Journal of Computer Applications ISSN: 0974 1925, Volume-5, Issue EICA2012-5, February 10, 2012

Indexing and Ranking in Spatial Database

Abstract : In reality spatial objects (e.g., Dams) not only have spatial locations but also have quality attributes (e.g., height, reservoir capacity).Given a spatial location S, Quality vector and a set of spatial objects D, a spatial query which retrieves and ranks the objects that intersect the region S and satisfies the quality vector. We proposed an efficient index called SKR tree and algorithms which performs 1) spatial filtering ,2) textual filtering and 3) object ranking in a fully integrated manner. The SKR tree is compared with the R tree. Keywords: R-tree, SKR-tree ,Query Processing, Spatial database.

I.

INTRODUCTION

Spatial database systems manage large collections of geographic entities, which apart from spatial attributes contain non spatial information. Spatial objects in reality are associated with multiple quality attributes in addition to their spatial locations. Traditional spatial queries and joins focus on manipulating only spatial locations and distances, but they ignore the importance of quality attributes. The dominance comparison is suitable for comparing two objects with respect to multiple quality attributes. For the sake of simplicity, we assume that the domain of each quality attribute is fully ordered (e.g., integer domain). An object A is said to dominate another object B, if A is no worse than B for all quality attributes and A is better than B for at least one quality attribute. In this system, we study an interesting type of spatial queries, which select the best spatial location with respect to the quality of facilities in its spatial neighborhood. Given a set D of interesting objects (e.g., candidate locations) and quality vector, a top k s p a t i a l preference queries retrieves the k objects in D with the highest scores. The score of an object is defined by the quality of features (e.g., facilities or services) in its spatial neighborhood. As a motivating example, consider a database containing all information of dams. Here feature refers to specific facilities or services. A customer may want to rank the contents of this database with respect to the quality of their locations, quantified by aggregating non spatial characteristics of other features (e.g., height of dam, reservoir capacity etc.,). As another example, the user (e.g., a tourist) wishes to find a hotel that is close to a high quality restaurant and a high quality cafe. For the ease of discussion, the qualities are normalized to values in The score T(p) of a hotel is defined in terms of: 1) the maximum quality for each feature in the neighborhood region of p, and 2) the aggregation of those qualities.

Data Mining

EICA 175

Indexing and Ranking in Spatial Database

In Figure1a: a simple score instance, called the range score, binds the neighborhood region to a circular region at p with radius (shown as a circle), and the aggregate function to SUM. For example, the maximum quality of p1(dam1) are 0.9 and 0.6. Hence the (p1) is 0.9+0.6=1.5.Similarly for the dam p2, (p2) is 1.0+0.1=1.1.Hence the dam p1 is returned as top result. Figure 1b: shows a dam p5 and three features s1, s2, s3 (with their quality values). The circles have their radii as multiples of Now, the score of si is computed by multiplying its quality with the weight 2^-j, where j is the order of the smallest circle containing si.[1] Traditionally, there are two basic ways for ranking objects: 1) spatial ranking, which orders the objects according to their distance from a reference point, and 2) non spatial ranking, which orders the objects by an aggregate function on their non spatial values. The top-k spatial preference query integrates these two types of ranking in an intuitive way.

II. PROBLEM STATEMENT Let c be the number of (numeric) quality attributes.Each object p in the object dataset D is composed of a set of quality attributes c, and is associated with a location Lp. A quality vector is a point in the c-dimensional space where each dimension refers to a quality attribute. As a shorthand notation, we use [i] to represent the ith (quality) attribute value of . The notion of dominance is used to compare quality vectors. A quality vector is said to dominate another one (denoted as <), if [i] is better than [i] and ,is not worse than [i]. A location is a pair (x, y) in the Euclidean space, where x and y are the coordinates values. A spatial object p=<loc,> consists of both a location o.loc and a quality vector o.. The notation dist (p, p) denotes the Euclidean distance between the locations of the spatial objects p and p. Given two spatial objects p and p is said to be a dominator of p when p.<p.. Given a query q that specifies a set of query keywords Wq a query spatial scope Sq, textual relevance and spatial relevance of object p to q are formalized in definitions 1 and 2, respectively.

Krishnaveni S , TamizhSelvi S P

535

Journal of Computer Applications ISSN: 0974 1925, Volume-5, Issue EICA2012-5, February 10, 2012

Definition 1 (Textual relevance) An object p is said to be textually relevant to a query if o contains queried keywords i.e., Wp Wq0. To quantify the relevance of p to q, a weighting function denoted by q(o) is adopted. Thus for a given q, q(p1) >q(p2) means object p1 is more textually relevant to q than object p2. Definition 2(Spatial relevance)An object o is said to spatially relevant to a query q if the location of object o overlays with the query spatial scope of q, i.e., Lp Sq0. Let q(o) be a scoring function to quantify the spatial relevance of o to q. Thus for a given q, q(p1) >q(p2) means object p1 is more spatially relevant to q than object p2. Accordingly, k objects are retrieved from D objects that are both textually and spatially relevant to a given query and ranks objects based on the quality features of their objects. The spatial relevance of an object p, denoted as (p) depends on the types of the spatial relationships defined between an object location Lq. and a spatial scope S. Commonly adopted relationships include 1. Enclosed. (p) is set to 1 if the corresponding location is fully enclosed by the query scope, i.e.,

2.

Overlapping. (p) is set to the fraction of the object location that is covered by the spatial scope,.

3.

Proximity. (p) is represented by the inverse of the distance between the center of Lp and that of S,i.e.,

Figure2: Example of Farthest Dominated Location

EICA 175

Indexing and Ranking in Spatial Database

Dams h1 h2 h3 h4 h5 h6

Height m 180 150 190 250 190 220 (a)

Reservoir capacity 4 3 4 3 4 5

Loc S1 S2 S3 S4

NN h3 h4 h4 h6 (b)

ND h3 h3 h5 h5

Figure3:Lists of dams and candidate locations, (a) Qualities of dams, (b) Candidate location at (200,4) Definition 3 (Nearest dominator, Nearest Dominator Distance).Given a location S .its quality vector , and a set of spatial objects D, the nearest dominator of S in D is defined as ND(S,,D)= argmindist(S,p) pD,p.< i.e., the nearest neighbor of S in D among those that dominate . The nearest dominator distance ndd (S,,D)=dist(S,ND(S,,D)). Refer to the example in Fig 2 and 3, the ND of S is the dam hj that minimizes the dist(Sj,hj) value, among those dams dominating the design competence .In Fig 2.b. lists the NN and ND of each location Sj. It is important to note that NN is not necessarily the same as ND. For example, the NN of S2 is h4 which, however, does not dominate S2 with respect to its design competence. Whereas its next nearest neighbor h3 does, which exactly is S2s ND. By considering the distance of each location Sj from its ND, we pick the largest one(i.e., dist(S3,h5)), and take its location(i.e., S3) as the result location for building the new dam. Definition 4 (Farthest Dominated Location Query) Given a set of (competitors) spatial objects D, a set of (candidate)locations L, and a quality vector as the design competence, the farthest dominated location query returns from L a location S such that the distance ndd (S,,D) is maximized, i.e.,

Refer to the dam example in Figs. 2 and 3. There are c = 2 Quality attributes (i.e., height, reservoir capacity). The set of objects is D={h1,h2h6} and the set of locations is L={S1,S2S4) dam h1 is a spatial object, with a fixed location in the Euclidean space and the quality vector h1.=(180, 4). Let the design competence be (200, 4). Location s3 is the farthest dominated location and its nearest dominator is h5. III. SPATIAL QUERY EVALUATION ON R TREE The most popular spatial access method is the R-tree[3], which indexes minimum bounding rectangles (MBRs) of objects. Figure 4showsaset D{p1,p2..p8} ofspatialobjects (e.g., points) and an R-tree that indexes them.

Krishnaveni S , TamizhSelvi S P

537

Journal of Computer Applications ISSN: 0974 1925, Volume-5, Issue EICA2012-5, February 10, 2012

R-trees can efficiently process main spatial query types, including spatial range queries, nearest neighbor queries, and spatial joins. Given a spatial region S, a spatial range query retrieves from D the objects that intersect.

Figure4: Spatial Query Evaluation on R trees Table I List of Notations Notations D M Meaning An entry in the R-tree The object dataset The number of features An object point of D The c-the score of p Minimum distance between p and e Upper bound score an R-tree entry Radius Min Heap and Max Heap Location of object p S Spatial Scope

Wk Lp

For instance, consider a range query that asks for all objects within the shaded area in Fig.4 Starting from the root of the tree, the query is processed by recursively following entries, having MBRs that intersect the query region. For instance, e1 does not intersect the query region, thus the sub tree pointed by e1 cannot contain any query result. In contrast, e2 is followed by the algorithm and the points in the corresponding node are examined recursively to find the query result p7. The upper bound score of an object is calculated using the formula

Krishnaveni S , TamizhSelvi S P

538

EICA 175

Indexing and Ranking in Spatial Database

IV. SPATIAL QUERY EVALUATION ON SKR TREE In Figure 5assume if a user Alice in Boston issues a geographic query Bostons pollution control irrigation and the top-3 dams are to be returned. In this query, pollution control and irrigation are query keywords, and Boston represents a location/area of her interest. Regardless of the order of their relevance, this example shows a set of candidate objects {p2, p3, p4, p5 p6}, with respect to both textual relevance and spatial relevance. Here, p1, although being within Boston, is not relevant because it contains neither pollution control nor irrigation. On the other hand, {p7, p8, p9, and p10} although being textually relevant to the query, are not within Boston. The top-3 dams in the candidate set that are most relevant are returned.

No pollution control and irrigation Pollution control and irrigation

only irrigation only pollution control

Figure5: Spatial Data Distribution For this we proposed an efficient indexing scheme called SKR tree which indexes both the textual and spatial contents of objects to support data retrievals based on their combined textual and spatial relevance, which, in turn, can be adjusted with different relative weights. Fig.6. Shows the structure of SKR tree where each node has both spatial and non spatial information of the data object. No is the Root node which has two child nodes N1 and N2. The child node has the keyword AN1,and spatial data DN1 .This indexing scheme performs textual filtering and spatial filtering and hence it is efficient when compared with the R tree indexing scheme.

Krishnaveni S , TamizhSelvi S P

539

Journal of Computer Applications ISSN: 0974 1925, Volume-5, Issue EICA2012-5, February 10, 2012

Figure6: Structure of SKR Tree

V. ALGORITHMS FOR SPATIAL PREFERENCE QUERY A. Range Score Algorithm Given a set of objects, the range score algorithm calculates the Nearest neighbour distance and ranks the objects.[4]. Input: A set of data D, query location , Root of R tree and the radius . Output: Top k data with shortest distance Procedure: 1: Group_Range(NodeN) 2: foreachentrye Ndo 3: IfNisnonleafthen 4: readthechildnodeNIpointedbye; 5: Group_Range(NI); 6: else foreachp Vsuchthatdist then 7: 8: update Wk by e; B. Branch and Bound Algorithm Branch and bound algorithm which calculates the score and ranks the objects based on the non spatial quality features of the objects. Input: A set of data D and Query location and root of R tree Output: Top k data with the highest score. Procedure: 1: Cal_Score(NodeN) 2: foreachentrye Ndo 3: IfNis nonleafthen 4: readthechildnodeNIpointedbye; 5: Cal_Score(NI);
Krishnaveni S , TamizhSelvi S P 540

EICA 175

Indexing and Ranking in Spatial Database

6: else 7: foreachp Vsuchthat 8: for c:=1tomdo 9: compute forall 10: updateWk by e

concurrently;

C. SKR Tree Construction Algorithm Input: Set of Objects D Output: Root of SKR tree Procedure: 1: Ne0 2: For each p D do 3: geocode p and represent Lpwith MBB mp 4: if for some eNe, me=mp then 5: add p to es dataset De; 6: else 7: create a new entry e; 8: set memp and De{p}; 9: NeNeU{e}; 10: End if 11: End for 12: For each eNedo 13: While 1Ne1>nmax do 14: Cluster the data according to min/max into nodes 15: NeNe 16: End while 17: Create the root node to cover Ne and their data summary D. NDD algorithm This algorithm which finds the nearest dominated locatiors that satisfies the quality vector . Input: Root of SKR tree , set of locations L and quality vector Output: A set nearest dominators Procedure:

Krishnaveni S , TamizhSelvi S P

541

Journal of Computer Applications ISSN: 0974 1925, Volume-5, Issue EICA2012-5, February 10, 2012

E. NDD Score algorithm This algorithm which finds the score of all nearest dominated locations and ranks the objects. Input: Root of SKR tree , set of locations L and quality vector Output: Top k objects with highest score Procedure:

VI. EXPERIMENTAL EVALUATION In this section, we conduct experiments on real object and feature data sets in order to demonstrate the application of top-k spatial preference queries. We obtained real spatial data fromhttp://www.fao.org/nr/water/aquastat/gis/index2.stm location in these data sets correspond to (longitude and latitude) coordinates in Africa and Middle East countries. This Spatial dataset containing information about dams in South Africa and middle east countries which has several features like irrigation, water supply, flood control, Hydro electricity, navigation , recreation, pollution control, livestock rearing and others.We used SQL server 2008 for spatial database design. Spatial query visualizer to visualize the spatial query and spatial uploader to upload the shape files of Africa and middle east countries. The proposed indexing scheme is evaluated with the R tree. .

Figure7:Effect of IO Access Time for Different Trees in ms

Krishnaveni S , TamizhSelvi S P

542

EICA 175

Indexing and Ranking in Spatial Database

VII. CONCLUSION In this paper, we studied top-k spatial preference queries, which provide a novel type of ranking for spatial objects based on qualities of features in their neighborhood. Given a spatial location S, Quality vector and a set of spatial objects D, a spatial query which retrieves and ranks the objects that intersect the region S and satisfies the quality vector. We proposed an efficient index called STR tree and algorithms which performs 1) spatial filtering ,2) textual filtering and 3) object ranking in a fully integrated manner. In the future, we will study the top-k spatial preference query on a road network, in which the distance between two points is defined by their shortest path distance rather than their euclidean distance. The challenge is to develop alternative methods for computing the upper bound scores for a group of points on a road network.

REFERENCES 1. M.L.Yiu, X.Dai,N. Mamoulis and M. Vaitis, Top k Spatial Preference Queries, Proc. IEEE Intl Conf. Data Eng.(ICDE),2007. 2. Man Lung Yiu, Hua Lu, Nikos Mamoulis, and MichailVaitis Ranking spatial data by quality preferences IEEE Trans.on Data Eng. March 2011. 3. A.Guttman, R-Trees:A Dynamic Index Structure for Spatial Searching, Proc. ACMSIGMOD, 1984. 4. Y. Chen and J.M. Patel, Efficient Evaluation of All-Nearest-Neighbor Queries, Proc. IEEE Intl Conf. Data Eng. (ICDE), 2007. 5. N. Mamoulis, M.L. Yiu, K.H. Cheng, and D.W. Cheung, Efficient Top-k Aggregation of Ranked Inputs, ACM Trans. Database Systems, vol. 32, no. 3, p. 19, 2007. 6. David Noack, Spatial variation in search engine Results,Intlconf on system sciences2010. 7. Zhicheng Li, Hu Xu, Yansheng Lu, Ailing QianAggregate Nearest Keyword Search in Spatial Databases Intl Asia specific web conf. 2010. 8. RamondT.Ng,Jiaweihan,Effective and efficient clustering in spatial th datamining,proceedings of 20 VLDB conf,santiago,chile. 9. Muhammad AamirCheema ,Xuemin Lin , Haixun Wang , Jianmin Wang , Wenjie Zhang A unified Approach for Computing Top-k Pairs in Multidimensional Space. IEEE Intl Conf. Data Eng. (ICDE),2011. 10. Xiang Lian and Lei ChenRanked Query Processing in Uncertain Databases, IEEE Trans Data Eng.

Krishnaveni S , TamizhSelvi S P

543