AN ENHANCED PRACTICAL PROGRAM ON DATA MINING

EDUCATION

Hiroyuki Morita, Masashi Kondo, Tomonori Ishigaki, Nagateru Araki and Yuji Nakayama

School of Economics, Osaka Prefecture University, 1-1 Gakuenmachi Nakaku Sakai, Osaka, Japan

Keywords: Data mining, Challenging education, Competition.

Abstract: To provide effective data mining education, practical data, high-level mining tools supported by computer,

and students’ interests are essential factors. To provide such education, we incorporated data analysis

competition into our data mining education. The competition highly motivates students; however, there was

no connection between their analysis and the practical usage. In our new educational program, we

incorporate a real sales promotion by the students themselves who propose a good analysis at a real shop.

By investigating some questionnaires for the students, we show that our program has a good evaluation.

1 INTRODUCTION

From the viewpoint of practical usage, data mining

education has been promoted since several years. To

implement the successful achievement of this

program, we need many factors than other existent

programs. In particular, in this course, a practical

data mining process supported by computer is highly

important. This is because experiments that analyze

extensive practical data are valuable for students to

develop a more comprehensive understanding.

However, it is not easy to generate practical data in a

real shop. To meet this necessity, from data of past

several years, we prepared a data of our University

Co-op shop, as real business data. By using this data,

we conducted a data analysis competition, which

provides common data to student participants and

they rival each other by their analysis.

The data that we use comprises about 1.5 million

records and several hundreds of items. Although the

shop is not big, its data is similar to those of an

outside real shop. A glance shot of the data is as

follows.

Figure 1 illustrates an overview of the point-of

sales (POS) data that we used. It has general data

attributes such as item code, purchasing time, and

receipt number. As an example of aggregation,

figures 2 and 3 illustrate the monthly sales and sales

by the hour for food and drinks, respectively.

Figure 1: An example of POS data.

Figures 2 and 3 illustrate that the sales are

unbalanced. In particular, around noon, we observe a

significant amount of sales.

In addition to the competition, we implement a

new trial which performs their analysis on the

University Co-op. It is an interesting and

challenging task.

As a related work, (Kay et al.. 2006) utilized data

mining method to extract effective communication

event patterns in student’s teamwork. They are

interesting, but we have a gap how to use data

mining to educational programs.

POSData(ScannerData)ofUniv Coop

Day/Time

NameofProduct

177

Morita H., Kondo M., Ishigaki T., Araki N. and Nakayama Y. (2010).

AN ENHANCED PRACTICAL PROGRAM ON DATA MINING EDUCATION.

In Proceedings of the 2nd International Conference on Computer Supported Education, pages 177-180

DOI: 10.5220/0002797201770180

 SciTePress

Figure 2: Monthly sales of food and drinks.

Figure 3: Food and drinks sales by the hour.

We explain our trial as follows. Section 2 describes

our aim and program, section 3 shows the practice in

this year, and section 4 draws a conclusion.

2 EDUCATION PROGRAM

2.1 The Aim and Distinctive Features

of this Education Program

By providing real-life business experience and an

opportunity to analyze business data, this academic

program aims to foster the development of

individuals who possess the abilities needed to

identify problems and come up with business

solutions by themselves. The distinctive feature of

this program is that with the cooperation of

University Co-op, actual on-campus business

premises are used as learning resources, and the

daily sales activities carried out in these premises,

along with the POS data obtained on a daily basis by

the Co-op, are used as teaching materials.

Figure 4: The aim of the education program.

Divided into teams of four or five, the

participating students carry out role-playing

exercises by acting as “distribution advisers” or

“business consultants”; this enables them to learn the

theory and application of data mining based on the

real-world experience. In addition to helping the

students learn the skills of teamwork, negotiation,

analysis and observation, and their importance there

of, these role-playing exercises enable them to

identify problems concerning real-life sales, which

would not be possible through lectures alone.

Furthermore, in this experiential learning process,

the students attempt various business solutions. In

addition, through their interaction with not only

faculty members but also real-life practitioners of

business analysis and postgraduate students working

in the corporate world, they learn how to present

their findings from new perspectives and how to put

forward practical business solutions.

2.2 Outline of the Education Program

Once the first- and second-year students have

completed their Liberal Arts Course and Specialized

Foundation Course or while they are taking them,

CSEDU 2010 - 2nd International Conference on Computer Supported Education

178

four classes in the second-year’s Seminar for Basic

Studies form the central locus of learning in this

program.

To enable second-year students to perform

problem-solving using large volumes of business

POS data, an advanced data mining system has been

introduced to ensure that liberal arts students as well

as those majoring in business subjects achieve

excellent results. The system consists of various

software-enabled data mining tools to be used

simultaneously on computers with a high processing

capacity.

The POS data analyzed consists of one to two

years’ worth of receipts data generated on a daily

basis in Osaka Prefecture University Co-op shops

and cafeterias. This includes POS data with ID,

which enables the tracking of individuals’

purchasing history after the data has been

anonymized.

Working in teams of four or five, students are

required to (1) adopt the viewpoint of those who

manage the business of University Co-op and (2)

find ways to use the Co-op’s data to improve its

business performance. For this, the students need to

observe how business is carried out at nearby stores

and eateries, perform various surveys, and interview

the members of the Co-op’s staff.

Presentation sessions are held in which students

report on their findings and have their analyses and

proposals evaluated. These sessions are attended not

only by the representatives of the Co-op but also by

specially invited business people with practical

experience of sales, purchasing, and data-handling in

various types of enterprises, thus bringing students

into contact with the real-life business world.

Challenge sessions verify the excellent team’s

proposal at the Co-op shop. Next, students try to

conduct sales at the Co-op shop as per their proposal

during one week. Students have to display the

commodities and carry out verification work by

order. Again, they identify the practice and their

proposal and have to report the results to the chief of

the Co-op shop and the members of this project.

3 APPLICATION OF DATA

ANALYSIS AT THE Co-oP

STORE

The announcement convention in which an all-star

team participates from a class (the current date is

September 24, 2009) is held. In this convention, the

report of an analysis proposal is presented between

the excellent (4 or 5) team of the class representative

of basic seminar C of second-year students and the

participating (3 or 4) team from special seminar,

which is conducted by the staff of this program.

Further, the content of the report is evaluated by

judges, including a judge outside the school, and

superiority or inferiority is decided.

3.1 Result Report Association

Excellence Announcement Outline

Report 1:

This team conducted sales promotion that gave a

10% discount of drinking products to customers who

bought dairy desserts such as yogurt and pudding.

Figure 5 shows that almost all the products of both

the categories were sold out. Thus, this sales

promotion seemed to work well, although a detailed

analysis that would compare the figure with that in

another period without the sales promotion must be

conducted.

Dairy dessert

in-out

Mon. Teu. Wed.

Thur.

Fri.

Cream puff (Monteur)

line-up

8 8

buy-out

8 8

Pudding (Glico)

line-up

5 5

buy-out

2 3 5

Caramel pudding

(Ropia)

line-up

buy-out

1 4 1

Chocolate crape

(1:Ropia)

line-up

buy-out

7 1

Chocolate crape

(2:Monteur)

line-up

buy-out

Yogurt (Meiji)

line-up

buy-out

1 1 1 2 1

Figure 5: Sales of dairy desserts.

Report 2:

This team analyzed the opportunity loss of sales.

They showed a prediction using the volume of sales

as a function of a day of the week and temperature.

In this analysis, the paper pack drink is applicable to

the analysis. They chose the goods of the annual

sales best 10, and expected the amount of

opportunity losses. The goods that have a slightly

less order quantity according to their analysis and

should increase the order quantity were ascertained.

3.2 Approach of Challenge

Introduction in Co-op Bread Shop

Site

The experience of the verification of the proposal

AN ENHANCED PRACTICAL PROGRAM ON DATA MINING EDUCATION

179

on the site in the bread shop by a result report

association excellent team and the sales site was

given by cooperation in the Co-op sales in the

current year (bread shop challenge). The testing

period was two weeks from October 19, 2009, to

October 30, 2009. The team of report 2 took charge

in the first week (20:30 was assumed to be business

hours from 8:30 on weekdays) and the team of

report 1 took charge of the second week.

The main work is as follows.

1: Order work (done two days before the sales day)

2: Carrying work (exhibition of confirmation and

commodity of order goods)

3: Verification work (set sales goods decision, sale

notice of time, and POP substitution)

4: Abandonment commodity (The one to be

abandoned with daily goods that remained

unsold on that day is selected).

3.3 Comment on a Challenge

Students in both the teams said that the period of one

week, during which they managed some part of the

Co-op store, was too short. They had to order

products that would be sold out within one week.

They also said that if they had had another week,

they could have ordered a variety of products, which

would have satisfied customers’ demands. We must

take these opinions into account in order to make our

educational project in future more valuable and

enjoyable for students.

4 CONCLUSIONS

In this program, students are requested to obtain a

useful finding from the extensive POS data collected

for a long-term period of one year or more by the

group work. Students should (1) understand the

features of the commodity, customer’s purchasing

pattern, and features of the store; (2) analyze data

through trial and error while combining several

analysis tools; and (3) solve this problem in the

limited class time. Here, it is understood that

information technology plays a prominent role. That

is, the data mining software with high speed of

computational speed and GUI known by intuition,

an excellent display, and the presentation

environment are needed in addition to a naturally

necessary data mining and statistical model analysis.

It is thought that the education effect that these

functions are the following is brought;

1. It is necessary to verify various hypotheses to

obtain a significant result from a large amount of

capricious POS data. Therefore, it is necessary to

analyze the data repeatedly. For this, the system

with strong calculation ability is useful.

2. It is effective in obtaining the analysis result in a

short time, correcting the hypothesis, making the

model easy, and sustaining students’ interest and

concentration. As a result, the possibility of

reaching a satisfactory result increases.

3. It is necessary to allot the analysis business for

the findings by team work. The easiness of the

operation by GUI software is lost in the

difference of the capacity for the analysis of the

students in the team and contributes to the

decrease of time loss as a team.

4. The computer network to share an individual

analysis result mutually makes the group work

extremely efficient.

5. A big display and the presentation device are

effective in bringing the result together and

obtaining a final finding.

Our system meets a necessary requirement for

executing this program.

There has been much discussion on the POS data

with respect to the reality of the business process. To

solve this and to obtain an effective finding, our

educational system has adequate power. However,

we think that we can use only a part of this power.

We wish to draw out the power kept secret by

teacher and student’s collaborations.

ACKNOWLEDGEMENTS

This program is financially supported by the

program for Promoting High-Quality University

Education of the Ministry of Education, Culture,

Sports, Science, and Technology in Japan.

REFERENCES

Kay,J., Maisonneuve, N., Yacef, K., Zaiane, O., Mining

Patterns of Events in Students' Teamwork Data,

Proceedings of Educational Data Mining Workshop,

held in conjunction with Intelligent Tutoring Systems

(ITS), Taiwan, June 26, 2006.

CSEDU 2010 - 2nd International Conference on Computer Supported Education

180