What is data mining ?

The Datamining Garden kick-off workshop
June, 19th 2007
Regus Pegasus, Diegem

What is data mining ?

Johan Blomme
Circulation Manager, AMP

1

1. Introduction : “Competing on Analytics”

2

• Thomas Davenport : organizations that have built their very business on
the ability to collect, analyze and act on data are consistently the leaders in
their industry.

• The demands of business today are creating an increasing need for access to data
and the use of it to maintain a sustainable competitive advantage :

– the rapid construction of data-driven analytics :
• descriptive statistics ;
• predictive modeling and optimization techniques ;

– the rapid deployment of knowledge derived from data ;

– the need to give end users access to results in a form that helps them gain the insights
they need to make critical business decisions.

3

Industrial Age Information Age

interwoven, collaborative
Processes:
linear, sequential

continuous, rapid
Tempo:
periodic, slow

Assets : intangibles
tangibles

4

2. Business drivers of data mining

6

Time and information drive the information age, and competitiveness will be
based on obtaining real-time information and acting on it promptly and effectively.
The following changes indicate how to compete in the information age :

• more complex business environments due to globalization and
deregulation ;
• greater impact of change from external causes ;
• a power shift from sellers to buyers, rapidly shifting customer
demands and subsequent reduced product life cycles ;
• constant technology change ;
• faster business cycles and temporary competitive advantage ;
• the need to explore collaborative strategies ;
• constant change at ever-increasing speeds and shrinking
strategy time horizons.

7

• Technology facilitates data gathering :

– e.g. RFID ;

– currently : applications mainly in production environment and logistics ;

– future possibilities : narrowcasting ;

– privacy issues !

8

• Technology transforms the way we live and interact :

– ubiquitous access to information is changing the economics of knowledge ;

– consumer preferences are becoming more complex and are changing more rapidly

– customers will increasingly choose how they would like to interact with organizations and will do only
business with componies that meet their interaction needs ;

– the customer takes the lead ;

– technology changes the behaviour of consumers ; consequently, it is very important to track customer
interactions and customer behaviour

9

3. Data mining defined

10

• Data mining is the extraction of actionable knowledge from large datasets to acquire
and sustain a competitive advantage.

• Data mining is about achieving the organization’s goals, not about the maths and the statistics.

11

• The introduction of data warehousing in the 90’s resulted in a wider acceptance of
data mining :

– operational data stored in corporate data warehouses has the potential to be exploited as
business intelligence ;
– data warehouses are multidimensional structures used for on line analytical processing ;
– OLAP :
• analyze information about past performance on an aggregate level
• verification-based approach : the user develops a hypothesis and then tests the data to prove or
disprove the hypothesis
– data mining :
• prospective data analysis
• predicting future trends, allowing businesses to make proactive, knowledge driven decisions

Data mining and statistics/OLAP can complement each other : the inductively revealed
relationships between variables can be used to formulate hypothesis and the insights gained

12

• Statistics vs. data mining :

– Statistical analysis is primarily concerned with confirmatory data analysis (model fitting) :
testing if a proposed model of hypothetical relationships between variables does or does not
provide a good explanation of the observed data.

 Statistical models are based on assumptions or some theory about relationships between
variables and assume a deductive process

– Data mining : rather than verifying hypothetical patterns, data mining uses the data itself to
detect such patterns.

 Data mining : computational algorithms play a much greater role in building model through
exploratory data analysis (EDA). The nature of the process is inductive.

14

optimization

business value
predictive modeling

forecasting

alerts

query / drill down

standard reports

degree of intelligence

16

The CRISP-DM model is an industry- and application-neutral standard for fitting
data mining into the general problem-solving strategy of a business.

17

4. An example of DM
The case of demand planning of magazines (AMP)

18

Distribution of press products :
2.8 mio copies every night

19

Business problem :

The market for printed magazines is declining. Key reasons :
- advertising is migrating to e-media ;
- publishers are not investing in the future of printed magazines at the same rate as they are in
in the future of e-media products ;
- the young generation is brought up in an e-media world and will be less inclined to read
printed products ;
- publishers’ drive to reduce costs makes e-media publishing an attractive proposition, since
paper, printing and distribution costs can be eliminated.

The big issue in single copy sales is that of unsolds. If sales volumes go down, the distribution cost/copy
increases, since the overhead of the distribution system have to be spread over fewer magazines, and
returns as a proportion of delivered magazines increases (the fee earned by distributors is based on cover
prices of magazines and number of copies sold (instead of a cost-to-serve model).

20

Objective :

How to build an intelligent supply chain to improve supply chain efficiency,
reduce costs and increase profits ?

21

Internet, WWW, Sales Force
Retail Catalog - Mail Kiosks

SAP
BUSINESS
WAREHOUSE
Product Planning Suppliers
& Development

Business
Understanding

• make-to stock environment
• lack of visibility of supply chain, esp. day-to-day demand and stock positions
• excessive inventory levels
• return rates of + 60 % are not uncommon in our industry

=> Information is key : integrate internal SC activities of AMP withthose of paterners to gain efficiencies across the supply chain

22

the traditional (linear) supply chain

23

the intelligent supply chain

Publisher Distributor Newsstand

• POS Data Sharing
Product Flow
• Inventory levels
• Forecasts
• Promotional Activities
Information Flow • New Product Introduction
• Production & delivery schedules

Information & Intelligence Sharing for Effectiveness
1

24


SAP
BUSINESS
WAREHOUSE
& Development

Business Data
Preprocessing
Understanding

. data normalization
. handling missing data

25


SAP
BUSINESS
WAREHOUSE
& Development

Business Data Develop
Understanding Preprocessing Forecast Model

. flat sales model
. intermittent data modeling
. discreta data : low volume model
. apply business rules

26


SAP
BUSINESS
WAREHOUSE
& Development

Business Data Develop Deploy
Understanding Preprocessing Forecast Model Forecasts

. interpret results : simulation
. workflow integration (operations)

27

service degree level 

monthly
titles

28

reference period Linear.(reference period) draw regulation Log.(draw regulation)

100,0000

75,0000
% weighted oos

50,0000
R² = 0,0213

25,0000

R² = 0,5696

0
0 25,0000 50,0000 75,0000 100,0000
% unsolds

29

Shared visibility across supply chain

Improved understanding, forecasting and analysis of consumer demand
Improved capability to respond and react to changes
Improved stability, predictability and efficiency of supply chain operations

Improved Fill Rates Reduced lead times Smoother SC execution
Improved on-shelf availability Reduced inventories More efficient processes
More effective demand generation Reduction of costs for handling
activities returns

Increased Reduced Reduced
Sales Inventories Costs

30

What is data mining ?

Recommended

More Related Content

What's hot (20)

Similar to What is data mining ? (20)

More from Johan Blomme (16)

Recently uploaded (20)

What is data mining ?