Permu-pattern :discovery of mutable permutation patterns with proximity constraints

1
Permu-pattern discovery of mutable permutation
patterns with proximity constraints

• Meng Hu , Jiong Yang , Wei Su
• KDD 2008

2
Outline

• Introduction
• Problem definition (Mutable set and permu-pattern
  )
• The permu-pattern algorithm
• Experimental results
• Conclusions

3
Introduction

• In some applications different symbols may be
  considered the same due to their similarity
• ex In text mining , a set of synonymous words is
  a group of distinct symbols that should be
  treated as the same
• Mutable permutation patterns the order of
  symbols in a pattern could be altered in
  sequences and some symbols are interchangeable
  with other symbols
• Ex Google bought Youtube
• Youtube was acquired by Google
• Google, buy, acquire, Youtube
• Advantage It can capture not only the total
  order of symbols but also the permutated order of
  symbols

4
Problem definition

• Sequence data
• lts1,g1,s2,g2,sngt si symbols gi gap
• In text si keyword gi the gap between two
  keywords
• This paper focus on
• (1).the order of the symbols may not be important
  
• (2).the gap between two symbols is important

5
Cont.
6
(No Transcript)
7
Cont.

• example
• PPs1,s2,s3,s4,Tgap500
• S1s1,100,s2,250,s4,300,s3
• (s1,s4,s3) as a subsequence because the gap
  between s1 and s4,s4 and s3 are both below 500
• then S1 support PP
• S2(s1,100,s5,500,s4,300,s3)
• The distance between s1 and s4 is 600gtTgap
• then S2 does not support PP

8
Cont.
9
The permu-pattern algorithm

• Apriori-based algorithm can not be used to
  discover the frequent permutation pattern
• Ex S1(A,100,C,100,B) ,S2(B,100,C,100,A)
• Tsup2 ,Tgap150
• A,B is not a frequent mutable
  permutation pattern
• A,B,C is a frequent mutable
  permutation pattern

10
Cont.

• Reachable
• exS1(s2,100,s1,500,s4,100,s6,100,s7,800,s3)
  

Intermediate set
11
Cont.

• S1(s2,100,s1,500,s4,100,s6,100,s7,800,s3)
• S2(s1,50,s4,100,s2,300,s7,150,s4,150,s6)
• S3(s2,100,s1,150,s4,500,s5,100,s6,100,s7)
• s3,s7 mutable set ,Tsup3, Tgap200

s1 s2 s3,s7 s4 s5 s6
3 3 3 3 1 3
Support of each mutable set
12
Pruning strategies
13
(No Transcript)
14
Data structure for reachable case
Starting mutable set
ending mutable set
intermediate mutable set
15
Experimental results
16
Cont.
17
Conclusions

• It can be used in many application such as
  biological sequence analysis and text mining
• The gap threshold can be avoid through the
  following methods
• 1.remove stopwords from the documents when
  transforming a document into a sequence
• 2.treating certain types of punctuation as large
  gap