Turkish Journal of Computer and Mathematics Education Vol.12 No. 11 (2021),672 - 680
Research Article
672
An Approach For Combining Spatial And Textual Skyline Querying Using Indexing
Mechanism
Swathi Sowmya Bavirthi1 K.P.Supreethi2,
1Assistant Professor, Chaitanya Bharathi Institute of Technology(A), Gandipet, Hyderabad-75, Telangana, India 2Professor, Department of CSE, JNTUH College of Engineering, JNTUH, Hyderabad-85, Telangana, India
Article History: Received: 11 January 2021; Revised: 12 February 2021; Accepted: 27 March 2021; Published online: 10 May 2021
ABSTRACT: Most research works focus on either static data query processing or spatial data querying processing due to its large-sized data and the complex challenges that occur during its exploration, which is imperative. There are many state-of-the-art spatial indexing mechanisms for skyline querying to support either point/static or spatial objects, an indexing mechanism for skyline querying that supports both textual and spatial object querying is explored in this paper using an R* tree indexing mechanism that inserts data faster than its variant R tree technique and takes minimum execution time than the traditional spatial indexing mechanisms. The textual data is indexed using inverted files and attached to the R* tree. The proposed algorithm also works competitively under dynamic databases.
KEYWORDS: Spatial data mining, skyline querying, textual objects, spatial objects, R* tree, spatial querying, textual querying.
1. INTRODUCTION
Spatial data mining [1] [2] [3] is a popular paradigm that utilizes multidimensional [4] [5] datasets for extracting interesting objects and identifying the association between spatial and non-spatial data. Skyline querying has attracted much attention and has a significant role in multicriteria decision-making. Given a set of n-dimensional objects, the skylines of that set are the undominated points in that set. For example, p dominates pi, where p is having importance as
equal as pi in all the attributes and better than at least one attribute.
The important applications [6] of skyline querying are geographical information systems, recommendation systems, etc., for supporting decision making in areas like recommending hotels, location-based services [7], utility development, mobile workforce management, and restaurants recommendation, etc. In recommendation systems, the query of the user and the location of the user is to be considered and based on these two parameters, appropriate results are to be found. For determining appropriate results there are several methods [8] [9] that depend on either point or spatial objects, but a method depending on both the point/static in this case textual object and spatial object has less research exposure. In this scenario, a tree-based indexing technique called R* tree and an Inverted index technique is exploited to address the above-mentioned issue.
This paper proposes a spatial skyline querying using keywords with an R* tree indexing mechanism called the SSKQR* algorithm that investigates both the textual and spatial object identification using keyword-based and spatial search methods. For indexing of objects, the R* tree spatial index mechanism is used and relevant depictions for the search query are obtained using the SSKQR* algorithm. The unnecessary depictions are pruned using an efficient sky R pruning strategy for obtaining the most relevant skylines for the user's search query in order to generate an IR tree. The proposed SSKQR* algorithm has obtained better performance results than the other state-of-art spatial indexing mechanisms and answering of queries is faster and yielded better results in terms of disk access, memory management and run time execution.
The remainder of the paper is sequenced as follows: Section 2 illustrates about literature review, the methodology is described in section 3, section 4 explains the environmental setup, and section 5 details about evaluation metrics and results, the conclusion and summary are exploited in section 6, and section 7 briefs about the future work.
2. LITERATURE REVIEW
In the past few years, extensive work has been done by many researchers on skyline query processing. L. Chen & Y. Gao et al [10] described a process for searching similarities and similarity joins using a space-filling curve and pivot-based B+ tree called SPB Tree that clusters data by using space-filling curve and leveraging B+ trees for efficient storage and similarity distance computations are reduced by using pivots and the proposed algorithm is compared with state-of-art methodologies and obtained better performance model in terms of cost and memory utilization. J. Zhang & X. Jiang et al [11] developed a solution for querying skylines on large datasets using a two-phase MapReduce framework which incorporates partitioning of data, filtering, and evaluating skylines parallelly. And proposed a grid-based presort partition algorithm to merge the skyline computations. The results showcased that the proposed framework performed efficiently for processing large datasets. L. Chen & X. Lian [12] illustrated a variant of querying skylines called metric skyline by using a pruning mechanism for metric skyline queries with metric index and the extensive experimentation showed that the proposed pruning and indexing technique outperformed in terms of answering metric skyline queries. Cong G. & Jensen CS et al [13] explained a framework for top-k spatial web object retrieval using the R-tree indexing technique by
673
incorporating an efficient pruning technique and the experimental results states that the proposed method has obtained better results in terms of scalability. David F. &Jin R. et al [14] addressed the problem of querying a data point that does not have any other points in the database with equal or lesser distance from all the query points then the data point is considered as the closest point to the query point, by introducing an efficient non-skyline pruning strategy with optimization techniques for minimizing distance computation costs and dominance tests. The experimentation states that the proposed model outperformed the state-of-art methodologies. Bartolini I., & P. Ciaccia et al [15] introduced a method called SaLSa that presorts the input data for limiting the tuples to read and for comparison using symmetric sorting functions and the results proved that the proposed algorithm outperformed state-of-art skyline algorithms. Kian-Lee Tan and Beng Chin Ooi et al [16] addressed the problem in distributed environment spatial joining processing by proposing an operator called spatial semijoin for pruning objects that are not related to joining conditions, also examined to distributed algorithms using multidimensional indexing R-tree and others using single dimensional object mapping. The results indicate that the proposed algorithms performed well on real datasets. H. Lu & C. S. Jensen et al [17] addressed the problem of actual cardinality to the desired cardinality of a skyline query by introducing a skyline ordering by partitioning dataset and maintaining the order in each partition, the results indicate that the demonstrated framework is efficient and scalable in terms of size constraints on querying skylines. A. poulivassilis& S. G. Hild [18] examined a hyper log graph-based technique for querying and updating databases by using nested graphs that are mapped to the database and discussed the process flow of formulating, evaluating, expressing, and optimizing of hyper log programs. D. Wu & M. L. Yiu et al [19] discussed the process of querying top-k spatial keywords using joint processing method with W-IR-Tree indexing technique, the empirical studies indicate that the discussed method performed well on real and synthetic datasets. K. Deng & X. Zhou et al [20] considered the problem of the multi-source querying process of skylines in a road network by reviewing different algorithms and found that the lower bound constraint algorithm outperformed the straightforward algorithms. W. Choi & L. Liu et al [21] presented an optimization technique for utilization of GPU for parallel computation of skylines over large datasets and the results showcased that the proposed technique outperformed the dominance tests conducted on CPU. R. C. wong and J. Pei et al [22] addressed the nominal attributes in dynamic preferencing and the effect of choosing the order of attributes in depicting skylines by developing two algorithms for finding in which order a point is in the skyline and to retrieve skylines in a specific order. The methods performed competitively under synthetic and real datasets.
According to the literature survey, many methods [23] [24] [28][29][30] were examined for skyline query processing using different indexing techniques. In this paper, a skyline querying mechanism using spatial and textual keywords based on R* tree indexing has been explored for improving the efficiency of the algorithm.
3. METHODOLOGY
The spatial skyline querying using textual keyword takes into account both the textual keywords and the user's location for querying the spatial data and returns the objects that are spatially and textually relevant to the keywords specified to the user and the location. Even though there were many state-of-the-art spatial skylines querying methodologies [25] [26] [27][31] have been proposed, a spatial skyline querying using textual keywords with R* tree indexing is exploited in this paper, which uses regular R-tree algorithm for operations like query and deletes but uses a different strategy for insertion. The architecture of the system is highlighted in Figure-1.
Figure-1. Architecture of the system
For defining the formal introduction to indexing using the R* tree mechanism, the following set of terms are introduced. • Area-value: area [bbx (1st group)] + area [bbx (2nd group)]
674
• Margin-value: margin [bbx (1st group)] + margin [bbx (2nd group)] • Overlap-value: area [bbx (1st group) ∩ bbx (2nd group)]
Here bbx represents the bounding box of the rectangles. These values are also used for splitting of the node. Let the entries of the current node be El and Ep, then
Overlap (Ek) = ∑ area(ElRectangle ∩ Ek Rectangle) P
l=1,γ≠k
, l ≤ k ≤ P
Beginning with the SubtreeSelection algorithm, it determines the most suitable subtree for accommodating the new entries. Unlike the previous versions of R-tree along with area parameter, margin and overlap parameters are also considered, whichare defined previously,
Algorithm: SubtreeSelection
N: ← root
1: If N is a leaf node then return N Else
If N’s child-pointers point to leaves, select from the entries of N which require minimum rectangle overlap for new data inclusion then
The entry with the rectangle of smallest area If N’s child-pointers do not point to leaves then
Select from the entries of N which require area enlargement from inclusion of new data End
From the chosen entry, set N as the child node which is pointed by the child-pointer and goto (1).
The input for the SubtreeSelection () algorithm is a level parameter. If the level passed is a leaf node then it will be returned otherwise if the node points to any other leaves or not, is verified. If yes, then the entries with the smallest area of the rectangle will be considered. Otherwise, entries of N that require area enlargement are considered and chosen as a child as node pointed by the child-pointer.
The process of inserting the data into the tree actually starts by calling SubtreeSelection () to find the appropriate node to insert the new entry. The number of entries to be inserted into the tree M is compared with the nodes of N. If N has minimum entries than the M entries then entry E will be inserted otherwise Overflow will be handled by invoking either split or reinsertion mechanism and finally rectangles bounding boxes are adjusted as per the updated results in the insertion path.
Algorithm: Insert
Determine node N for placing new data E with level as parameter by calling SubtreeSelection. If entries of N are less than M then
accommodate E in N If N and M have equal entries then
with N as level parameter, invoke Overflow If split is performed invoking Overflow then
propagation of overflow in upwards is necessary If root is splitted by invoking Overflow then
new root is created
In insertion path, adjustment of rectangles is to be done such that minimum bounding rectangles enclosing their children For the user’s query location, the nodes that are close to it are obtained by searching the R* tree, and then an inverted file index data structure is constructed using textual keywords of the query. The obtained nodes relevant to the textual keywords are identified for pruning and then passed on to the skyline querying algorithm for the generation of skylines. The sample inverted file structure after R* tree indexing is shown in Figure-2.For the user’s query location, the nodes that are close to it are obtained by searching the R* tree, and then an inverted file index data structure is constructed using textual keywords of the query. The obtained nodes relevant to the textual keywords are identified for pruning and then passed on to the skyline querying algorithm for the generation of skylines. The sample inverted file structure after R* tree indexing is shown in Figure-2.
675
Figure-2. Inverted file structure after R* tree
The inverted file structure for the given query location is given as input for the SSKQR* algorithm for obtaining the skylines. The spatial skyline querying algorithm using textual keywords with R* tree indexing technique is as shown below:
Algorithm: SSKQR*
With rStarTree.root create priority heap Initialize empty skyline set
Until heap becomes empty loop through it
for next iteration, get (priority, key) minimum heap key from heap item get key value
for dominance of the key, loop all skylines
check items of the skyline set for dominance key if dominated key is found then
proceed with another key
if dominated key is found and removed from the heap then proceed with next key
if dominated key is not found
expansions are to be done, as we are not at leaf node else
insert tuple into skyline set, as we are at leaf node return skylines
For spatial skyline querying using keywords with the R* tree indexing technique, the priority heap, and empty skyline set are created for skylines. Then minimum heap and key are obtained and checked for dominance of the key, if dominated by a skyline then the process is continued with another key. If the key is removed from the heap even though it is dominated then continue the process with the next key. If no skyline dominates the key then proceed with expansions as we are away from the leaf node. Otherwise, if we are at leaf node insert into the skyline set.
4. EXPERIMENTAL SETUP
The entire experimentation was implemented in a system having 64-bit Windows 10 operating system with Intel® Core™ i5 CPU @ 2.4 GHz Processor, 8 GB RAM, and 1 TB Hard Disk installed with PyCharm, Python platform with its supporting packages as its configuration.
The overall experimentation was carried out using a publicly available Kaggle dataset namely TripAdvisor, consisting of 2328 reviews of a restaurant with its reference phrase as London. The dataset describes the reviews and reviewer's information of hotels as hotel name, hotel position in the city, number of remarks, URLs, etc., with a total size of 1.2 MB. The whole dataset is sub-divided into different chunks and each of which comprises 200, 400, 600, and 800, reviews of hotels having 45.8 KB, 56.0 KB, 64.0 KB, and 72.0 KB as each sample size, for measuring the efficiency of the proposed spatial skyline querying using keywords in terms of memory usage and execution time.
5. PERFORMANCE ANALYSIS AND EXPERIMENTAL RESULTS
The proposed spatial skyline querying using textual keywords with R* tree indexing technique is tested for its performance in terms of disk access rates, memory usage, and runtime execution using different sizes of the same dataset for performance measurement in terms of small and large data streams. Different query locations and different keywords
676
are also extensively used for performance analysis. From Fig-1 to Fig-3, the analysis results of the proposed spatial skyline querying using the R* tree indexing technique when compared with the R-tree and other indexing techniques are presented. The results state that the proposed SSKQR* algorithm has obtained better results in terms of disk access rates when compared to RS_IR, SKQ, and SSKQR algorithms.
Fig-1. Different query locations for disk access Fig-2. Different query keywords for disk access
Fig-3. Different sized datasets for disk access
Fig-4 to Fig-9 represents the performance analysis of the proposed SSKQR* algorithm with RS_IR, SKQ, and SSKQR. The results showcased that the proposed SSKQR* algorithm achieved better performance in terms of runtime execution and approximately similar performance in terms of memory utilization.
677
Fig-4. Different query locations for memory usage Fig-5. Different query keywords for memory usage
678
Fig-8. Different query keywords for runtime Fig-9. Different sized datasets for runtime
6. CONCLUSION AND SUMMARY
In this paper, an effective method using spatial skyline querying using textual keywords with R* tree indexing technique is exploited, for determining spatial objects that are related to the textual keywords in the query based on spatial and keyword-based search mechanisms, and to address the challenges of R-tree indexing. To address the issues, a TripAdvisor dataset which is publicly available is utilized and the R* tree is constructed and a sky-R pruning strategy has been used for unnecessary depictions removal that is obtained during querying of textual queries that are relevant. The experimental results indicate that the proposed spatial skyline querying using textual keywords with R* tree indexing mechanism has shown minimum disk access and better runtime execution & memory utilization than the other state-of-art methods, alongside the SSKQR*, has shown its efficiency in dynamic streaming databases.
7. FUTURE WORK
As the SSKQR* algorithm has been implemented using the R* tree indexing data structure, the other variants of R tree indexing mechanisms are to be investigated for supporting different skyline queries.
REFERENCES
[1] Sharifzadeh, M., C. Shahabi and L. Kazemi. “Processing spatial skyline queries in both vector spaces and spatial network databases.” ACM Trans. Database Syst. 34 (2009): 14:1-14:45.
[2] Hui Lin and Bo Huang, "SQL/SDA: a query language for supporting spatial data analysis and its Web-based implementation," in IEEE Transactions on Knowledge and Data Engineering, vol. 13, no. 4, pp. 671-682, July-Aug. 2001, doi: 10.1109/69.940739.
[3] G. Proietti and C. Faloutsos, "Analysis of range queries and self-spatial join queries on real region datasets stored using an R-tree," in IEEE Transactions on Knowledge and Data Engineering, vol. 12, no. 5, pp. 751-762, Sept.-Oct. 2000, doi: 10.1109/69.877506.
[4] X. Miao, Y. Gao, B. Zheng, G. Chen and H. Cui, "Top-k dominating queries on incomplete data," 2016 IEEE 32nd International Conference on Data Engineering (ICDE), Helsinki, 2016, pp. 1500-1501, doi: 10.1109/ICDE.2016.7498394.
[5] X. Liu, D. Yang, M. Ye and W. Lee, "U-Skyline: A New Skyline Query for Uncertain Databases," in IEEE Transactions on Knowledge and Data Engineering, vol. 25, no. 4, pp. 945-960, April 2013, doi: 10.1109/TKDE.2012.33.
[6] X. Lin, J. Xu and H. Hu, "Range-Based Skyline Queries in Mobile Environments," in IEEE Transactions on Knowledge and Data Engineering, vol. 25, no. 4, pp. 835-849, April 2013, doi: 10.1109/TKDE.2011.229.
679
[7] Son W., Lee MW., Ahn HK., Hwang S. (2009) Spatial Skyline Queries: An Efficient Geometric Algorithm. In: Mamoulis N., Seidl T., Pedersen T.B., Torp K., Assent I. (eds) Advances in Spatial and Temporal Databases. SSTD 2009. Lecture Notes in Computer Science, vol 5644. Springer, Berlin, Heidelberg. https://doi.org/10.1007/978-3-642-02982-0_17.
[8] E. Tiakas, A. N. Papadopoulos and Y. Manolopoulos, "Skyline queries: An introduction," 2015 6th International Conference on Information, Intelligence, Systems and Applications (IISA), Corfu, 2015, pp. 1-6, doi: 10.1109/IISA.2015.7388053.
[9] S. Borzsony, D. Kossmann and K. Stocker, "The Skyline operator," Proceedings 17th International Conference on Data Engineering, Heidelberg, Germany, 2001, pp. 421-430, doi: 10.1109/ICDE.2001.914855.
[10] L. Chen, Y. Gao, X. Li, C. S. Jensen and G. Chen, "Efficient Metric Indexing for Similarity Search and Similarity Joins," IEEE Transactions on Knowledge and Data Engineering, vol. 29, no. 3, pp. 556-571, 1 March 2017, doi: 10.1109/TKDE.2015.2506556.
[11] J. Zhang, X. Jiang, W. Ku and X. Qin, "Efficient Parallel Skyline Evaluation Using MapReduce," IEEE Transactions on Parallel and Distributed Systems, vol. 27, no. 7, pp. 1996-2009, 1 July 2016, doi: 10.1109/TPDS.2015.2472016.
[12] L. Chen and X. Lian, "Efficient Processing of Metric Skyline Queries," IEEE Transactions on Knowledge and Data Engineering, vol. 21, no. 3, pp. 351-365, March 2009, doi: 10.1109/TKDE.2008.146.
[13] Cong G, Jensen CS, Wu D. “Efficient Retrieval of the Top-k Most Relevant Spatial Web Objects,” International Conference on Very Large Data Bases. Proceedings. 2009; vol. 2, no. 1, pp. 337-348 August 2009, doi: 10.14778/1687627.1687666.
[14] David F., Jin R., Donghui D., "Efficient skyline computation in metric space", Proceedings of the 12th International Conference on Extending Database Technology: Advances in Database Technology, March 2009, pp. 1042-1051, doi: /10.1145/1516360.1516479.
[15] Bartolini, I., P. Ciaccia and M. Patella. “Efficient sort-based skyline evaluation.” ACM Trans. Database Syst. December 2008, vol. 33, no. 4, doi: 10.1145/1412331.1412343.
[16] Kian-Lee Tan, Beng Chin Ooi and D. J. Abel, "Exploiting spatial indexes for semi join-based join processing in distributed spatial databases," in IEEE Transactions on Knowledge and Data Engineering, vol. 12, no. 6, pp. 920-937, Nov.-Dec. 2000, doi: 10.1109/69.895802.
[17] H. Lu, C. S. Jensen and Z. Zhang, "Flexible and Efficient Resolution of Skyline Query Size Constraints," in IEEE Transactions on Knowledge and Data Engineering, vol. 23, no. 7, pp. 991-1005, July 2011, doi: 10.1109/TKDE.2010.47.
[18] A. Poulovassilis and S. G. Hild, "Hyperlog: a graph-based system for database browsing, querying, and update," in IEEE Transactions on Knowledge and Data Engineering, vol. 13, no. 2, pp. 316-333, March-April 2001, doi: 10.1109/69.917568.
[19] D. Wu, M. L. Yiu, G. Cong and C. S. Jensen, "Joint Top-K Spatial Keyword Query Processing," in IEEE Transactions on Knowledge and Data Engineering, vol. 24, no. 10, pp. 1889-1903, Oct. 2012, doi: 10.1109/TKDE.2011.172.
[20] K. Deng, X. Zhou and H. T. Shen, "Multi-source Skyline Query Processing in Road Networks," 2007 IEEE 23rd International Conference on Data Engineering, Istanbul, 2007, pp. 796-805, doi: 10.1109/ICDE.2007.367925. [21] W. Choi, L. Liu and B. Yu, "Multi-criteria decision making with skyline computation," 2012 IEEE 13th
International Conference on Information Reuse & Integration (IRI), Las Vegas, NV, 2012, pp. 316-323, doi: 10.1109/IRI.2012.6303026.
[22] R. C. Wong, J. Pei, A. W. Fu and K. Wang, "Online Skyline Analysis with Dynamic Preferences on Nominal Attributes," in IEEE Transactions on Knowledge and Data Engineering, vol. 21, no. 1, pp. 35-49, Jan. 2009, doi: 10.1109/TKDE.2008.115.
[23] Y. Hsueh, R. Zimmermann, W. Ku and Y. Jin, "SkyEngine: Efficient Skyline search engine for Continuous Skyline computations," 2011 IEEE 27th International Conference on Data Engineering, Hannover, 2011, pp. 1316-1319, doi: 10.1109/ICDE.2011.5767944.
[24] Y. Li, W. Qu, Z. Li, Y. Xu, C. Ji and J. Wu, "Skyline Query Based on User Preference with MapReduce," 2014 IEEE 12th International Conference on Dependable, Autonomic and Secure Computing, Dalian, 2014, pp. 153-158, doi: 10.1109/DASC.2014.36.
[25] D. Kim, H. Im and S. Park, "Computing Exact Skyline Probabilities for Uncertain Databases," in IEEE Transactions on Knowledge and Data Engineering, vol. 24, no. 12, pp. 2113-2126, Dec. 2012, doi: 10.1109/TKDE.2011.164.
[26] Zhiyong Huang, Hua Lu, Beng Chin Ooi and A. K. H. Tung, "Continuous Skyline Queries for Moving Objects," in IEEE Transactions on Knowledge and Data Engineering, vol. 18, no. 12, pp. 1645-1658, Dec. 2006, doi: 10.1109/TKDE.2006.185.
[27] Zou L., Chen L., Özsu M.T., Zhao D., “Dynamic Skyline Queries in Large Graphs”,in Kitagawa H., Ishikawa Y., Li Q., Watanabe C. (eds) Database Systems for Advanced Applications, DASFAA 2010, vol 5982, 2010, Springer, Berlin, Heidelberg. https://doi.org/10.1007/978-3-642-12098-5_5.
[28] Divanu Sameera, Vedavathi K, Durga Prasad Kavadi. (2020). A Novel Approach for Text Classification using Recurrent Neural Networks. International Journal of Advanced Science and Technology, 29(05), 9371-9386. . Vol. 29 No. 05 (2020): Vol. 29 No. 05 (2020)
680
[29] B. Madhuravani, Divya Priya Degala, M. Anjaneyulu, B.Dhanalaxmi, Prediction Exploration for Coronary Heart Disease Aid of Machine Learning, Turkish Journal of Computer and Mathematics Education, Vol.12 No.9 (2021), 312-316
[30] B. Madhuravani, , Gudipati Sri Sowmya, P. Sowjanya.Automatic prediction of silent heart attacks using neural networks based hybrid classification system, International Journal of Advanced Trends in Computer Science and Engineering, 2020, 9(5), pp. 7165–7169, 40
[31] B. Madhuravani, Dr DSR Murthy, A novel secure authentication approach for wireless communication using chaotic maps, Proceedings - International Conference on Trends in Electronics and Informatics, ICEI 2017, 2018, 2018-January, pp. 360–363
