A high-level class to perform the outlier detection. More...

#include <OutlierDetector.h>

Public Member Functions
	COutlierDetector ()
	Default constructor.

	COutlierDetector (TrajData *data)
	Constructor that accepts data.

bool	ResetOutlierDetector ()
	Resets the outlier detector object. More...

bool	PartitionTrajectory ()
	Partitions the trajectories in the data. More...

bool	DetectOutlier ()
	Finds outliers and stores them in the data member. More...

Public Attributes
TrajData *	m_data

Private Member Functions
bool	FindOptimalPartition (CTrajectory *pTrajectory)
	Finds optimal partitioning of a trajectory. More...

bool	StoreTrajectoryPartitionIntoIndex ()
	Saves the trajectory partitioning and precomputes some calculations (such as distances)

bool	CheckOutlyingProportion (CTrajectory *pTrajectory, float minProportion)
	Checks if the trajectory contains a significant portion of outlying partitions.

bool	GenerateAndSetupOutlier (CTrajectory *pTrajectory, int outlierId)
	Marks the given trajectory as an outlier in the data member. More...

int	ComputeModelCost (CTrajectory *pTrajectory, int startPIndex, int endPIndex)
	Measures the model cost of using a partition to represent part of a trajectory. More...

int	ComputeEncodingCost (CTrajectory *pTrajectory, int startPIndex, int endPIndex)
	Measures the encoding cost of using a partition to represent part of a trajectory. More...

float	MeasurePerpendicularDistance (CMDPoint s1, CMDPoint e1, CMDPoint s2, CMDPoint e2)
	Measures the perpendicular distance between two line segments (see paper for definition of perpendicular distance)

float	MeasureDistanceFromPointToLineSegment (CMDPoint s, CMDPoint e, CMDPoint *p)
	Measures the distance between a point and a line segment.

float	MeasureDistanceFromPointToPoint (CMDPoint point1, CMDPoint point2)
	Measures the distance between two points.

float	ComputeVectorLength (CMDPoint *vector)
	Computes the length of a vector.

float	ComputeInnerProduct (CMDPoint vector1, CMDPoint vector2)
	Computes the inner product between two vectors.

float	MeasureAngleDisntance (CMDPoint s1, CMDPoint e1, CMDPoint s2, CMDPoint e2)
	Measures the angle distance between two line segments (see paper for definition of angle distance)

float	ComputeDistanceBetweenTwoLineSegments (CMDPoint startPoint1, CMDPoint endPoint1, CMDPoint startPoint2, CMDPoint endPoint2)
	Computes the weighted distance between two line segments from component distances (see paper for definition of distance)

void	ComputeDistanceBetweenTwoLineSegments (CMDPoint startPoint1, CMDPoint endPoint1, CMDPoint startPoint2, CMDPoint endPoint2, float *distanceVector)
	Measures the perpendicular distance between two line segments (see paper for definition of perpendicular distance) and stores the result in distanceVector.

void	SubComputeDistanceBetweenTwoLineSegments (CMDPoint startPoint1, CMDPoint endPoint1, CMDPoint startPoint2, CMDPoint endPoint2, float &perpendicularDistance, float &parallelDistance, float &angleDistance)
	Measures component distances and stores them in the variables perpendicularDistance, parallelDistance, and angleDistance.

Private Attributes
int	m_nTotalLineSegments

vector< IntIntPair >	m_idArray

vector< PointPointPair >	m_lineSegmentArray

vector< FloatArray >	m_distanceIndex

FloatArray	m_lengthArray

vector< PartitionInfo * >	m_partitionInfoArray

CMDPoint	m_vector1

CMDPoint	m_vector2

CMDPoint	m_projectionPoint

float	m_coefficient

float	m_coefficient1

float	m_coefficient2

PartitionInfo	m_partitionInfo

float	m_distance1

float	m_distance2

float	m_theta

float	m_cosTheta

float	m_lowerBoundPerp

float	m_lowerBoundPara

float	m_lowerBoundAngle

float	m_upperBoundPerp

float	m_upperBoundPara

float	m_upperBoundAngle

map< IntIntPair, int >	m_closePartitionArrayIndexMap

vector< vector< IntIntPair > >	m_closePartitionArray

map< IntIntPair, float >	m_finePartitionLengthMap

map< IntIntPair, int >	m_coarsePartitionIdMap

Detailed Description

A high-level class to perform the outlier detection.

Mainly does management of the outlier detection (prunining of partitions for optimization, flagging trajectories as outliers, etc.) Much of the actual computation for outlier detection occurs in CDistanceOutlier.

Member Function Documentation

int COutlierDetector::ComputeEncodingCost	(	CTrajectory *	pTrajectory,
		int	startPIndex,
		int	endPIndex
	)

private

Measures the encoding cost of using a partition to represent part of a trajectory.

In the MDL description of the paper this is known as L(D|H). It is the distance between the original trajectory and the partition representation.

Parameters

[in]	pTrajectory	the trajectory that is being partitioned
[in]	startPIndex	the index of the staring point of the partition
[in]	endPIndex	the index of the ending point of the partition

int COutlierDetector::ComputeModelCost	(	CTrajectory *	pTrajectory,
		int	startPIndex,
		int	endPIndex
	)

private

Measures the model cost of using a partition to represent part of a trajectory.

In the MDL description of the paper this is known as L(H). It is the length of the partition.

Parameters

[in]	pTrajectory	the trajectory that is being partitioned
[in]	startPIndex	the index of the staring point of the partition
[in]	endPIndex	the index of the ending point of the partition

bool COutlierDetector::DetectOutlier ( )

Finds outliers and stores them in the data member.

Finds outliers in the data that is contained in the m_data field of the object. The outlying trajectories are stored in the m_outlierList member of the supplied data.

Returns: Successful or not.

bool COutlierDetector::FindOptimalPartition ( CTrajectory * pTrajectory )

private

Finds optimal partitioning of a trajectory.

Consults PARTITION_PRUNING_OPTIMIZATION to determine if the optimizations to the two-level partitioning should be used.

Parameters

[in] pTrajectory the trajectory to partition

bool COutlierDetector::GenerateAndSetupOutlier	(	CTrajectory *	pTrajectory,
		int	outlierId
	)

private

Marks the given trajectory as an outlier in the data member.

Returns: Successful or not.

bool COutlierDetector::PartitionTrajectory ( )

Partitions the trajectories in the data.

Consults parameter USE_CONVENTIONAL_PARTITONING to determine whether or not to use two-level partitioning scheme

bool COutlierDetector::ResetOutlierDetector ( )

Resets the outlier detector object.

Clears all of the internal arrays and resets the counters. Does not clear the trajectory data.

Returns: Successful or not

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

OutlierDetector.h
Measure.cpp
OutlierDetector.cpp

Public Member Functions

Public Attributes

Private Member Functions

Private Attributes

Detailed Description

Member Function Documentation