ai.libs.jaicore.ml.tsc.distances

Class ShotgunDistance

java.lang.Object

ai.libs.jaicore.ml.tsc.distances.ShotgunDistance

All Implemented Interfaces:

ITimeSeriesDistance

public class ShotgunDistance
extends java.lang.Object
implements ITimeSeriesDistance

Implementation of Shotgun Distance measure as published in "Towards Time Series Classfication without Human Preprocessing" by Patrick Schäfer (2014). To make many of the standard methods to calculate the distance betwenn two time series applicable, a lot of time and effort has to be spent by a domain expert to filter the data and extrdact equal-length, equal-scale and alighned patterns. The Shotgun Distance avoids preprocessing the data for alignment, scaling or length. This is achieved by breaking the query time series into disjoint windows (subsequences) of fixed length. These windows are slid along the sample time series (with stride 1) to find the best matching position in terms of minimizing a distance metric (e.g. Euclidean distance or DTW distance).

Constructor Summary

Constructors

Constructor and Description
ShotgunDistance(int windowLength, boolean meanNormalization) Constructor for the Shotgun Distance.

Method Summary

Modifier and Type	Method and Description
double	distance(double[] a, double[] b) Calculates the distance between two time series.
void	setWindowLength(int windowLength) Sets the window length.

Methods inherited from class java.lang.Object

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

Constructor Detail

ShotgunDistance

public ShotgunDistance(int windowLength,
                       boolean meanNormalization)

Constructor for the Shotgun Distance.

Parameters:

windowLength - The window length.

meanNormalization - Mean normalization. If false, no mean substraction at vertical alignment of windows.

Method Detail

distance

public double distance(double[] a,
                       double[] b)

Description copied from interface: ITimeSeriesDistance

Calculates the distance between two time series.

Specified by:

distance in interface ITimeSeriesDistance

Parameters:

a - First time series.

b - Second time series.

Returns:

Distance between the first and second time series.

setWindowLength

public void setWindowLength(int windowLength)

Sets the window length.

Parameters:

windowLength - @see #windowLength