md. rashidujjaman rifat, shubrami moutushy, abu wasif, dr ...cse.buet.ac.bd/news/poster/40.pdfmd....
TRANSCRIPT
k-Dense POI Search in Road Network
Md. Rashidujjaman Rifat, Shubrami Moutushy, Abu Wasif, Dr. Mohammed Eunus Ali
Department of Computer Science and Engineering (CSE), BUET
In a road network while travelling
from source to destination, k-
Dense POI Search finds a path
containing k- POIs where total
distance from current location to
destination including travel cost of
k-clustered POIs is minimum.
METHODOLOGY
OBJECTIVE This work aims to give flexibility to users’
for searching a facility point.
The main contributions are:
Finding the best set of k POIs
Minimize total distance from current
location to destination including travel
cost of clustered k POIs
Finding travel order among selected
POIs
Defining a distance for different types
of POIs which distinguishes a target
POIs for selection in the set of k-POI
from pruned one
PROBLEM DEFINITION
OUTCOME The major outcomes of our approach are:
• A new query processing to find a
path containing k-POIs where total
distance including the travel cost of
k-POIs is minimum
• An algorithm that identifies the
best set of POI to form k-cluster
and update the threshold value
efficiently
Input Output
n1
n5
n4
n6
n3
n2
tB
A
C
D
s
n1
n5
n4
n6
n3
n2
tB
A
C
D
s
KDPS(s,t,Q((q.x,q.y),k,G(V,E))
s: Source
q: Query point
t: destination
T: pharmacy(object type)
k=3
nodeobject/POI
n2
n6
n7
n4
o1
o2
s
o3
o4o13
o12
o9o10 o11
o7
o8
o14
o15
o16
o17
o6
o5
q
n3
n1
t
n5 n2
n6
n7
n4
o1
o2
s
o3
o4o13
o12
o9o10 o11
o7
o8
o14
o15
o16
o17
o6
o5
q
n3
n1
t
n5
base POItarget POI
Query point q
2
1
3
4 5
q
q Kdps outputs a preferred path
P={q, n1, n4, n5, n6, t} containing
3(k=3) pharmacy where the
particular drug is found..
O9 is the nearest POI to q. It is initial base
POI obi, based on which cluster and
threshold is calculated. Other target objects
are selected as base POI for clustering based
on threshold.
For any target object oi in target POI set, if the sum of linear
distance from source and from destination to oi is greater than
threshold, then oi can be pruned from search region.
For o5, Dlinear= dc+dd ≥ Dthreshold. So o5 is pruned.
o10 is selected as next base POI for finding another ‘better’ cluster. If new POI
exists (not already in target POI set),will be inserted into target POI set.
Process is repeated through finding base POI from target POI set, finding new
cluster and choosing better one and updating threshold until it finds the best
POI.
n6
n7
n4
o1
o2
s
o3
o4o13
o12
o9o10 o11
o7
o8
o14
o15
o16
o17
o6
o5
q
n3
n1
t
n5
4 4 5
node
Clustered
POIs2
8
n
RERERENCE [1] J. S. E. Martin, H. P. Kriegel and X. Xu.
A density-based algorithm for discovering
clusters in large spatial databases with
noise. In ACM SIGKDD, 1996.
n2
n6
n7
n4
o1
o2
s
o3
o4o13
o12
o9o10 o11
o7
o8
o14
o15
o16
o17
o6
o5
q
n3
n1
t
n5Cd
dd
nodepruned
POI
n2
n6
n7
n4
o1
o2
s
o3
o4o13
o12
o9o10 o11
o7
o8
o14
o15
o16
o17
o6
o5
q
n3
n1
t
n5
initial base POItarget POI
Finding cluster using €-link algorithm[1].
For €=4 and k=3,
Distance from o9 to o10, DL(o9, o10)≤4
Distance from o10 to o11, DL(o10, o11)≤5
Distance from o9 to n1, DL(o9, n1)≤4
Distance from n1 to o12, DL(n1, o12)≤4
Cluster C={o9, o10, o12}