International Journal of Engineering Trends and Technology (IJETT) – Special Issue – April 2017

Case Studies On Data Mining In Market

Analysis
Shravan Kumar Manthri 1 and Hari Priyanka Chilakalapudi 2
Department of Computer Science and Engineering
1
Bhoj Reddy Engineering College for Women,
Hyderabad-500059, Telangana,India
Department of Computer Science and Engineering
2
Bhoj Reddy Engineering College for Women,
Hyderabad-500059, Telangana,India
Abstract- A huge chunk of data is generated each minute track who is buying what, when they are
in enterprise business. Extracting information from piles buying it and at what price. The stores can
of data helps in extracting patterns that can predict and
guide future behaviour of the enterprise. Data mining then use this data, after analyzing it, for
techniques filter through large amounts of raw data and multiple purposes, such as offering customers
extract useful information that gives enterprise coupons targeted to their buying habits and
businesses a competitive edge in the market. Various deciding when to put items on sale or when to
cases on customer purchasing habits have been
presented and also used in real problems. Data mining
sell them at full price. Data mining can be a
techniques are highly effective in analysing consumer cause for concern when only selected
behaviours. It helps enterprises to make informed information, which is not representative of the
business decisions, enhances business intelligence, overall sample group, is used to prove a
thereby improving the company’s revenue, detecting certain hypothesis.
anomalies, fraud detection and reducing cost overheads.
In this paper the author reviews various such techniques.
Further, the application of these techniques in various Power of hidden information in data:
scenarios is analyzed. Enterprises generate terabytes of data each day
that is stored in databases, data warehouses, or
I. Introduction various other kinds of data repositories. Most
of the valuable information may be hiding
Data mining behind such data; the overwhelming data
Data mining is a process used by companies to volume makes it difficult for human beings to
turn raw data into useful information. By using extract them without the help of powerful tools
software to look for patterns in large batches and techniques. At the beginning of the last
decade information was only available on
of data, businesses can learn more about their
papers and only at a specific time. In this age,
customers and develop more effective information is now easily accessible, as
marketing strategies as well as increase sales content providers, content locators and
and decrease costs. Data mining depends on powerful search engines have enabled access
effective data collection and warehousing as to huge amounts of information in no time.
well as computer processing. To study a Natural Language Processing has allowed
customer psychological mindset and access to the same information in a language
familiar to the person. From the time when
converting this into statistical format and see people used to wait in huge queues to pay
that if there is any technical format by which bills, taxes or for movie tickets and
we can analyze his buying behaviour. entertainment, we have reached a time where
everything happens in a few clicks. All these
BREAKING DOWN 'Data Mining in changing factors and trends made a huge
impact on the enterprises which are highly
marketing' dependent on the consumer behaviour and
basket trends. Every enterprise needs to be
Grocery stores are well-known users of data
mining techniques. Many supermarkets offer highly scalable and be able to predict the
free loyalty cards to customers that give them future behaviour of the customer to make
access to reduced prices not available to non- profits in their business. This very needs lead
members. The cards make it easy for stores to to the Market Analysis. Earlier techniques like
questioners, surveys, test marketing, previous

ISSN: 2231-5381 http://www.ijettjournal.org Page 180

 International Journal of Engineering Trends and Technology (IJETT) – Special Issue – April 2017

sales data and leading indicators were used to those in other groups (clusters).
predict the future of a product. All these eg: Finding customer segments in a company
techniques although proved to be effective, based on transaction, web and customer call
were highly time consuming and required a lot data.
of manpower. Data mining techniques have
completely changed the scenario. Information Association analysis
which was once gathered by travelling miles is Association is a data mining function that
now available in few seconds. discovers the probability of the co-occurrence
II. Data Mining Tasks of items in a collection. The relationships
between co-occurring items are expressed as
Classification:
association rules.
Classification is the process of finding a model Eg: Find cross selling opportunities for a
that describes the data classes or concepts. The retailer based on transaction purchase history.
purpose is to be able to use this model to
predict the class of objects whose class label is Exploring and Expanding Business
unknown. This derived model is based on the Data mining is defined as a business process
analysis of sets of training data. for exploring large amounts of data to discover
Eg: 1.Assigning voters into known buckets by meaningful patterns and rules [4]. Companies
political parties . can apply data mining in order to improve
2.Bucketing new customers into one of their business and gain advantages over the
known customer groups. competitors. The most important business
areas that successfully apply data mining,
Regression: presented in below figure.
In statistical modeling, regression analysis is a
statistical process for estimating the
relationships among variables. It includes
many techniques for modeling and analyzing
several variables, when the focus is on the
relationship between a dependent variable and
one or more independent variables.
Eg:Predicting unemployment rate for next
year. Estimating insurance premium.

Anomaly detection:
Anomaly detection (also outlier detection) is
the identification of items, events or
observations which do not conform to an
expected pattern or other items in a dataset.
Eg:Fraud transaction detection in credit cards.

Time series
A time series is a series of data points indexed Digital marketing is a method of marketing
(or listed or graphed) in time order. Most from the viewpoint of customers. Digital
commonly, a time series is a sequence taken at information is not only more easily integrated,
successive equally spaced points in time. Thus sorted, and spread, but also enables providers
it is a sequence of discrete-time data. and consumers to interact more quickly. In the
Eg:Sales forecasting, production forecasting, past, it usually took a long time for marketing
virtually any growth phenomenon that needs to analyze and achieve effectiveness. Today,
to be extrapolated. digital marketing enables marketing promotion
to have a higher synergistic effect.
Clustering With the rapid changes in the business
clustering is the task of grouping a set of environment, technology progress, and digital
objects in such a way that objects in the same transmission, the marketing of businesses
group (called a cluster) are more similar (in should be changed rapidly. Similarly, the
some sense or another) to each other than to

ISSN: 2231-5381 http://www.ijettjournal.org Page 181

 International Journal of Engineering Trends and Technology (IJETT) – Special Issue – April 2017

strategy of the market should be changed from some scenarios that have implemented the
the Red Sea Strategy to the Blue Ocean above methods.
Strategy. With the environment changing, not Case study 1 : Application of Association Rule
only market space continues to expand, but mining in Recommender systems
also the market environment becomes more Recommender systems are hugely popular
tightened. From traditional store sales, these days in variety of fields. To name a few
telephone marketing, and face-to-face Movies, music, books, research articles, search
marketing, to the development of Internet mar- queries, social tags, etc. These systems help
keting, such as sale and purchase through the the enterprise by combining the ideas of
Website, keyword marketing, blog marketing, intelligent systems, machine learning,
and so on, and with wireless network information retrieval to predict the customer
development, the ubiquitous Internet will be behaviour. There are two approaches in
bringing the world . It makes the development recommender systems, one is collaborative
of digital marketing communications seem filtering and content-based filtering.
increasingly important. The business is no Collaborative filtering methods collect and
longer limited by traditional ways such as time analyze a large amount of information on
and space, but increases the opportunities of users’ behaviors, activities or preferences and
contact and interactions with customers. predict what users will like based on their
similarity to other users. One of the
Apriori Algorithm approaches is to use the Apriori algorithm.
In 1994, the Apriori algorithm was proposed In this case study Apriori algorithm is used to
by Agrawal and Srikant [3]. It is a classic extract association rules from user profiles. As
algorithm for learning association rules. an example PVT system is used. PVT system
Apriori is designed to operate on databases is a recommender program that suggests TV
containing transactions (for example, channels to users based on their viewing
collections of items bought by customers, or habits. This system maintains both positively
details of commerce website frequentation). and negatively rated TV channels. Treating
As is common in association rule mining, user profiles as transactions and the program
given a set of itemsets (for instance, sets of ratings therein as itemsets, the Apriori
retail transactions, each listing individual algorithm can be used to derive a set of rules
items purchased), the algorithm attempts to and associated confidence levels between
find subsets which are common to at least a programs. The confidence values are taken as
minimum number of the item sets. Apriori similarity scores and used to fill in a program
uses a “bottom up” approach, where frequent similarity matrix. The procedure goes as
subsets are extended one item at a time (a step follows, the relationship between programs is
known as candidate generation), and groups of identified beyond a simple overlap. Like for
candidates are tested against the data. The example a person who watches Reality shows
algorithm terminates when no further like Rodies and Big Boss may not be
successful extensions are found. interested in shows like KBC and Indian Idol.
The purpose of the Apriori Algorithm is to But if a relation between Rodies and Indian
find associations between different sets of Idol can be established then it can provide a
data. It is sometimes referred to as “Market basis for pattern matching. The relation can be
Basket Analysis”. There are a number identified by finding the support and
of items in each set, and is called a transaction. confidence values. In this case study, the
The output of Apriori is sets of rules that tell confidence values are taken as the similarity
us how often items are contained in sets of scores and recommended to the user. Using
data. direct program similarity we can derive rules
and further by chaining these rules together we
III. Predicting the customer behaviour can get new results.

Predicting the customer behaviour is the most Case Study 2: Classification model for Target
important activity in enterprise business.All selection in direct marketing
the previously described methods give
enterprises huge amounts of useful Using historical purchase data, a predictive
information. the following section presents response model with data mining techniques

ISSN: 2231-5381 http://www.ijettjournal.org Page 182

 International Journal of Engineering Trends and Technology (IJETT) – Special Issue – April 2017

was developed to predict a probability that a Since every business produces many products,
customer in Ebedi Microfinance bank standing out in a competitive market is a great
(Nigeria) will respond to a promotion or an issue to be solved for businesses. The
offer.[2] To achieve this purpose, a predictive advantage of operating businesses in addition
response model using customers’ historical to their innovative products, services, brands,
purchase data was built with data mining quality, etc., how to combine marketing
techniques. The data were stored in a data techniques with products, make everyone
warehouse to serve as management decision know about the product, and increase
support system. The response model was built customer's purchase intention and interests is
from customers’ historic purchases and one of the key success factors of operating a
demographic dataset. The purchase behaviour business.Overall, the digital marketing will
variables used in the model development are suggest a proper communication method with
as follow. Recency: This is the number of consumers based on analyzing the marketing
months since the last purchase and first information of the product, history records,
purchase. It is typically the most powerful of and purchase behavior of consumers. Among
the three characteristics for predicting them, the product information includes the
response to a subsequent offer. This seems type, price, place, and promotion of the
quite logical. It says that if you have recently product. History records contain the previous
purchased something from a company, you are marketing strategies, practices and market
more likely to make another purchase than reactions to estimate, consult, and explore the
someone who did not recently make a potential or unknown influence factors of
purchase. Frequency: This is the number of marketing. In terms of consumer behavior,
purchases. It can be the total of purchases feedbacks can be received and suggesting
within a specific time frame or include all them for the products of interest. Different
purchases. This characteristic is second to characteristics of consumer behavior, the
recency in predictive power for response. marketing strategy should be ingratiated with
Again, it is quite intuitive as to why it relates the different of consumer’s age, sex,
to future purchases. Monetary value: This is occupation, income, lifestyle. In other words,
the total amount. Similar to frequency, it can product information, history record, and
be within a specific time frame or include all consumer purchase behavior in terms of
purchases. Of the three, this characteristic is products association will affect the business to
the least powerful when it comes to predicting formulate the marketing strategy for different
response. But when used in combination, it products.
can add another dimension of understanding.
Demographic information includes customers’
personal characteristics and information such REFERENCES
as age, sex, address, profession etc Bayesian
algorithm precisely Naïve Bayes algorithm [1] Barry Smyth, Kevin McCarthy, James
was employed in constructing the classifier Reilly, Derry O’Sullivan and Lorraine
system. Both filter and wrapper feature McGinty, “Case-Studies in Association Rule
selection techniques were employed in Mining for Recommender Systems”. Science
determining inputs to the model. The results Foundation Ireland (SFI) under Grant No.
obtained shows that Ebedi Microfinance bank 03/IN.3/I361
can plan effective marketing of their products
and services by obtaining a guiding report on [2] Eniafe Festus Ayetiran; “A Data Mining-
the status of their customers which will go a Based Response Model for Target Selection in
long way in assisting management in saving Direct Marketing”, I.J. Information
significant amount of money that could have Technology and Computer Science, 2012, 1,
been spent on wasteful promotional 9-18
campaigns.
[3] Shu-Ching Wang,Shun-Sheng Wang
IV. Conclusions Chih-Ming Chang,”Systematic Approach for
Digital Marketing Strategy through Data
In today’s business world, grabbing a Mining Technology”.Department of Informa
customer attention initiates an important role. tion Management, Chaoyang University of

ISSN: 2231-5381 http://www.ijettjournal.org Page 183

 International Journal of Engineering Trends and Technology (IJETT) – Special Issue – April 2017

Technology Taichung 409, Taiwan,2014,4-1. Economic Studies, Bucharest, Romania.

Database Systems Journal vol.IV, no.4/2013
[4] Ruxandra PETRE,” Data Mining Solutions
for the Business Environment”. University of

ISSN: 2231-5381 http://www.ijettjournal.org Page 184