ai.libs.jaicore.ml.tsc.distances

Class MoveSplitMerge

java.lang.Object

ai.libs.jaicore.ml.tsc.distances.MoveSplitMerge

All Implemented Interfaces:

ITimeSeriesDistance

public class MoveSplitMerge
extends java.lang.Object
implements ITimeSeriesDistance

Implementation of the Move-Split-Merge (MSM) measure as published in "The Move-Split-Merge Metric for Time Series" by Alexandra Stefan, Vassilis Athitsos and Gautam Das (2013). The idea behind the MSM metric is to define a set of operations that can be used to transform any time series into any other.

• A Move operation changes the value of a single point of the time series.
• A Split operation splits a single point of the time series into two consecutive points that have the same value as the original point.
• A Merge operation merges two consecutive points that have the same value int a single point that has that value.

Each operation has an associated cost. The cost for a Move for x to y is |x-y|. The cost for a Split and Merge is defined by a constant c. Let S = (s_1, s_2, .., s_n) be a sequence of Move/Split/Merge operations with s_i either Move, Split or a Merge. The Move-Split-Merge distance between to time series A and B is defined be the cost of the lowest-cost transformation S*, such that transform(S*, A) = B.

Constructor Summary

Constructors

Constructor and Description
MoveSplitMerge(double c) Constructor.

Method Summary

Modifier and Type	Method and Description
double	distance(double[] a, double[] b) Calculates the distance between two time series.

Methods inherited from class java.lang.Object

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

Constructor Detail

MoveSplitMerge

public MoveSplitMerge(double c)

Constructor.

Parameters:

c - The constant cost for Split and Merge operations.

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.