International Journal of Advanced Computational Engineering and Networking, ISSN: 2320-2106, Volume-4, Issue-12, Dec.

-2016

COMPARATIVE STUDY BETWEEN DENSITY BASED CLUSTERING

- DBSCAN AND OPTICS
1
PRANJAL DUBEY, 2ANAND RAJAVAT
1
PG Scholar, Department of CSE, SVITS, Indore, India
2
HOD, Department of CSE, SVITS, Indore, India
E-mail: [email protected], [email protected]

Abstract— Data mining is process of retrieving data and patterns from large database. Clustering is a phase of data mining
that cumulates the data and finds a proven structure from the database. A good clustering approach plays a major role in
detecting clusters of arbitrary shapes. In this paper I have discuss about the Density Based Clustering Spatial Clustering of
Applications with Noise (DBSACN) which finds out clusters of different shapes and size from a large database and
improves scalability and efficiency in a multiphase clustering. With this Ordering Points to Identify the Clustering Structure
(OPTICS) have been compared to identify similar objects based on their density, here one produces clusters and the other
outputs augmented ordering representing density-based structure of a database. The parameters and their optimisations are
also discussed.

Keywords— Clustering, density-based clustering, DBSCAN, OPTICS.

I. INTRODUCTION An innovative technique which is used to compare in

between two different clustering algorithms
Clustering is the primary data mining technique. It (DBSCAN and SNN) described several
can even be a stand-alone tool or a pre-processing implementations of the DBSCAN and SNN
step in other data mining applications. Clustering is a algorithms, two density-based clustering algorithms.
process of evolving similar objects from database and These implementations can be used to cluster sets of
grouping them into valid clusters. The different points based on their spatial density. The results
cluster formation follows different attributes and obtained through the use of these algorithms show
algorithm to form clusters therefore it may results in that SNN performs better than DBSCAN since it can
different outcomes. Clustering algorithm is applied in detect clusters with different densities while
many fields: pattern recognition, information DBSCAN cannot [3].
retrieval, image processing, machine learning. Many clustering algorithm have been proposed,
Density-based algorithm is simple and high efficiency seldom was focused on high dimensional and
algorithm [1]. incremental databases. An incremental approach on
Various methods are best suited for different Grid Density-Based Clustering Algorithm (GDCA)
databases. Here we are dealing with DBSCAN and discovers clusters with arbitrary shape in spatial
OPTICS which are used to detect clusters of different databases. It first partitions the data space into a
densities, shapes and sizes in spatial datasets with number of units, and then deals with units instead of
noise. points. Only those units with the density no less than
The paper is structured in following sections: section a given minimum density threshold are useful in
2 presents review density-based clustering. Section 3 extending clusters [4].
discusses about DBSCAN and clustering over it. An innovative approach presents a new density-based
Section 4 discusses about OPTICS and cluster clustering algorithm, ST-DBSCAN, which is based
structuring and formation. Section 5 concludes with on DBSCAN. It proposes three marginal extensions
parameters with respect to optimizing performance of to DBSCAN related with the identification of (i) core
density-based algorithms. objects, (ii) noise objects, and (iii) adjacent clusters.
II. LITERATURE REVIEW In contrast to the existing density-based clustering
Density-based clustering is highlighted by number of algorithms, this algorithm has the ability of
applications. Significant work has been done in this discovering clusters according to non-spatial, spatial
field of Density based clustering. One approach has and temporal values of the objects [5].
been developed the incremental clustering for mining The new concept is presented on clustering technique
large database. This approach present the first which provides an effective method for Clustering
incremental clustering algorithm based on DBSCAN Incremental Gene Expression data. It is designed
which is applicable on any database containing data based on density based approach where the efficiency
in a metric space. Due to the density-based nature of of GenClus in detecting quality clusters over gene
DBSCAN, the insertion or deletion of an object expression data. This work presents a density based
affects the current clustering only in the clustering approach which finds useful subgroups of
neighbourhood of this object. Thus, efficient highly coherent genes within a cluster and obtains a
algorithm scan be given for incremental insertions hierarchical structure of the dataset where the sub
and deletions to an existing clustering [2]. clusters give the finer clustering of the dataset [6].
Comparative Study Between Density Based Clustering - DBSCAN and Optics

34
International Journal of Advanced Computational Engineering and Networking, ISSN: 2320-2106, Volume-4, Issue-12, Dec.-2016

Comparative Study Between Density Based Clustering - DBSCAN and Optics

35
International Journal of Advanced Computational Engineering and Networking, ISSN: 2320-2106, Volume-4, Issue-12, Dec.-2016

Comparative Study Between Density Based Clustering - DBSCAN and Optics

36
International Journal of Advanced Computational Engineering and Networking, ISSN: 2320-2106, Volume-4, Issue-12, Dec.-2016

ACKNOWLEDGMENT

The authors would like to appreciate their teacher’s

for their valuable guidance and would thanks to their
family and friends for their valuable support.

REFERENCES

[1] Yaminee S. Patil, M. B. Vaidya “A technical survey on

clustering analysis in data mining” International Journal of
Emerging Technology and Advanced Engineering.
[2] Martin Ester,Hans-Peter Kriegel,Jorg Sander,Michael
Wimmer,Xiaowei Xu,”Incremental clustering for mining
in a data ware housing”, University of Munich
Oettingenstr. 67, D-80538 München, Germany.
[3] Adriano Morira, Marible Y.Santos, Sofia Carneiro,”
Density based clustering algorithms DBSCAN and SNN”,
University of Minho – Portugal, Version 1.0, 25.07.2005.
[4] CHEN Ning , CHEN An, ZHOU Long-xiang,”An
Incremental Grid Density-Based Clustering Algorithm”,
Journal of Software, Vol.13, No.1,2002.
[5] Naresh kumar Nagwani and Ashok Bhansali, “An Object
Oriented Email Clustering Model Using Weighted
Similarities between Emails Attributes”, International
Journal of Research and Reviews in Computer science
(IJRRCS), Vol. 1, No. 2, June 2010.
[6] Sauravjyoti Sarmah , Dhruba K. Bhattacharyya,”An
Effective Technique for Clustering Incremental Gene
Expression data”, IJCSI International Journal of Computer
Science Issues, Vol. 7, Issue 3, No 3, May 2010.
[7] Gray, G ‘Lecture 7&8: Proximity measures & clustering’,
2013, Lecture Notes.
[8] Ankerst, M & Breunig, MM & Kriegel, HP & Sander, J,
‘OPTICS: Ordering points to identify the clustering
structure’, 1999, ACM SIGMO International Conference
on Management of Data, 1999, pp. 49–60.
[9] Nidhi Suthar, Indrjeet Rajput, Vinit kumar Gupta” A
technical survey on DBSCAN clustering algorithm”
International Journal of Scientific and Engineering
Research, Volume 4, Issue 5, May 2013.
[10] Izabela Anna Wowczko” Density Based Clustering with
DBSCAN and OPTICS” Business Intelligence and Data
Mining, 2013.

Comparative Study Between Density Based Clustering - DBSCAN and Optics

37