Continous k-Dominant Skyline Computation on Multidimensional Data Streams

1
Continous k-Dominant Skyline Computation on
Multidimensional Data Streams

• SAC08

2
Outline

• Introduction
• Background Material
• CoSMuQ Algorithm
• Experimental Result
• Conclusion

2
3
Introduction

• Assume that a customer is interested in
  purchasing a PDA device.
• The customer focus on two important
  characteristic of a PDA, he screen size and the
  battery autonomy.

3
4
Background Material

• Given a d-dimensional space S s1 ,s2 ,, sD
• k-dominate
• k-dominant skyline

4
5
Example

• P4 is 5-dominanted by P2 . s1 s2 s3 s5 s6
  
• K-dominance is cyclic unlike d-dominance

Oid s1 s2 s3 s4 s5 s6
P1 4 4 4 2 2 2
P2 2 2 2 4 4 4
P3 3 3 3 1 3 3
P4 1 1 1 5 1 1
P5 2 2 2 3 3 3
5
6
Cont.

6
7
Cont.

• Subquery skylines

7
8
Cont.

• Query k-dominant skyline
• CoSMuQ Approach
• Update the grids and the skylines of the grids.
• Update the subqueries.
• Update the queries.

8
9
CosMuQ Approach

• Assume that there are 4-dimensional points,
  therefore T D 4.
• Query

9
10
Cont.
t5
10
t5
11
Cont.
11
12
Cont.

• Since t5 is skyline point in g1,2 , g2,4 , g3,4
  it will inserted in the (2, sq.ds)
• It will be inserted in the ( 2, sq.ds) subqueries
  sq2, sq3 and sq4 .

12
13
Cont.

• In our example t5 is not skyline point of g1,2,
  g1,3 and g2,3 ,therefore it is not
  (2,sq.ds)-skyline point of sq1.
• Then we check if the new point is dominated by
  the (2,sq.ds)-skyline point of sq1.
• It is not dominated, thus it is inserted in the
  candidate skyline.
• It is not dominated by any (sq.ds,sq.ds)-skyline
  point, thus it is inserted in them.
• Finally, we compare t5 with t4 and we discard
  t4

13
14
Performance Evaluation
14
15
Conclusion

• Paremeter k.
• In high dimensional Space.
• Dimension Cardinality Domains of Grid.

15