	useStratifiedSampling = false
	)

This function prepares the dataset for k-fold cross validation and should be called prior to calling the getTrainingFold(UINT foldIndex) or getTestingFold(UINT foldIndex) functions. It will spilt the dataset into K-folds, as long as K < M, where M is the number of samples in the dataset.

Parameters

const	UINT K: the number of folds the dataset will be split into, K should be less than the number of samples in the dataset
const	bool useStratifiedSampling: sets if the dataset should be broken into homogeneous groups first before randomly being spilt, default value is false

Returns

returns true if the dataset was split correctly, false otherwise

Definition at line 814 of file TimeSeriesClassificationData.cpp.

---

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

• TimeSeriesClassificationData.h
• TimeSeriesClassificationData.cpp