FORM NO.

- F/ TL / 021
Rev.00 Date 20.03.2020

RECORD NOTEBOOK

DATA MINING LAB–(BCS18L11)

2023-2024(ODD SEMESTER)

DEPARTMENT

Of

COMPUTER SCIENCE AND ENGINEERING

NAME : M.SUDHARSHAN

REGISTERNO : 201061101117

COURSE : B.TECH-CSE

YEAR/SEM/SEC : IV/VII/B
 FORM NO. - F/ TL / 021
Rev.00 Date 20.03.2020

BONAFIDE CERTIFICATE

REGISTER NO : 201061101117
NAME OF THE LAB : DATA MINING LAB(BCS18L11)
DEPARTMENT : COMPUTER SCIENCE AND ENGINEERING

Certified that this is the bonafide record of work done by M.SUDHARSHAN

of IV Year B.Tech(CSE), Sec-‘B’ in the DATA MINING LAB(BCS18L11)
during the year 2023-2024.

Signature of Lab-in-Charge Signature of Head of Dept

Submitted for the Practical Examination held on -------------------------

Internal Examiner External Examiner

 INDEX

Staff
S.No List Of Experiments Page No Signature
Date

INTRODUCTION TO THE WEKA MACHINE

1 13/06/23 1
LEARNING TOOLKIT

2 INTRODUCTION TO EXPLORATORY DATA

20/06/23 ANALYSIS USING RATTLE AN OPEN 08
SOURCE TOOL(R)

INTRODUCTION TO REGRESSION USING

3 27/06/23 13
RATTLE AN OPEN SOURCE TOOL. ( R)

CLASSIFICATION USING THE WEKA

4 04/07/23 17
TOOLKIT – PART 1

CLASSIFICATION USING THE WEKA

5 11/07/23 21
TOOLKIT - PART-2

PERFORMING DATA PREPROCESSING FOR

6 18/07/23 24
DATA MINING IN WEKA

7 25/07/23 PERFORMING CLUSTERING IN WEKA 27

8 01/08/23 ASSOCIATION RULE ANALYSIS IN WEKA 31

DATA MINING CASE STUDY USING THE

9 08/08/23 33
CRISP-DM STANDARD

DATA MINING CASE STUDY USING THE

10 22/08/23 45
CRISP- DM STANDARD
 EXP NO. 1 DATE: 13/06/23

INTRODUCTION TO THE WEKA MACHINE

LEARNING TOOLKIT

AIM:

To study WEKA Tool Kit

WEKA is opensource java code created by researchers at the University of Waikato in New Zealand.

It provides many different machine learning algorithms, including the following classifiers:

● Decision tree (j4.8, an extension of C4.5)

● MLP, aka multiple layer perceptron (a type of neural net)
● Naïve bayes
● Rule induction algorithms such as JRip
● Support vector machine
● And many more…

The GUI WEKA

The Weka GUI Chooser (class weka.gui.GUIChooser) provides a starting point for launching
Weka’s main GUI applications and supporting tools. If one prefers a MDI (“multiple document
interface”) appearance, then this is provided by an alternative launcher called “Main” (class
weka.gui.Main).The GUI Chooser consists of four buttons—one for each of the four major Weka
applications—and four menus.

The buttons can be used to start the following applications:

• Explorer : An environment for exploring data with WEKA (the rest of this documentation deals
with this application in more detail).
• Experimenter: An environment for performing experiments and conducting statistical tests
between learning schemes.
• KnowledgeFlow: This environment supports essentially the same functions as the Explorer but
with a drag-and-drop interface. One advantage is that it supports incremental learning.

M.SUDHARSHAN 1 201061101117
 • SimpleCLI: Provides a simple command-line interface that allows direct execution of
WEKA commands for operating systems that do not provide their own command line
interface.

The menu consists of four sections:

WEKA Explorer The

user interface

Section Tabs

At the very top of the window, just below the title bar, is a row of tabs. When the Explorer is first
started only the first tab is active; the others are grayed out. This is because it is necessaryto
open (and potentially pre-process) a data set before starting to explore the data.

The tabs are as follows:

1. Preprocess. Choose and modify the data being acted on.

2. Classify. Train and test learning schemes that classify or perform regression.
3. Cluster. Learn clusters for the data.
4. Associate. Learn association rules for the data.
5. Select attributes. Select the most relevant attributes in the data.
6. Visualize. View an interactive 2D plot of the data.

Once the tabs are active, clicking on them flicks between different screens, on which the respective
actions can be performed. The bottom area of the window (including the status box, thelog button,
and the Weka bird) stays visible regardless of which section you are in. The Explorer can be easily
extended with custom tabs.

1. Preprocessing

M.SUDHARSHAN 2 201061101117
 OPTIONS

1. All. All boxes are ticked.

2. None. All boxes are cleared (unticked).
3. Invert. Boxes that are ticked become unticked and vice versa.
4. Pattern. Enables the user to select attributes based on a Perl 5 Regular Expression. E.g., .* id
selects all attributes which name ends with id. Once the desired attributes have been selected, they
can be removed by clicking the Remove button below the list of attributes. Note that this can be
undone by clicking the Undo button, which is located next to the Edit button in the top- right
corner of the Preprocess panel.

Loading Data

The first four buttons at the top of the preprocess section enable you to load data into WEKA:

1. Open file. Brings up a dialog box allowing you to browse for the data file on the local file system.
2. Open URL. Asks for a Uniform Resource Locator address for where the data is stored.
3. Open DB. Reads data from a database. (Note that to make this work you might have to edit the
file in weka/experiment/DatabaseUtils.props.)
4. Generate. Enables you to generate artificial data from a variety of Data Generators. Using the
Open file. button you can read files in a variety of formats: WEKA’s ARFF format, CSV format,
C4.5 format, or serialized Instances format. ARFF files typically have a .arff extension, CSV files
a .csv extension,C4.5 files a .data and .names extension and serialized Instances objects a .bsi
extension.

Working with Filters

The preprocess section allows filters to be defined that transform the data in various ways. The Filter
box is used to set up the filters that are required. At the left of the Filter box is a Choose button. By
clicking this button it is possible to select one of the filters in WEKA. Once a filter has been selected,
its name and options are shown in the field next to the Choose button. Clicking on this box with the
left mouse button brings up a GenericObjectEditor dialog box. A click with the right mouse button
(or Alt+Shift+left click) brings up a menu where you can choose, either to

M.SUDHARSHAN 3 201061101117
 display the properties in a GenericObjectEditor dialog box, or to copy the current setup string to the
clipboard.

2. Classification

Selecting a Classifier

At the top of the classify section is the Classifier box. This box has a text field that gives the name
of the currently selected classifier, and its options. Clicking on the text box with the left mouse
button brings up a GenericObjectEditor dialog box, just the same as for filters that you can use to
configure the options of the current classifier. With a right click (or Alt+Shift+left click) you can
once again copy the setup string to the clipboard or display the properties in a GenericObjectEditor
dialog box. The Choose button allows you to choose one of the classifiers that are available in
WEKA.
3. Clustering

Selecting a Clusterer

By now you will be familiar with the process of selecting and configuring objects. Clicking on the
clustering scheme listed in the Clusterer box at the top of the window brings up a
GenericObjectEditor dialog with which to choose a new clustering scheme.

M.SUDHARSHAN 4 201061101117
 Cluster Modes

The Cluster mode box is used to choose what to cluster and how to evaluate the results. The first
three options are the same as for classification: Use training set, Supplied test set and Percentage
split except that now the data is assigned to clusters instead of trying to predict a specific class. The
fourth mode, Classes to clusters evaluation, compares how well the chosen clusters match up with a
pre-assigned class in the data. The drop-down box below this option selects the class, just asin the
Classify panel. An additional option in the Cluster mode box, the Store clusters for visualization tick
box, determines whether or not it will be possible to visualize the clusters once training is complete.
When dealing with datasets that are so large that memory becomes a problem it may be helpful to
disable this option.
4. Associating
Setting Up
This panel contains schemes for learning association rules, and the learners are chosen and
configured in the same way as the clusterers, filters, and classifiers in the other panels.

Learning Associations

Once appropriate parameters for the association rule learner have been set, click the Start button.
When complete, right-clicking on an entry in the result list allows the results to be viewed or
saved.

Selecting Attributes
Searching and Evaluating

Attribute selection involves searching through all possible combinations of attributes in the data
to find which subset of attributes works best for prediction. To do this, two objects must be set up:

M.SUDHARSHAN 5 201061101117
 an attribute evaluator and a search method. The evaluator determines what method is used to assign
a worth to each subset of attributes. The search method determines what style of search is
performed.

Options

The Attribute Selection Mode box has two options:

1. Use full training set. The worth of the attribute subset is determined using the full set of
training data.
2. Cross-validation. The worth of the attribute subset is determined by a process of cross-
validation. The Fold and Seed fields set the number of folds to use and the random seed used when
shuffling the data. As with Classify , there is a drop-down box that can be used to specify which
attribute to treat as the class.

M.SUDHARSHAN 6 201061101117
 RESULT:
The Weka tool kit is studied and output is verified.

M.SUDHARSHAN 7 201061101117
 EXP NO. 2 DATE: 20/06/23

INTRODUCTION TO EXPLORATORY DATA ANALYSIS USING

RATTLE AN OPEN SOURCE TOOL(R)
AIM :

To Conduct Exploratory Data Analysis Using Rattle

(a) SUMMARY

The Summary check box provides numerous measures for each variable, including, in the first
instance, the minimum, maximum, median, mean, and the first and third quartiles. Generally,
if the mean and median are significantly different then we would think that there are some
entities with very large values in the data pulling the mean in one direction. It does not seem to
be the case for Age but is for Income.

M.SUDHARSHAN 8 201061101117
 (b) DESCRIBE

The Describe check

box

(c) BASICS

The Basics check box

M.SUDHARSHAN 9 201061101117
 (d) KURTOSIS

The kurtosis is a measure of the nature of the peaks in the distribution of the data. A larger value for the
kurtosis will indicate that the distribution has a sharper peak, as we can see in comparing the
distributions of Income and Adjustment. A lower kurtosis indicates a smoother peak.

M.SUDHARSHAN 10 201061101117
 (e) SKEWNESS

The skewness is a measure of how asymmetrical our data is distributed. A positive skew indicates
that the tail to the right is longer, and a negative skew that the tail to the left is longer.

(f) MISSING

Missing values present challenges to data mining. The Show Missing check button of the Summary option
of the Explore tab provides a summary of missing values in our dataset. Following figure illustrates the
missing value summary. Such information is useful in understanding structure in the missing values.

M.SUDHARSHAN 11 201061101117
 Figure : Missing value summary for a version of the audit modified to include
missing values.

The missing value summary table is presented with the variables listed along the top. Each row
corresponds to a pattern of missing values. A 1 indicates a value is present, whereas a 0 indicates a value
is missing.

RESULT :

Classification introduction to exploratory data analysis using rattle an open source tool(r) is
executed and results verified

M.SUDHARSHAN 12 201061101117
 EXP NO.3 DATE: 27/06/23

INTRODUCTION TO REGRESSION USING RATTLE AN

OPENSOURCE TOOL.( R)

AIM :

\To perform Correlation and linear regression analysis that are the most commonly used
techniques for investigating the relationship between two quantitative variables.

Correlation Analysis

A correlation plot will display correlations between the values of variables in the dataset. In addition to
the usual correlation calculated between values of different variables, the correlation between missing
values can be explored by checking the Explore Missing check box. The first thing to notice for this
correlation plot is that only the numeric variables appear. Rattle only computes correlations between
numeric variables at this time. The second thing to note
about the graphic is that it is symmetric about the diagonal. The correlation between two variables is
the same, irrespective of the order in which we view the two variables. The third thing to note is that
the order of the variables does not correspond to the order in the dataset, but to the order of the strength
of any correlations,from the least to the greatest. This is done simply to achieve a more pleasing
graphic which is easier to take in.

M.SUDHARSHAN 13 201061101117
 We interpret the degree of any correlation by both the shape and colour of the graphic elements.
Any variable is, of course, perfectly correlated with itself, and this is reflected as the diagonal lies
on the diagonal of the graphic. Where the graphic element is a perfect circle, then there is no
correlation between the variables, as is the case in the correlation between Hours and Deductions
although in fact there is a correlation,just a very weak one. The colours used to shade the circles
give another (if perhaps redundant) clue to the strength of the correlation. The intensity of the colour
is maximal for a perfect correlation, and minimal (white) if there is no correlation. Shades of red
are used for negative correlations and blue for positive correlations.

By selecting the Explore Missing check box you can obtain a correlation plot that will show any
correlations between the missing values of variables. This is particularly useful to understand how
missing values in one variable are related to missing values in another. We notice immediately that
only three variables are included in this correlation plot. Rattle has identified
that the other variables in fact have no missing values, and so there is no point including them
in the plot. We also notice that a categorical variable, Accounts, is included in the plot even though it
was not included in the usual correlation plot. In this case we can obtain a correlation for categorical
variables since we only measure missing and presence of a value, which is easily interpreted as
numeric.

The graphic shows us that Employment and Occupation are highly correlated in their presence of
missing values. That is, when Employment has a missing value, so does Occupation, and vice versa,
at least in general. The actual correlation is 0.995 (which can be read from the Rattle text view
window), which is very close to 1. On the other hand, there is no (in fact very little at 0.013) correlation
between Accounts and the other two variables, with regard missing values. It is important to note that
the correlations showing missing values may be based on very small samples, and this information is
included in the text view of the Rattle window. For example, in this example we can see that there are
only 100, 101, and 43 missing values, respectively, for each of the three variables having any missing
values. This corresponds to approximately 5%, 5%, and 2% of the entities, respectively, having
missing values for these variables.

M.SUDHARSHAN 14 201061101117
 Rattle uses the default R correlation calculation known as Pearson's correlation, a common measure of
correlation.

M.SUDHARSHAN 15 201061101117
 Hierarchical Correlation

Rattle 2007−11−25 16:08:20

RESULT :

Introduction to regression using rattle an open source tool is executed and results are verified.

M.SUDHARSHAN 16 201061101117
 EXP NO. 4 DATE: 04/07/23

CLASSIFICATION USING THE WEKA TOOLKIT – PART 1

AIM :

Demonstration of classification rule process on dataset student.arff using j48 algorithm

This experiment illustrates the use of j-48 classifier in weka. The sample data set used in this experiment is
“student” data available at arff format. This document assumes that appropriate data pre processing has been
performed.

Steps involved in this experiment:

Step-1: We begin the experiment by loading the data (student.arff)into weka.

Step2: Next we select the “classify” tab and click “choose” button t o select the “j48”classifier.

Step3: Now we specify the various parameters. These can be specified by clicking in the text box to the right
of the chose button. In this example, we accept the default values. The default version does perform some
pruning but does not perform error pruning.

Step4: Under the “text” options in the main panel. We select the 10-fold cross validation as our evaluation
approach. Since we don’t have separate evaluation data set, this is necessary to get a reasonable idea of
accuracy of generated model.

Step-5: We now click ”start” to generate the model .the Ascii version of the tree as well as evaluation
statistic will appear in the right panel when the model construction is complete.

Step-6: Note that the classification accuracy of model is about 69%.this indicates that we may find more
work. (Either in preprocessing or in selecting current parameters for the classification)

Step-7: Now weka also lets us a view a graphical version of the classification tree. This can be done by right
clicking the last result set and selecting “visualize tree” from the pop-up menu.

Step-8: We will use our model to classify the new instances.

Step-9: In the main panel under “text” options click the “supplied test set” radio button and then click the
“set” button. This wills pop-up a window which will allow you to open the file containing test instances.

Dataset student .arff

@relation student
@attribute age {<30,30-40,>40}
@attribute Income {low, medium, high}
@attribute Student {yes, no}
@attribute credit-rating {fair, excellent}
@attribute buyspc {yes, no}
@data
%
<30, high, no, fair, no

M.SUDHARSHAN 17 201061101117
 <30, high, no, excellent, no
30-40, high, no, fair, yes
>40, medium, no, fair, yes
>40, low, yes, fair, yes
>40, low, yes, excellent, no
30-40, low, yes, excellent, yes
<30, medium, no, fair, no
<30, low, yes, fair, no
>40, medium, yes, fair, yes
<30, medium, yes, excellent, yes
30-40, medium, no, excellent, yes
30-40, high, yes, fair, yes
>40, medium, no, excellent, no
%

M.SUDHARSHAN 18 201061101117
 The following screenshot shows the classification rules that were generated when j48 algorithm is
applied on the given dataset.

M.SUDHARSHAN 19 201061101117
 Result :

Demonstration of classification rule process on dataset student.arff using j48 algorithm is

executed and result are verified.

M.SUDHARSHAN 20 201061101117
 EXP NO. 5 DATE: 11/07/23

CLASSIFICATION USING THE WEKA TOOLKIT – PART 2

AIM :

Demonstration of classification rule process on dataset employee.arff using id3 algorithm

This experiment illustrates the use of id3 classifier in weka. The sample data set used in this experiment is
“employee”data available at arff format. This document assumes that appropriate data pre processing has
been performed.

Steps involved in this experiment:

Step 1: We begin the experiment by loading the data (employee.arff) into weka.

Step2: next we select the “classify” tab and click “choose” button to select the “id3”classifier.

Step3: now we specify the various parameters. These can be specified by clicking in the text box to the
right of the chose button. In this example, we accept the default values his default version does perform
some pruning but does not perform error pruning.

Step4: under the “text “options in the main panel. We select the 10-fold cross validation as our evaluation
approach. Since we don’t have separate evaluation data set, this is necessary to get a reasonable idea of
accuracy of generated model.

Step-5: we now click”start”to generate the model .the ASCII version of the tree as well as evaluation statistic
will appear in the right panel when the model construction is complete.

Step-6: note that the classification accuracy of model is about 69%.this indicates that we may find more
work. (Either in preprocessing or in selecting current parameters for the classification)

Step-7: now weka also lets us a view a graphical version of the classification tree. This can be done by right
clicking the last result set and selecting “visualize tree” from the pop-up menu.

Step-8: we will use our model to classify the new instances.

Step-9: In the main panel under “text “options click the “supplied test set” radio button and then click the
“set” button. This will show pop-up window which will allow you to open the file containing test instances.

Data set employee.arff:

@relation employee

@attribute age {25, 27, 28, 29, 30, 35, 48}

@attribute salary{10k,15k,17k,20k,25k,30k,35k,32k}

@attribute performance {good, avg, poor}

@data

M.SUDHARSHAN 21 201061101117
 %

25, 10k, poor

27, 15k, poor

27, 17k, poor

28, 17k, poor

29, 20k, avg

30, 25k, avg

29, 25k, avg

30, 20k, avg

35, 32k, good

48, 35k, good

48, 32k, good

The following screenshot shows the classification rules that were generated when id3 algorithm is
applied on the given dataset.

M.SUDHARSHAN 22 201061101117
 RESULT :

Demonstration of classification rule process on dataset student.arff using id3 algorithm is

executed and result are verified.

M.SUDHARSHAN 23 201061101117
 EXP NO. 6 DATE: 18/07/23

PERFORMING DATA PREPROCESSING FOR DATA MINING IN

WEKA
AIM :

Demonstration of preprocessing on dataset student.arff

This experiment illustrates some of the basic data preprocessing operations that can be performed using
WEKA-Explorer. The sample dataset used for this example is the student data available in arff format.
Step1: Loading the data. We can load the dataset into weka by clicking on open button in preprocessing
interface and selecting the appropriate file.

Step2: Once the data is loaded, weka will recognize the attributes and during the scan of the data weka
will compute some basic strategies on each attribute. The left panel in the above figure shows the list of
recognized attributes while the top panel indicates the names of the base relation or table and the
current working relation (which are same initially).

Step3:Clicking on an attribute in the left panel will show the basic statistics on the attributes for the
categorical attributes the frequency of each attribute value is shown, while for continuous attributes we
can obtain min, max, mean, standard deviation and deviation etc.,

Step4:The visualization in the right button panel in the form of cross-tabulation across two attributes.
Note:we can select another attribute using the dropdown list.

Step5:Selecting or filtering attributes

Removing an attribute-When we need to remove an attribute,we can do this by using the attribute filters
in weka.In the filter model panel,click on choose button,This will show a popup window with a list of
available filters.
Scroll down the list and select the “weka.filters.unsupervised.attribute.remove” filters.

Step 6:a)Next click the textbox immediately to the right of the choose button.In the resulting dialog box
enter the index of the attribute to be filtered out.
b) Make sure that invert selection option is set to false.The click
OK now in the filter box.you will see“Remove-R-7”.
c) Click the apply button to apply filter to this data.This will
remove the attribute and create newworking relation.
Save the new working relation as an arff file by clicking save button on the
top(button)panel.(student.arff)

Discretization
1. Sometimes association rule mining can only be performed on categorical data.This requires
performing discretization on numeric or continuous attributes.

M.SUDHARSHAN 24 201061101117
 In the following example let us discretize age attribute.
• Let us divide the values of age attribute into three bins(intervals).
• First load the dataset into weka(student.arff).
• Select the age attribute.
• Activate filter-dialog box and select “WEKA.filters.unsupervised.attribute.discretize”from the list.
• To change the defaults for the filters,click on the box immediately to the right of the choose button.
• We enter the index for the attribute to be discretized.In this case the attribute is age.So we must
enter ‘1’ corresponding to the age attribute.
• Enter ‘3’ as the number of bins.Leave the remaining field values as they are.
• Click OK button.
• Click apply in the filter panel.This will result in a new working relation with the selected attribute
partition into 3 bins.
• Save the new working relation in a file called student-data-discretized.arff.

Dataset student .arff

@relation student

@attribute age {<30,30-40,>40}

@attribute income {low, medium, high}

@attribute student {yes, no}

@attribute credit-rating {fair, excellent}

@attribute buyspc {yes, no}

@data

<30, high, no, fair, no

<30, high, no, excellent, no

30-40, high, no, fair, yes

>40, medium, no, fair, yes

>40, low, yes, fair, yes

>40, low, yes, excellent, no

30-40, low, yes, excellent, yes

<30, medium, no, fair, no

<30, low, yes, fair, no

>40, medium, yes, fair, yes

<30, medium, yes, excellent, yes

M.SUDHARSHAN 25 201061101117
 30-40, medium, no, excellent, yes

30-40, high, yes, fair, yes

>40, medium, no, excellent, no

The following screenshot shows the effect of discretization.

RESULT :

Performing data preprocessing for data mining in weka is executed and result are verified.

M.SUDHARSHAN 26 201061101117
 EXP NO. 7 DATE: 25/07/23

PERFORMING CLUSTERING IN WEKA

Aim :

Demonstration of clustering rule process on dataset student.arff using simple k-means

This experiment illustrates the use of simple k-mean clustering with Weka explorer. The sample data set
used for this example is based on the student data available in ARFF format. This document assumes that
appropriate preprocessing has been performed. This istudent dataset includes 14 instances.

Steps involved in this Experiment

Step 1: Run the Weka explorer and load the data file student.arff in preprocessing interface.

Step 2: Inorder to perform clustering select the ‘cluster’ tab in the explorer and click on the choose button.
This step results in a dropdown list of available clustering algorithms.

Step 3 : In this case we select ‘simple k-means’.

Step 4: Next click in text button to the right of the choose button to get popup window shown in the
screenshots. In this window we enter six on the number of clusters and we leave the value of the seed on
as it is. The seed value is used in generating a random number which is used for making the internal
assignments of instances of clusters.

Step 5 : Once of the option have been specified. We run the clustering algorithm there we must make sure
that they are in the ‘cluster mode’ panel. The use of training set option is selected and then we click‘start’
button. This process and resulting window are shown in the following screenshots.

Step 6 : The result window shows the centroid of each cluster as well as statistics on the number and the
percent of instances assigned to different clusters. Here clusters centroid are means vectors for each
clusters. This clusters can be used to characterized the cluster.

Step 7: Another way of understanding characterstics of each cluster through visualization ,we can do this,
try right clicking the result set on the result. List panel and selecting the visualize cluster assignments.

Interpretation of the above visualization

From the above visualization, we can understand the distribution of age and instance number in each
cluster. For instance, for each cluster is dominated by age. In this case by changing the color dimension
to other attributes we can see their distribution with in each of the cluster.

Step 8: We can assure that resulting dataset which included each instance along with its assign cluster. To
do so we click the save button in the visualization window and save the result student k- mean .The top
portion of this file is shown in the following figure.
Dataset student .arff
@relation student

@attribute age {<30,30-40,>40}

@attribute income {low,medium,high}

M.SUDHARSHAN 27 201061101117
 @attribute student {yes,no}

@attribute credit-rating {fair,excellent}

@attribute buyspc {yes,no}

@data

<30, high, no, fair, no

<30, high, no, excellent, no

30-40, high, no, fair, yes

>40, medium, no, fair, yes

>40, low, yes, fair, yes

>40, low, yes, excellent, no

30-40, low, yes, excellent, yes

<30, medium, no, fair, no

<30, low, yes, fair, no

>40, medium, yes, fair, yes

<30, medium, yes, excellent, yes

30-40, medium, no, excellent, yes

30-40, high, yes, fair, yes

>40, medium, no, excellent, no

M.SUDHARSHAN 28 201061101117
 The following screenshot shows the clustering rules that were generated when simple k-means
algorithm is applied on the given dataset.

M.SUDHARSHAN 29 201061101117
 RESULT :

Demonstration of classification rule process on dataset student.arff using simpe k-means is executed
and result are verified.

M.SUDHARSHAN 30 201061101117
 EXP NO. 8 DATE:01/08/23

ASSOCIATION RULE ANALYSIS IN WEKA

AIM :

Demonstration of Association rule process on dataset contactlenses.arff using apriori algorithm

This experiment illustrates some of the basic elements of asscociation rule mining using WEKA. The sample
dataset used for this example is contactlenses.arff

Step1: Open the data file in Weka Explorer. It is presumed that the required data fields have been discretized.
In this example it is age attribute.

Step2: Clicking on the associate tab will bring up the interface for association rule algorithm.

Step3: We will use apriori algorithm. This is the default algorithm.

Step4: Inorder to change the parameters for the run (example support, confidence etc) we click on the text
box immediately to the right of the choose button.

Dataset contactlenses.arff

The following screenshot shows the association rules that were generated when apriori algorithm is applied
on the given dataset.

M.SUDHARSHAN 31 201061101117
 RESULT:

Demonstration of classification rule process on dataset student.arff using apriori algorithm is executed
and result are verified.

M.SUDHARSHAN 32 201061101117
 EXP NO. 9 DATE: 08/08/23

DATA MINING CASE STUDY USING THE CRISP-DM STANDARD

AIM :

CRISP- DM CASE STUDY – Early Prediction Of Student Success : Mining Students Enrolment
Data

This case study explores the socio-demographic variables (age, gender, ethnicity, education, work
status, and disability) and study environment (course programme and course block), that may influence
persistence or dropout of students at the Open Polytechnic of New Zealand. This case study examines
to what extent these factors, i.e. enrolment data help us in pre-identifying successful and unsuccessful
students. The data stored in the Open Polytechnic student management system covers over 450 students
who enrolled to 71150 Information Systems course was used to perform a quantitative analysis of study
outcome. Based on a data mining techniques (such as feature selection and classification trees), the most
important factors for student success and a profile of the typical successful and unsuccessful students
are identified. The empirical results show the following: (i) the most important factors separating
successful from unsuccessful students are: ethnicity, course programme and course block; (ii) among
classification tree growing methods Classification and Regression Tree (CART) was the most
successful in growing the tree with an overall percentage of correct classification of 60.5%; and
(iii) both the risk estimated by the cross-validation and the gain diagram suggests that all trees, based
only on enrolment data are not quite good in separating successful from unsuccessful students. The
implications of these results for academic and administrative staff are discussed.

INTRODUCTION
Increasing student retention or persistence is a long term goal in all academic institutions. The
consequences of student attrition are significant for students, academic and administrative staff. The
importance of this issue for students is obvious: school leavers are more likely to earn less than those
who graduated. Since one of the criteria for government funding in the tertiary education environment in
New Zealand is the level of retention rate, both academic and administrative staff are under pressure to
come up with strategies that could increase retention rates on their courses and programmes. The most
vulnerable students to low student retention at all institutions of higher education are the first-year
students, who are at greatest risk of dropping out in the first term or semester of study or not completing
their programme/ degree on time.
Therefore most retention studies address the retention of first-year students. Consequently, the early
identification of vulnerable students who are prone to drop their courses is crucial for the success of any
retention strategy. This would allow educational institutions to undertake timely and pro-active measures.
Once identified, these ‘at-risk’ students can be then targeted with academic and administrative support
to increase their chance of staying on the course.

The main objective of this study is to explore factors that may impact the study outcome in the
Information Systems course at the Open Polytechnic. The Information Systems course is a core course
for those majoring in IT and for most students an entry point, i.e. the first course they are taking with the
Open Polytechnic. This issue have not been examined so far for Open Polytechnic and this case study
attempts to fill the gap.

M.SUDHARSHAN 33 201061101117
 More specifically the enrolment data were used to achieve the following objectives:

• Build models for early prediction of study outcome using the student enrolment data
• Evaluate the models using cross-validation and misclassification errors to decide
which model outperforms other models in term of classification accuracy
•Present results which can be easily understood by the users (students, academic and
administrativestaff)

At the time of enrolment in the Open Polytechnic of New Zealand, the only information. i.e. variables
we have about students are those contained in their enrolment forms. The question we are trying to
address in this case study is whether we can use the enrolment data alone to predict study outcome for
newly enrolled student.

Framework for Data Mining Process

Framework for data mining applications is based on the CRISP-DM Model created by a consortium of
NCR, SPSS, and Daimler-Benz companies. The modified version of the CRISP- DM model is presented
on Figure 1, following the project through the general life cycle from business and data understanding,
data preparation, modeling, evaluation and deployment. The feedback from deploymentto data and
business understanding illustrates the iterative nature of a data mining process.

Figure 1: Modified CRISP-DM Model Version 1

Business Understanding
The business understanding phase begins with the setting up of goals for the data mining project.
In this paper that would be an increasing understanding of the pre-enrolment factors that may
prevent students from successfully complete the course.
Because we are planning to increase completion rate on the Information Systems course
understanding its students population and patterns in the pre-enrolment data becomes necessary
before we start developing a predictive model. In this phase we are coming up with the following
questions:
what is the profile of a student who successfully completes this course? Can the successful vs. unsuccessful
student be distinguished in terms of demographic features (such as gender, age or ethnic origin) or study

M.SUDHARSHAN 34 201061101117
 environment (such as course programme, faculty or course block)? Depending on the answers to these
questions we consider the methods and approaches that can be adopted to increase the completion rate.
Data Understanding
The scope of our research in terms of data used is limited by the data available in the Open
Polytechnic Student Management System (known as Integrator) and the enrolment form used for
collecting data from newly enrolled students. It is important to have a full understanding of the
nature of the data and how itwas collected and entered before proceeding further. In this phase an
initial data exploration using a pivot table was also conducted to get some insight in the data.
The enrolment form asks students to enter the following information: demographic (gender, date
of birth,ethnic origins, disability, and work status), academic (secondary school qualification,
course programme, course faculty, and course block), and contact details. Once the data from
the enrolment form is entered into Integrator and the Enrolment Section process the application,
the enrolment date is recorded and student becomes enrolled on the course.
Data Preparation
In this phase the data are put into a form suitable for the modeling phase. If required some selected
variables are combined, transformed or used to create a new variables. For example, enrolment
date and the course block start date were used to generate a variable labeled as “early
enrolment”. Any data excluded from the data set is documented and their removal explained. For
example, only a few students enrolled to Bachelor of Arts programme were on the Information
Systems course. They were removed from further analysis for the reason explained later in Data
and Methodology section. Data are cleaned for any duplication of records. For example, in case
ofInformation Systems course, the course code changed in the past. If student enrolled in the time
when thechange in the course code happened and then re-enrolled on the same course, two
records exist in the data set for the same student and the same course, but under two different
course codes. In this case datafor this student were merged into one, single record. The
dependent variable “study outcome” with twopossible outcomes (labeled as Pass and Fail)
indicates whether students successfully completed thecourse or failed the course due to
voluntary transfer/ withdrawal or academic withdrawal or simplybecause not fulfilling course
pass requirements.
Modeling
In this phase we choose and ran models on the training data set. Then we decided whether a
suitable model for the data set was found that was acceptable from both analytical and managerial
standpoint. In this phase we decided to use classification tree models with four different tree
growing criteria.
Evaluation
The final models from the previous phase are then applied on a testing, i.e. a validation data set
with the aim of assessing their predictive accuracy and consistency with the results obtained for
the training data set. This phase involves an iterative process of fitting different versions of
models to training and testing data set, each time evaluating their predictive performance.
Deployment
Once we decided on the final model we can apply it to current data not used during the modeling
and evaluation phase. This process is known as scoring. The model results are used to address the
issues identified in the business understanding phase. The results should be presented in the user
friendly format and prepared for use by administrative staff. The final model should retain the
highest predictive accuracy and if it is to be continuously used it should be regularly updated,
particularly if some organizational changes occur or if new factors are brought in. For example, if the
new information about financial support is added to the student record or enrolment form, then the
model should consider the new factor that might be relevant for study outcome.

M.SUDHARSHAN 35 201061101117
 Data and Methodology

Data preparation
Variables definition and their domains are presented in Table 1.

From the initial dataset all students granted cross-credit or credit were excluded because they didn’t
actually study this course. The courses they have previously completed were recognized and credited for
this course. We also removed 6 students enrolled under Bachelor of Arts program. Such a small number
of students in this category do not allow generalization of the results for this particular category. The
total number of data was reduced to 453.

We needed to clarify the definition of study outcome that we used in our analysis. We considered only
two possible outcomes, labeled as: Pass and Fail. Students labeled Pass successfully completed the
course. Students labeled Fail transferred or withdrew from the course voluntarily or academic registry
withdrew them for not completing the in-course assessments. Those students who stayed on the course
until the end of the course but scored less than the course pass mark were also allocated into this category.
Because of the data set size (only 453 students) numeric continuous variable such as age was converted
into a categorical variable with only three age groups. In Ethnicity we combined Maoriand Pacific Island
students because of two reasons: they were found to be no different in preliminary bivariate analysis and
combined together constitute a small proportion of the data (less than 10%). Combining them into one
ethnic group helps with model parsimony. The Secondary school variable combines all students with no
secondary school up to NCEA Level 2 on the New Zealand National Qualifications Framework into one
group.

Though the software applications skills (spreadsheet and database in particular) are very important on
this course we do not take into account the different skill levels present on the Information Systems
course. Since the students enrolling on this course have different backgrounds and levels of interest in
computing, we would expect that skill level has also a significant impact on the study outcome. However,
the information about their Office skills level is not available in the moment of enrolment, sonot included
this factor in our analysis.
Variable Description (Domain)
tudent Demographics
Gender Student gender (binary: female or male)
Age Student’s age (numeric: 1 – under 30, 2 – 30 to 40 or 3 – over 40)

Ethnicity Student’s ethnic group (nominal: Pakeha, Maori & Pacific Islanders
or Others)

Disability Student has a disability (binary: yes or no)

Secondary school Student’s highest level of achievement from a secondary school on
the New Zealand National Qualifications Framework (nominal: No
secondary qualification, NCEA1, NCEA2, University entrance,
NCEA3, Overseas or Other)
Work status Student is working (binary: yes or no)
Early enrolment Student enrolled for the first time in the course before start of
the course (binary: yes or no)

M.SUDHARSHAN 36 201061101117
 Study Environment
Course programme Programme (nominal: Bachelor of Business or Bachelor of Applied
Science)
Course block Semester in which a course is offered (Semester 1, Semester 2
or Semester 3)
Study outcome Study outcome (nominal: Pass – successful completion, Fail –
unsuccessful completion includes also withdrawals, academic
withdrawals and transfers)
Table 1: Description of variables and their domains

Methodology
Three types of data mining approaches were conducted in this study. The first approach is
descriptive which is concerned with the nature of the dataset such as the frequency table and the
relationship betweenthe attributes obtained using cross tabulation analysis (contingency tables).
In addition, feature selectionis conducted to determine the importance of the prediction variables
for modeling study outcome. The third type of data mining approach, i.e. predictive data mining
is conducted by using four different classification trees. Finally, a comparison between these
classification tree models was conducted to determine the best model for the dataset.

We decided to use the classification tree models because of some advantages they may have over traditional
statistical models such as logistic regression and discriminant analysis traditionally used in retention
studies. First, they can handle a large number of predictor variables, far more than the logistic regression
and discriminant analysis would allow. Secondly, the classification tree models are non- parametric and
can capture nonlinear relationships and complex interactions between predictors and dependent variable.

Results and Discussion

Before growing the classification trees we summarized the variables by categories and by study outcome,
i.e. whether students passed or failed the course. Feature selection was used to rank the variables
by their importance for further analysis. Finally, the classification tree results for four different
growing methods are presented.

Summary Statistics
As part of the data understanding phase we carried out the cross-tabulation for each variable and the
study outcome after preparing and cleaning the data. The Table 2 reports the results. Based on the
results shown majority of Information Systems students are female (over 63%). However, percentage
of female students who successfully complete the course are higher (65%) which suggests that female
students are more likely to pass the course than their male counterpart. When it comes to age over 68%
of students are above 30 with the age group between 30 and 40 being majority. This age group is also
more likely to fail the course because their percentage of students who failed the course in this age
group (39.7%) is higher than their overall participation in the student population (38.6%).

Table 2: Descriptive statistics (percentage) – Study outcome (453 students)

Variable Domain Count Total Pass Fail
Gender Female 286 63.1 65.0 61.5
Male 167 36.9 35.0 38.5
Age Under 30 136 30.0 30.4 29.7

M.SUDHARSHAN 37 201061101117
 Between 30 and 40 175 38.6 37.4 39.7
Above 40 142 31.3 32.2 30.5
Disability Yes 19 4.2 3.3 5.0
No 434 95.8 96.7 95.0
Ethnicity Pakeha 318 70.2 75.7 65.3
Maori & Pacific Islanders 41 9.1 2.8 14.6
Others 94 20.8 21.5 20.1
Secondary school No secondary school / NCEA Level 1 or 2 183 40.4 36.9 43.5
University Entrance / NCEA Level 3 163 36.0 38.3 33.9
Overseas Qualification or Other 107 23.6 24.8 22.6
Work status Yes 351 77.5 78.0 77.0
No 102 22.5 22.0 23.0
Early enrolment Yes 317 70.0 72.4 67.8
No 136 30.0 27.6 32.2
Course programme Bachelor of Business 305 67.3 73.8 61.5
Bachelor of Applied Sciences 148 32.7 26.2 38.5
Course block First semester 139 30.7 31.3 30.1
Second semester 201 44.4 48.6 40.6
Third semester 113 24.9 20.1 29.3
Disability was shown to be a disadvantage for Information Systems students. Students with it are
more likely to fail than those without it. There are huge differences in percentage of students who
successfully completed the course depending on their ethnic origin. Though Maori and Pacific
Islanders make 9.1% ofall students on this course their participation is significantly lower in the
“Pass” subpopulation (i.e. 2.8%)and higher in the “Fail” subpopulation (14.6%). Based on these
results we can say that students with this ethnic origin are identified as students “at- risk”. Further
methods of data mining will confirm this statement.

A substantial number of students (over 40%) don’t have secondary school qualification higher
than NCEA Level 2 on the New Zealand National Qualification Framework and they are more vulnerable than
the other two categories in this variable. Over three-fourths of Information Systems students are working and
studying at the same time. Though the difference between those who work and those who do not is not statistically
significant, it is interesting to note that the students who are working are more likely to pass the course than those
not working.

We are using early enrolment as a proxy for motivation and good time management skills. Students who are
motivated and are planning their study in advance will also enroll well before the enrolment closing date. The
opposite category (late comers) participates with 30% in the total
number of students, but these students are more likely to fail the course. Their participation in the “Fail”
subpopulation increased from 30% to 32.2%.

One third of students on this course enrolled on the Bachelor of Applied Sciences program. They are
more likely to fail the course when compared with students enrolled on the Bachelor of Business
program. Finally, students studying this course in the summer semester are more likely to fail than
those studying in the second and first semester.

Feature Selection
The number of predictor variables is not so large and we don’t have to select the subset of variables for further
analysis which is the main purpose of applying feature selection to data. However, feature selection could be
also used as a pre-processor for predictive data mining to rank
predictors according to the strength of their relationship with dependent or outcome variable. During the
feature selection process no specific form of relationship, neither linear nor nonlinear, is assumed. The
outcome of the feature selection would be a rank list of predictors according to their importance for further
analysis of the dependent variable with the other methods for regression and classification.

M.SUDHARSHAN 38 201061101117
 Results of feature selection are presented in Figure 2 on the importance plot and also in Table 3. The top
three predictors for the study outcome are: ethnic origin of students, course programme they are enrolled
on and course block, i.e. semester in which they are study.

Table 3: Best predictors for dependent variable

Variable Chi- P-
square value

Ethnicity 19.35 0.00006

Course programme 7.800.00523
Course block 5.510.06354
Secondary school 2.060.35748
Early enrolment 1.160.28131
Disability 0.860.35363
Gender 0.580.44774
Age 0.280.86750
Work status 0.070.78940

From Table 3, P-values from the last column only the first three chi-square values are significant at 10% level.
Though the results of the feature selection suggested continuing analysis with only the subset of predictors, which
includes ethnicity, course programme and course block, we have included all available predictors in our
classification tree analysis. We follow an advice given in Luan & Zhao (2006) who suggested that even though
some variables may have little significance to the overall prediction outcome, they can be essential to a specific
record.

Classification Trees
The objective of an analysis based on a classification tree is to identify factors that contribute the most to separation
of successful from unsuccessful students. When the classification tree is formed we can calculate the

M.SUDHARSHAN 39 201061101117
 probability of each student being successful. Once the classification tree is formed, it could be used
in the newdata set to predict the study outcome for newly en-rolled students.

The classification trees for study outcome are given in Figure 2 and 4. In each tree node the num - ber
of th rd successful students (4 line, last column) and unsuccessful students
(line, last column) is given, aswell as nd th the percentages for each category
(2column) and relative and absolute size of the node(5 line). The variable names above the nodes are
the predictors that provided the best split for the node accordingto the classification and regression
tree-style exhaustive search for uni variate splits method. This method looks atall possible splits for
each predictor variable at each node. The search stops when the split with the largestimprovement in
goodness of fit, based on the Gini measure of node impurity, is found. Immediately above thenodes
are categories which describe these nodes. Note that all available predictor variables in the dataset
wereincluded in the classification tree analysis in spite their insignificance detected in the feature
selection section.

CHAID, exhaustive CHAID and QUEST

Three classification tree growing methods, namely: CHAID, exhaustive CHAID and QUEST
generated exactly the same tree structure presented in Figure 3. It shows that only 2 variables
were used to construct the tree: (1) ethnicity and (2) course program. All the other student
demographics variables were used but not included in the final model. We could change the stopping
criteria to allow further growing of the tree, but that would result in nodes with just a few students. In the
most extreme case we can continue splitting the tree until we create a terminal node for every student.
However, we would get a model, i.e. classification tree that fits data better, but with more likely poor
performance when used on a new data set. This phenomenon is known as overfitting the tree. The largest
successful group (i.e. students who successfully completed the course) consists of 274 (60.5%) students
(Node 3). Ethnic origin of students in this group is either Pakeha or other ethnic groups (excluding Maori
and Pacific Islands students). Students in this group opted for the Bachelor of Business programme. The
largest unsuccessful group (i.e. students who were unsuccessful) contains 138 students (30.5% of all
participants) (Node 4). They belong to either Pakeha or other ethnic groups (excluding Maori and Pacific
Islands students). The next largest group considered also as unsuccessful students, contains 41, i.e. 9.1%
of all students, where 75.4% of them are unsuccessful (Node 1). They are described as Maori and Pacific
Islands students.

M.SUDHARSHAN 40 201061101117
 Figure 3: CHAID, exhaustive CHAID and QUEST classification tree

The overall percentage of correct classification for the study outcome is only 59.4%
(Table 4).This percentage of correct classification was achieved with 2 variables only.
The cross-validation estimate of the risk is 0.406 indicates that the category predicted by the model
(successfulor unsuccessful student) is wrong for 40.6% of the cases. So the risk of misclassifying a
student is approximately 41%. This result is consistent with the results in the CHAID classification
matrix (Table 4). The Overall percentage shows that the model only classified correctly 59% of
students.
Table 4: CHAID classification matrix

Observed
Fail

Pass
Percent
correct
Fail 117 122 49.0%
Pass 62 152 71.0%
Overall percentage 65.4% 55.5% 59.4%
With large numbers of false positives (122) and few false negatives (62), the CHAID model is in itself poor
at identifying an unsuccessful student (positive predictive value is only 49%). It will, however, pick up
65.4% of all unsuccessful students (the sensitivity). The predictive values,
which take into account the prevalence of failing the course, are generally more important in determining the
usefulness of a prediction model. The negative predictive value was of more concern to the course because the

M.SUDHARSHAN 41 201061101117
 objective was to minimize the probability of being in error when
deciding that a student is not at risk for not completing the course.
However the CHAID model, as a classification tool, will Predicted
pick-up

with high probability successful students (negative predictive

value is 71%) and correctly identifies 55.5% of those who pass the course (the specificity).

The classification matrix indicates also another problem with the model. For unsuccessful students it
predicts failure for only 49% of them, which means that 51% of unsuccessful students are inaccurately
classified with the successful students. Practical consequence of this misclassification is that these
students would not received additional learning support provided to the students “at risk”, simply because
they will be classified by the model among successful students. This feature of the model is more critical
than misclassification of the successful students among unsuccessful students (29% of successful
students belong to this category). In this case these students may receive additional learning support even
though they don’t need it. One option to increase percentage of correctly classified unsuccessful students
is to change the misclassification cost matrix. With this option there is always a trade-off between
increasing the percentage of correct classification of unsuccessful students and decreasing percentage of
correct classification for successful students as well as decreasing the percentage of overall correct
classification. In this case the increased cost for misclassification of unsuccessful to the successful group
of students decreased significantly both remaining percentage of correct classification whichwas not
compensated in equivalent increase in the initial 49%.

CART
Figure 5 shows the CART classification tree for study outcome. It shows that only three variables
were used to construct the tree: (1) ethnicity, (2) course programme and (3) course block.The
largest successful group (i.e. students who successfully completed the course) consists of
215(47.5%) students (Node 5). The ethnic origin of students in this group is either Pakeha or
otherethnic groups (excluding Maori and Pacific Islands students). Students in this group enrolled
on the Bachelor of Business programme in either Semester 1 or Semester 2. The largest
unsuccessful group (i.e. students who were unsuccessful) contains 138 students (30.5% of all
participants)(Node 4). They belong to either Pakeha or other ethnic groups (excluding Maori and
Pacific Islands students). The next largest group considered also as unsuccessful students,
contains 41,i.e 9.1% of all students, where 75.4% of them are unsuccessful (Node 1). They are
described as Maori and PacificIslands students.

Table 5: CART classification matrix

Observed

Fail

Pass Percent
correct
Fail 149 90 62.3%

M.SUDHARSHAN 42 201061101117
 Pass 89 125 58.4%
Overall percentage 62.6% 58.1% 60.5%

Figure 5: CART classification tree

he cross-validation estimate of the risk is 0.446 indicates that the category predicted by the model
(successful or unsuccessful student) is wrong for 44.6% of the cases. The CART classification
matrix (Table 5) shows that model classify correctly 61% of students. This is slight increase in
comparison to the CHAID model. The numbers of false positives (90) for the CART model
decreases and therefore increasing the positive predictive value to 62.3%. In other words it will
work better than the CHAID model at identifying an unsuccessful student. The price paid for
increasing accuracy is reflected in decreasing sensitivity. The CART model will pick up 62.6%
of all unsuccessful students (CHAID model 65.4%). At the same time the specificity will increase
to 58.1% (CHAID model 55.5%).

M.SUDHARSHAN 43 201061101117
 Concluding Remarks
This study examines the background information from enrolment data that impacts upon the study
outcome of Information Systems students at the Open Polytechnic. Based on results from feature
selection (Figure 2 and Table 3), the CHAID tree presented in Table 4 and Figures 3 and 4, the
CART trees presented in Table 5 and Figures 5 and 6 it was found that the most important factors
that help separate successful from unsuccessful students are ethnicity, course programme and
course block. Demographic data such as gender and age though significantly related to the study
outcome, according to the feature selection result, were not used in the classification trees.
Unfortunately the classification accuracy from the classification trees was not very high. In the
case of the CHAID tree the overall classification accuracy was 59.4% and in the case of the CART
tree slightly higher at 60.5%. This would suggest that the background information (gender, age,
ethnicity, disability, secondary school, work status,and early enrolment) gathered during the
enrolment process, does not contain sufficient information for an accurately separation of
successful and unsuccessful students.

Classifying students based on pre-enrollment information and the rules presented for each node
would allow the administrative and academic staff to identify students who would be “at risk” of
dropping the course even before they start with their study. Then the student support systems,
such as orientation, advising, and mentoring programs, could be used to positively impact the
academic successes of such students.

This study is limited in three main ways that future research can perhaps address. Firstly, this
research is based on background information only. Leaving out other important factors (academic
achievement, numberof courses completed, motivation, financial aids, etc.) that may affect study
outcome, could distort results obtained with classification trees. For example, including the
assignment mark after the submission of the first course assignment would probably improve
predictive accuracy of the models. To improve the model, more attributes could be included to obtain
prediction models with lower misclassification errors. However, the model in this case would not
be a tool for pre-enrolment, i.e. early identification of ‘at-risk’ students. Secondly, we used a
dichotomous variable for the study outcome with only two categories: pass and fail. However,
splitting the fail category into those who stayed on the course but eventually failed the course and
those who voluntary transfer or were withdrawn from the course would probably provide better
profiling for each of the three categories of study outcomes. The only problem we might have with
the three categories for study outcome is a low prediction accurate as a result of relatively small data
set for the course. Thirdly, from a methodological point of view an alternative to a classification tree
should be considered. The prime candidates to be used with this data set are logistic regression and
neural networks.

RESULT :

Analysis done on the above case study , examined the background information from enrolment data
that impactsupon the study outcome of Information Systems students at the Open Polytechnic.

M.SUDHARSHAN 44 201061101117
 EXP NO. 10 DATE: 22/08/23

DATA MINING CASE STUDY USING THE CRISP-DM

STANDARD
AIM :

CRISP-DM CASE STUDY - Analysing Automobile Warranty Claims: Example Of The CRISP–
DM Industry Standard Process In Action

Quality assurance continues to be a priority for automobile manufacturers, including Daimler

Chrysler. Jochen Hipp of the University of Tubingen, Germany, and Guido Lindner of
DaimlerChrysler AG, Germany, investigated patterns in the warranty claims for DaimlerChrysler
automobiles.

1.Business Understanding Phase

DaimlerChrysler’s objectives are to reduce costs associated with warranty claims and improve
customer satisfaction. Through conversations with plant engineers, who are the technical experts
in vehicle manufacturing, the researchers are able to formulate specific business problems, such
as the following:
• Are there interdependencies among warranty claims?
• Are past warranty claims associated with similar claims in the future?
• Is there an association between a certain type of claim and a particular garage?
The plan is to apply appropriate data mining techniques to try to uncover these and other
possibleassociations.

2.Data Understanding Phase

The researchers make use of DaimlerChrysler’s Quality Information System (QUIS), which contains
information on over 7 million vehicles and is about 40 gigabytes in size. QUIS contains production
details about how and where a particular vehicle was constructed, including an average of 30 or more
sales codes for each vehicle. QUIS also includes warranty claim information, which the garage
supplies, in the form of oneof more than 5000 possible potential causes.
The researchers stressed the fact that the database was entirely unintelligible to domain non-experts:
“So experts from different departments had to be located and consulted; in brief a task that turned
out to be rather costly.” They emphasise that analysts should not underestimate the importance,
difficulty and potential cost of this early phase of the data mining process, and that shortcuts here
may lead to expensive reiterations of the process downstream.

3. Data Preparation Phase

The researchers found that although relational, the QUIS database had limited SQL access. They
needed to select the cases and variables of interest manually, and then manually derive new variables
that could be used for the modelling phase. For example, the variable number of days from selling
date until first claim had to be derived from the appropriate date attributes.
They then turned to proprietary data mining software, which had been used at DaimlerChrysler on
earlier projects. Here they ran into a common roadblock – that the data format requirements varied
from algorithm to algorithm. The result was further exhaustive pre-processing of the data, to
transform the attributes into a form usable for model algorithms. The researchers mention that the
data preparation phase took much longer than they had planned.

M.SUDHARSHAN 45 201061101117
 4.Modelling Phase
Since the overall business problem from phase 1 was to investigate dependence among the warranty
claims, the researchers chose to apply the following techniques: (1) Bayesian networks and (2)
Association rules. Bayesian networks model uncertainty by explicitly representing the conditional
dependencies among various components, thus providing a graphical visualisation of the dependency
relationships among the components.
As such, Bayesian networks represent a natural choice for modelling dependence among warranty claims.
Association rules are also a natural way to investigate dependence among warranty claims since the
confidence measure represents a type of conditional probability, similar to Bayesian networks.

The details of the results are confidential, but we can get a general idea of the type of dependencies
uncovered by the models. One insight the researchers uncovered was that a particular combination
of construction specifications doubles the probability of encountering an automobile electrical
cable problem. DaimlerChrysler engineers have begun to investigate how this combination of
factors can result in an increase in cable problems.

The researchers investigated whether certain garages had more warranty claims of a certain type than
did other garages. Their association rule results showed that, indeed, the confidence levels for the
rule “If garage X, then cable problem,” varied considerably from garage to garage. They state that
further investigation is warranted to reveal the reasons for the disparity.

5.Evaluation Phase
The researchers were disappointed that the support for sequential-type association rules was relatively
small, thus precluding generalisation of the results, in their opinion. Overall, in fact, the researchers
state: “In fact, we did not find any rule that our domain experts would judge as interesting, at least at
first sight.” According to this criterion, then, the models were found to be lacking in effectiveness and
to fall short of the objectives set for them in the business understanding phase. To account for this, the
researchers point to the “legacy” structure of the database, for which automobile parts were
categorised by garages and factories for historic or technical reasons and not designed for data mining.
They suggest adapting and redesigning the database to make it more amenable to knowledge
discovery.

6.Deployment Phase
The researchers have identified the foregoing project as a pilot project, and as such, do not intend to
deploy any large-scale models from this first iteration. After the pilot project, however, they have
applied the lessons learned from this project, with the goal of integrating their methods with the
existing information technology environment at DaimlerChrysler. To further support the original goal
of lowering claims costs, they intend to develop an intranet offering mining capability of QUIS for all
corporate employees.

46
M.SUDHARSHAN 201061101117
 DISCUSSION : Lessons drawn from this case study.

First, the general impression one draws is that uncovering hidden nuggets of knowledge in databases is a
rocky road. In nearly every phase, the researchers ran into unexpected roadblocks and difficulties. This
tells us that actually applying data mining for the first time in a company requires asking people to do
something new and different, which is not always welcome. Therefore, if they expect results, corporate
management must be 100% supportive of new data mining initiatives. Another lesson to draw is that
intense human participation and supervision is required at every stage of the data mining process. For
example, the algorithms require specific data formats, which may require substantial pre-processing.
Regardless of what some software vendor advertisements may claim, you can’t just purchase some data
mining software, install it, sit back and watch it solve all your problems. Data mining is not magic.
Without skilled human supervision, blind use of data mining software will only provide you with the
wrong answer to the wrong question applied to the wrong type of data. The wrong analysis is worse
than no analysis, since it leads to policy recommendations that will probably turn out to be expensive
failures.

Finally, from this case study we can draw the lesson that there is no guarantee of positive results
when mining data for actionable knowledge, any more than when one is mining for gold. Data
mining is not a panacea for solving business problems. But used properly, by people who understand
the models involved, the data requirements and the overall project objectives, data mining can
indeed provide actionable and highly profitable results.

RESULT:

CRISP-DM CASE STUDTY- Analysing Automobile Warranty Claims: Examples of the CRISP-DM
Industry Standard Process in action is done and output verified.

47
M.SUDHARSHAN 201061101117