Core.CBR.Distances.Locals

Class LocalDistanceSequencesDIST4MADCAPPablo

java.lang.Object

Core.CBR.Distances.Locals.LocalDistance

Core.CBR.Distances.Locals.LocalDistanceSequencesDIST4MADCAPPablo

public class LocalDistanceSequencesDIST4MADCAPPablo
extends LocalDistance

Specific implementation of LocalDistance

See Also:

LocalDistance

Constructor Summary

Constructors

Constructor and Description
LocalDistanceSequencesDIST4MADCAPPablo()

Method Summary

Methods

Modifier and Type	Method and Description
float	distance(java.lang.Object b, java.lang.Object a) Metric function used to assert if a sequence contains another.
float	similitude(java.util.ArrayList<java.lang.String> s1, java.util.ArrayList<java.lang.String> s2) Returns the similitude between a container sequence and a contained one.

Methods inherited from class Core.CBR.Distances.Locals.LocalDistance

computeDistance, setMissingMethod

Methods inherited from class java.lang.Object

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

Constructor Detail

LocalDistanceSequencesDIST4MADCAPPablo

public LocalDistanceSequencesDIST4MADCAPPablo()

Method Detail

distance

public float distance(java.lang.Object b,
             java.lang.Object a)

Metric function used to assert if a sequence contains another.

The algorithm has been designed with the purpose of return a weighted output depending on the following properties:

• A complete match of a pattern within a sequence returns a lower value than a partial match.
• A recent match (end of the sequence) returns a lower value than an old match (beginning of the sequence).
• A match where the events are closer between them returns a lower value than a match between distant events.
• The lowest value, the better.

Our notation assumes that within sequences, elements located at position 0 will be the last ones recorded, the newest ones, and the ones located at |s1|-1 or |s2|-1 the oldest ones. Basically, the algorithm to estimate the distances is composed of two parts. The first one is the index function, which return where the contained sequence element s2_j is located inside the stream sequence s1. The search will be done after the specified position i, since it represents where s2_j-1 was found. If s2_j cannot be found in stream s1, it will be assigned the maximum value possible |s1|. The second part is the algorithm itself, where, for each event pattern inside s2, positions retrieved by function index are exponentially accumulated. If the last contained sequence element was not found, the index i remains the same and we continue looking from there.

Specified by:

distance in class LocalDistance

Parameters:

b - Object.

a - Object.

Returns:

computeDistance(s1, s2)

similitude

public float similitude(java.util.ArrayList<java.lang.String> s1,
               java.util.ArrayList<java.lang.String> s2)

Returns the similitude between a container sequence and a contained one. The estimation is performed using the computeDistance function as base.

Parameters:

s1 - Container sequence

s2 - Contained sequence

Returns:

1-computeDistance(s1, s2)