IRJET- Customer Segmentation from Massive Customer Transaction Data

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 06 Issue: 06 | June 2019 www.irjet.net p-ISSN: 2395-0072
© 2019, IRJET | Impact Factor value: 7.211 | ISO 9001:2008 Certified Journal | Page 432
CUSTOMER SEGMENTATION FROM MASSIVE CUSTOMER TRANSACTION
DATA
Neethu CM1, Anitha Abraham2
1Post Graduate Student
2Assistant Professor, College of Engineering Kidangoor
---------------------------------------------------------------------***----------------------------------------------------------------------
Abstract - In this internet era, more and more people use
online shopping. Analysingmassivecustomertransaction data
about these online activities can be used to improve the
business and to satisfy customer demands in a better way. In
this research paper we try to study different methods
employed to analyse the customer transaction data. In our
study we have studied methods like K-Means clustering, PAM
clustering, Agglomerative, Divisive and Density Based
clustering methods. Based on ourstudywehave identifiedthat
K-Means is the widely used clustering method.
Key Words: Clustering, Partitional clustering, Hierarchical
Clustering
1. INTRODUCTION
Customer Segmentation is also known as clustering of
customers. Clustering can be considered themostimportant
unsupervised learning problem. It deals with finding a
structure in a collection of unlabelled data [1] et al says the
definition of clustering could be the process of organizing
objects into groups whose membersaresimilarinsome way.
A cluster is therefore a collection ofobjectswhicharesimilar
between them and are dissimilar to the objects belonging to
other clusters. [2] Clustering is most important technique
used in data mining. The most commonly used algorithms in
Clustering are Hierarchical and Partitioning. [3] Pramod
Gupta says that applications of clusteringinvariousfieldlike
In Biology, clustering has been used to find groups of genes
that have similar functions. In Information Retrieval,
clustering can be used to group search resultsofa queryinto
a small number of clusters, each of which capturing a
particular aspect of the query. In Geology, cluster analysis
has been applied to find patterns in the atmospheric
pressure of polar regions and areas of the ocean that have a
significant impact on land climate. In Medicine, cluster
analysis can also be used to detect patterns in the spatial or
temporal distribution of diseases like cancer, autism, etc.
2. RELATED WORK
Customer Segmentation is also called clustering of
customers. It achieving successful modern marketing and
customer relationship management. There are different
clustering methods available. Data objects are divided into
non overlapping clusters so that every and each object is in
one subset in partitional method. [12] explained K-Medoids,
PAM, CLARA (Clustering Large Applications) and
CLARANS. K-means is the most easy and wide usedpartition
method which describes details in [14],
[15],[16],[17],[18],[19]. L.Rokach et al [19] says that the K-
means algorithm may be viewed as a gradient-decent
procedure. PAM is similar to K-means and find median
instead of doing mean. PAM is to find the representative
object for each cluster, this representative object called
medoids which means that the most centrally located point
in the cluster. Consider Oj is data object and Om is selected
medoids [24] we can say that the Oj belongs to the cluster
Om. The quality of clustering is determined by average
dissimilarity between an object and the medoids of its
cluster. RuiXu et al proposed algorithm which is in [21].
J.Han et al [15] proposed CLARA algorithm is used to handle
large data set and relies on sampling. CLARAdrawn multiple
samples and gives best clustering results. Quality of
clustering is determined with average dissimilarity of all
objects in the entire data set in [19] describes the algorithm.
Han et al [15] proposed CLARANS method for clustering
polygonal objects. CLARANS is main memory clustering
techniques and it uses K-medoids method for clustering.
Because K-medoids is very robust and to avoid outliers.
Application is that CLARANS is used to cluster spatial
coordinates. Since the parameter maxneighbour is set to
high, it is very effectively same as the quality of the
clustering produced by PAM. Same as for lower value of
maxneighbour produces a lower clustering quality.
Hierarchical methodisoneoftheclusteringmethod.
It can be categorized into two, they are Agglomerative
method and Divisive method. Cen Li et al [22] proposed an
agglomerative method. It starts with many small clusters so
it is called bottom up approach. Each and every iteration
each smaller clusters combine to form large clusters. And
finally we got a large cluster. [15] The hierarchical clustering
methods could be further divided according to the measure
is calculated. Single-link cluster, Complete-link cluster,
Average-link cluster. [23] et al proposed Balanced Iterative
Reducing and Clustering using Hierarchies(ie)BIRCHitisan
agglomerate technique, it is used for big databases. This
method clusters the incomingmulti-dimensional metricdata
points incrementally and dynamically to provide quality
clusters. Guha et al [4] developed another agglomerative
hierarchical clustering algorithm, ROCK. Groupingdata with
attributes or distinct non-metric attributes. Thismethodisa
measurement of link used to reveal the relation between
combine of objects and their common behaviour. George
Karypis et al [24] developed a technique known as

CHAMELEON is type of clustering algorithm based on
agglomerative HC with k-nearest neighbour graph. In which
an edge is eliminated if both vertices are not within the k-
nearest closest points related to each other. In this method,
authors ignored the issue of scaling to large data sets that
cannot fit in the main memory. CLICK is another
agglomerative algorithm based on the calculation of the
minimum weight cut to form clusters [30]. Here, we use the
weighted graph and the edge weights are assigned a new
interpretation. By combining probability and graph theory,
the edge weight between two node is calculated. CLICK
assumes the similarity values within clusters and between
the clusters and follow Gaussian distributions. CLICK
recursively checks the current sub graph and generates a
kernel list, which consists ofthecomponentssatisfyingsome
criterion function. Using kernels as basic set of clusters and
CLICK carried out singleton clusters. These clusters contain
only one node and merge to generate resulting clusters.
Additional heuristics are provided to accelerate the
algorithm performance. CAST is another agglomerative
algorithm used probabilistic model for graph based theory
clustering algorithms [34]. Cast is the heuristics original
theoretical version and creates clusters sequentially and
each cluster begin with unassigned data point randomly.
SudiptoGuha et al proposed CURE algorithm [26]. CURE is a
novel hierarchical clustering Algorithm. In CURE, a constant
number c of scattered points in a clusterarefirstchosen. The
scattered points capture the shape and extent of the cluster
and these points are next shrunk towards the centroidofthe
cluster by a fraction.
[5] TengkeXiong et al proposed an DHCC (Divisive
Hierarchical Clustering of Categorical Data), it is a divisive
hierarchical algorithm for categorical data.Thismethoduses
MCA (Multiple Correspondence Analysis) is carried out by
performing a standard correspondence analysis (CA) on an
indicator matrix. DHCC starts with an all clusters containing
all the categorical objects, and repeatedly chooses one
cluster to split into two sub clusters. A binary tree is
employed to represent the hierarchical structure of the
clustering results. In DHCC, divides the cluster Cp involves
finding a sub optimal solution to the optimization problem
on the data set Cp with K =2. DHCC consists of two phases,
initial splitting phase and refinement phase. [29] R.Datta et
al and [36] M s Lew et al proposed one of the current
techniques for image retrieval is content-based image
retrieval. In content based image retrieval approach visual
features such as color feature, texture feature, shape feature
and local features are automatically extracted from the
image objects and organized as feature vectors. Then at
search phase, after selecting the query image by user,
retrieval engine retrieves the most similar images to the
query image by performing similarity comparison between
query feature vector and all the feature vectors in database.
Jorg Sander et.al proposed [33] Density-Based
Clustering for spatial databases.InwhichDBSCAN reliesona
density based notion of clusters and is designed to discover
clusters of arbitrary shape as well as to distinguish noise.
The generalized algorithm calledGDBSCAN canclusterpoint
objects as well as spatially extended objects according to
both their spatial and their non-spatial attributes. [35]
Markus M. Breunig et.al proposed LOF: Identifying Density-
Based Local Outliers. It introduce a new method for finding
outliers in a multidimensional data set. We introduce a local
outlier (LOF) for each object in the data set, indicating its
degree of outlierness. This is, to the best of our knowledge,
the first concept of an outlier which also quantifies how
outlying an object is. The outlier factor is local in the sense
that only a restricted neighbourhood of each object is taken
into account.
3. PROPOSED FRAMEWORK
In this experiment created an online shopping site. For that
used product images and its descriptions from the internet.
Download the images of each category such as Electronics,
Clothes, Home Appliance etc.
3.1 Problem Statement
Let S = (S1, S2-------Sn) be n transaction record and (i1,i2-----im)
be m items in S. Transaction record Si is an item set
represented as (x1, x2------xz) where xj € I for1≤j≤z.
Efficiently create a product clusters (c1, c2-----cm) from the
product I based on the transaction set S.
3.2 System Architecture
Figure 1: System Architecture for online shopping
activities
3.3 Experimental Details
In this experiment we formed clusters based on product
wise, season wise, and the combinationofseasonandgender
wise. We formed clusters from the transaction details of
customers. The main modules are User, Admin and Vendor.
In the new method we adopt k-means algorithm and build a
product tree where the leaf nodes usually denote the
provided products and the internal nodes are multiple

product categories. A product tree often consists of several
levels and thousands of nodes, and the number of nodes
increases rapidly from the top level to the bottom level. In
transaction data, each product (item) bought by a customer
corresponds to a leaf node in the product tree. To facilitate
the analysis and visualization of customer’s behaviour, we
build a “personalized product tree” for eachcustomer,called
purchase tree. The purchase tree can be built by aggregating
all products in a customer’s transactions, and pruning the
product tree by only retaining the corresponding leaf nodes
and all paths from root to leaf node.
Figure 2: Product Tree
The products are oftenorganizedaccordingtomultipletypes
of categories. In this paper, we propose the product tree to
systematically organize the products in a company.
Figure 3: Purchase Tree for the figure 2.
We often created a transactional databases for this
experiment. User module contains the registration, sign in,
purchase products as their own needs. And Suggestions to
the particular user based on the items purchased first time.
For Example, If a user purchased a mobile phone, then later
login he will be see the suggestions about different mobile
phones. Clusters are formed basedonthepurchasinghistory
details. Partitional clustering method is used for cluster
formation. Mainly used K-means method for cluster
formation. It formed clusters like product wise, season wise,
and the combination of season and gender wise. Admin
module contains some extra privilege including all privilege
of visitor and user. Add products, edit product information
and remove product. Ship order to user based on order
placed by sending confirmation mail. See the clusters and
analyze those clusters. He can understand which categoryof
product most sold. Vendor module containsregistrationand
login to the system. If he will add their products only his
registration accepted by administrator. Vendorscanalsosee
the clusters.
K-Means Algorithm
Input: S (instance set), K (number of cluster)
Output: clusters
1. Select a point randomly from the database
as initial centroids.
2. Choose clusters as k=3;
3. Compute similarity distance between center point
and each data points.
4. Form k clusters by assigning all data points to the
closest centroids.
3.4 Results and Analysis
Partition method is recommended for large data set and
hierarchical method is for small data set. Performance of
partition is better than hierarchical algorithms. Easy to
implement. Comparatively more efficient than hierarchical
method. Finally product wise, month wise, season wise and
region wise clusters were formed.
4. CONCLUSION
Customer segmentation is effectively implemented by using
K-means partitioning method. Segmentation implemented
like product wise, season wise, region wise, combination of
season wise and gender wise. Partition method is work well
large databases. Segmentation provides loyalty between
marketer and customers and provide customer relationship
management, improves business and get more profits.

REFERENCES
[1] Saroj, T. C., and Chaudhary, T., 2015. “Study on various
clustering techniques”. International Journal of Computer
Science and Information Technologies, 6(3),pp. 3031–3033.
[2] Tan, P.-N., Steinbach, M., and Kumar, V., 2013. “Data
mining cluster analysis: basic concepts and algorithms”.
Introduction to data mining.
[3] Gupta, P., 2011. “Robust clustering algorithms”. PhD
thesis, Georgia Institute of Technology.
[4] Gupta, P., 2011. “Robust clustering algorithms”. PhD
thesis, Georgia Institute of Technology.
[5] Xiong, T., Wang, S., Mayers, A., and Monga, E., 2012.
“Dhcc: Divisive hierarchical clustering of categorical data”.
Data Mining and Knowledge Discovery, 24(1), pp. 103–135.
[6] Berkhin, P., 2006. “A survey of clustering data mining
techniques”. In Grouping multidimensional data. Springer,
pp. 25–71
[7] Everitt, B., Landau, S., Leese, M., and Stahl, D.,
2001.“Cluster analysis. 4th”. Arnold, London.
[8] Hansen, P., and Jaumard, B., 1997. “Cluster analysis and
mathematical programming”. Mathematical programming,
79(1-3), pp. 191–215.
[9] Jain, A. K., and Dubes, R. C., 1988. “Algorithms for
clustering data”.
[10] Jain, A. K., Murty, M. N., and Flynn, P. J., 1999. “Data
clustering: a review”.ACMcomputingsurveys(CSUR),31(3),
pp. 264–323.
[11] Kolatch, E., et al., 2001. “Clustering algorithms for
spatial databases: A survey”. PDF is availableontheWeb, pp.
1–22.
[12] Wilks, D. S., 2011. “Cluster analysis”. In International
geophysics, Vol. 100. Elsevier, pp. 603–616.
[13] Grabmeier, J., and Rudolph, A., 2002. “Techniques of
cluster algorithms in data mining”. Data Mining and
knowledge discovery, 6(4), pp. 303–360.
[14] Migu´eis, V. L., Camanho, A. S., and e Cunha, 2012.
“Customer data mining for lifestyle segmentation”. Expert
Systems with Applications, 39(10), pp. 9359– 9366.
[15] Ng, R. T., and Han, J., 2002. “Clarans: A method for
clustering objects for spatial data mining”.IEEEtransactions
on knowledge and data engineering, 14(5), pp. 1003–1016.
[16] Huang, J. Z., Ng, M. K., Rong, H., and Li, Z.,
2005.“Automated variable weighting in k-means type
clustering”.IEEE Transactions on Pattern Analysis
andMachine Intelligence, 27(5), pp. 657–668.
[17] Tsai, C.-Y., and Chiu, C.-C., 2008. “Developing a feature
weight self-adjustment mechanism for a k-means clustering
algorithm”. Computational statistics & data analysis,52(10),
pp. 4658 4672.
[18] Chen, X., Xu, X., Huang, J. Z., and Ye, Y., 2013. “Tw-k-
means: automated two-level variable weighting clustering
algorithm formultiview data”. IEEE Transactions on
Knowledge and Data Engineering, 25(4), pp. 932–944.
[19] Rokach, L., and Maimon, O., 2005. “Clustering
methods”.In Data mining and knowledge discovery
handbook.Springer, pp. 321–352.
[20] Green, P. E., Kim, J., and Carmone, F. J., 1990.
“Apreliminary study of optimal variable weighting in k-
meansclustering”.Journal of Classification,7(2),pp.271–285.
[21] Xu, R., and Wunsch, D., 2005. “Survey of
clusteringalgorithms”. IEEE Transactions on neural
networks,16(3), pp. 645–678.
[22] Li, C., and Biswas, G., 2002. “Unsupervised learningwith
mixed numeric and nominal data”. IEEE Transactionson
Knowledge & Data Engineering(4), pp. 673–690
[23] Zhang, T., Ramakrishnan, R., and Livny,M.,1996.“Birch:
an efficient data clustering methodforverylargedatabases”.
In ACM Sigmod Record, Vol. 25, ACM, pp. 103–114.
[24] Karypis, G., Han, E.-H.,andKumar,V.,1999.“Chameleon:
Hierarchical clustering using dynamicmodeling”. Computer,
32(8), pp. 68–75.
[25] Karypis, G., and Kumar, V., 1998. “Multilevel k-way
partitioning scheme for irregular graphs”. Journal ofParallel
and Distributed computing, 48(1), pp. 96–129.
[26] Guha, S., Rastogi, R., and Shim, K., 1998. “Cure: an
efficient clustering algorithm for large databases”. In ACM
Sigmod Record, Vol. 27, ACM, pp. 73–84.
[27] Roux, M., 2015. “A comparative study of divisive
hierarchical clustering algorithms”. arXiv preprint
arXiv:1506.08977.
[28] Izadpanah, N., 2015. “A divisive hierarchical clustering-
based method for indexing image information”.arXiv
preprint arXiv:1503.03607.
[29] Datta, R., Joshi, D., Li, J., and Wang, J. Z., 2008. “Image
retrieval: Ideas, influences, and trends of the new age”. ACM
Computing Surveys (Csur), 40(2), p. 5.

[30] Lew, M. S., Sebe, N., Djeraba, C., and Jain, R., 2006.
“Content-based multimedia information retrieval: State of
the art and challenges”. ACM Transactions on Multimedia
Computing, Communications, and Applications (TOMM),
2(1), pp. 1–19.
[31] MacQueen, J., et al., 1967. “Some methods for
classification and analysis of multivariate observations”. In
Proceedings of the fifth Berkeley symposium on
mathematical statistics and probability, Vol. 1, Oakland, CA,
USA, pp. 281–297.
[32] Reddy, M. V., Vivekananda, M., and Satish, R. “Divisive
hierarchical clustering with k-means and agglomerative
hierarchical clustering”.
[33] Sander, J., Ester, M., Kriegel, H.-P., and Xu, X., 1998.
“Density-based clustering in spatial databases: The
algorithm gdbscan and its applications”. Data mining and
knowledge discovery, 2(2), pp. 169–194.
[34] Ben-Dor, A., and Yakhini, Z., 1999. “Clustering gene
expression patterns”. In Proceedings of the third annual
international conference on Computational
molecularbiology, ACM, pp. 33–42.
[35] Breunig, M. M., Kriegel, H.-P., Ng, R. T., and Sander, J.,
2000. “Lof: identifying density-based local outliers”. In ACM
sigmod record, Vol. 29, ACM, pp. 93–104.
[36] M. S. Lew, N. Sebe, C. Djeraba, R. Jain, Content-based
multimedia information retrieval: State of the art and
challenges, ACM Transactions on Multimedia Computing,
Communications, and Applications (TOMM) 2 (1) (2006)
1{19.
Author Profile
Neethu C M ,post graduation student in college of
engineering kidangoor. Specialization in computer and
information science. Received the graduation in computer
science and Engineering from Chennai University.
Anitha Abraham, She is working as an Asst.Professor at
college of engineering kidangoor. She receivedgraduationin
computer science and engineering from college of
engineering kidangoor and post graduation in
communication and network technology from MG
University. Her area ofinterestincludesCryptography,Image
Processing, Artificial Intelligence.

IRJET- Customer Segmentation from Massive Customer Transaction Data

More Related Content

What's hot (20)

Similar to IRJET- Customer Segmentation from Massive Customer Transaction Data (20)

More from IRJET Journal (20)

Recently uploaded (20)

IRJET- Customer Segmentation from Massive Customer Transaction Data