dynamic skylines considering range queries
DESCRIPTION
Dynamic Skylines Considering Range Queries. Speaker: Adam Adviser: Yuling Hsueh. 16th International Conference, DASFAA 2011. Wen-Chi Wang En Tzu Wang Arbee L.P. Chen3. INTRODUCTION. What is “Skyline” ?. INTRODUCTION. Dynamic skyline considering query - PowerPoint PPT PresentationTRANSCRIPT
Data Management+ Laboratory
Dynamic Skylines Considering Range Queries
Speaker: AdamAdviser: Yuling Hsueh
16th International Conference, DASFAA 2011
Wen-Chi Wang En Tzu Wang Arbee L.P. Chen3
INTRODUCTION
What is “Skyline” ?
DM+ Page 2
INTRODUCTION
Dynamic skyline considering query- Dynamic skyline query regarding query q retrieves the data points
not dynamically dominated by any other data points, with respect to q.
Dynamically dominated- A data point t (t[1], t[2],…,t[n]) is defined to dynamically dominate
another data point s (s[1], s[2],…,s[n]), with respect to query q (q[1], q[2],…,q[n]), iff
1) |t[i] − q[i]| ≤ |s[i] − q[i]|, i = 1 to n, and ∀2) at least in one dimension, say j, |t[j] − q[j]| < |s[j] − q[j]|.
DM+ Page 3
INTRODUCTION
1) |t[i] − q[i]| ≤ |s[i] − q[i]|, i = 1 to n, and ∀2) at least in one dimension, say j, |t[j] − q[j]| < |s[j] − q[j]|.
DM+ Page 4
INTRODUCTION
We turn to find the skyline in a transferred dataset in which all of the data points in the original space are transferred to the other space whose origin is equal to query.
DM+ Page 5
INTRODUCTION
Query=(2000, 4), C1=(1992, 8), C2=(1995, 8), C3=(1998, 3) = (|1992 − 2000|, |8 − 4|) = (8, 4), = (5, 4) and = (2, 1)
DM+ Page 6
INTRODUCTION
Dynamic skyline considering range queries
DM+ Page 7
PRELIMINARIES
Problem Formulation- Given an n-dimensional dataset D and a range query q ([q1, q1'],
[q2, q2'], …, [qn, qn']), where [qi, qi'] is an interval representing the user interests in the ith dimension, i = 1 to n∀ , the dynamic skyline query regarding q returns the data points from D, not dynamically dominated by any other data points, with respect to q.
DM+ Page 8
PRELIMINARIES
DM+ Page 9
PRELIMINARIES
DM+ Page 10
query q ([15, 20], [20, 25]), p8 = (17, 30)(|17 − 17|, |30 − 25|) = (0, 5) P7(|25 − 20|, |25 - 25|) = (5, 0), p3(|25 − 20|, |5 − 20|) = (5, 15)
PRELIMINARIES
DM+ Page 11
Data Structures Used in Algorithm- Grid index- Multidirectional Z-order curves
Grid index- Each dimension of the n-dimensional space is partitioned into b
blocks, each associated with an equal domain range of r.
PRELIMINARIES
DM+ Page 12
PRELIMINARIES
DM+ Page 13
Query cells: (3, 4), (3, 5), (4, 4), and (4, 5), range form: ([3, 4], [4, 5]) Pivot cells:([0, 2], [4, 5]), ([5, 7], [4, 5]), ([3, 4], [0, 3]), and ([3, 4], [6, 7])
PRELIMINARIES
DM+ Page 14
Z-order curve- point (5, 4) = (101, 100)- the Z-address of (5, 4) is (110010)
Monotonic Ordering of Z-order curve- a data point in a cell with a former order cannot be dominated by
the data points in the cells with the latter order
PRELIMINARIES
DM+ Page 15
Query (3, 4), p4 = (4, 4) (1, 0), p1 = (1, 6 ) (2, 2)
PRELIMINARIES
DM+ Page 16
Dynamic Skyline Processing
DM+ Page 17
Principle of Pruning Strategies
Dynamic Skyline Processing
DM+ Page 18
Principle of Pruning Strategies
Dynamic Skyline Processing
DM+ Page 19
Principle of Pruning Strategies
ALGORITHM
DM+ Page 20
EXPERIMENT
DM+ Page 21
EXPERIMENT
DM+ Page 22
EXPERIMENT
DM+ Page 23
CONCLUSIONS
DM+ Page 24
Author propose a new problem on dynamic skyline computation regarding a range query.
To efficiently answer this query, Author propose an approach based on the gird index and a newly designed variant of the well-known Z-order curve. By these two components, three efficient pruning strategies are devised, thus avoiding the need to scan the whole dataset for generating the transferred dataset and also reducing the times of dominance checking.
THE END
Thank you for listening!
DM+ Page 25
THE END
Q & A
DM+ Page 26