Tài liệu Module 17: Introduction to Data Mining pptx

Contents Overview 1 Training a Data Mining Model 12 Building a Data Mining Model with Browsing the Dependency Network 23 Lab A: Creating a Decision Tree with Review 32 Module 17: I

Contents

Overview 1

Training a Data Mining Model 12

Building a Data Mining Model with

Browsing the Dependency Network 23

Lab A: Creating a Decision Tree with

Review 32

Module 17: Introduction

to Data Mining

purpose, without the express written permission of Microsoft Corporation If, however, your only means of access is electronic, permission to print one copy is hereby granted

Microsoft may have patents, patent applications, trademarks, copyrights, or other intellectual property rights covering subject matter in this document Except as expressly provided in any written license agreement from Microsoft, the furnishing of this document does not give you any license to these patents, trademarks, copyrights, or other intellectual property

 2000 Microsoft Corporation All rights reserved

Microsoft, BackOffice, MS-DOS, Windows, Windows NT, <plus other appropriate product

names or titles Replace this example list with list of trademarks provided by copy editor Microsoft is listed first, followed by all other Microsoft trademarks in alphabetical order > are either registered trademarks or trademarks of Microsoft Corporation in the U.S.A and/or other countries

<This is where mention of specific, contractually obligated to, third party trademarks, which are added by the Copy Editor>

The names of companies, products, people, characters, and/or data mentioned herein are fictitious and are in no way intended to represent any real individual, company, product, or event, unless otherwise noted

Other product and company names mentioned herein may be the trademarks of their respective owners

Instructor Notes

This module introduces students to data mining and explains how to build and browse data mining models by using Microsoft® SQL Server™ 2000 Analysis Services Students will learn fundamental data mining terminology, concepts, techniques, and algorithms

This is an overview module that focuses on the use of built-in Analysis Manager wizards It is not intended to provide in-depth knowledge of data mining

After completing this module, students will be able to:

! Describe data mining characteristics, applications, and modeling techniques

! Describe the process of training a model

! Use the online analytical processing (OLAP) Mining Model Wizard to edit, process, and explore the decision trees

! Analyze relational data relationships in the dependency network browser

! Describe the steps required to build a clustering model by using OLAP data

Materials and Preparation

This section lists the required materials and preparation tasks that you need to teach this module

Required Materials

To teach this module, you need Microsoft PowerPoint® file 2074A_17.ppt

Preparation Tasks

To prepare for this module, you should:

! Read all the materials for this module

! Read the instructor notes and margin notes

! Practice combining the lecture with the demonstrations

! Complete the lab

! Review the Trainer Preparation presentation for this module on the Trainer Materials compact disc

! Review any relevant white papers that are located on the Trainer Materials compact disc

Presentation:

40 Minutes

Lab:

20 Minutes

Demonstration: Determining Why Students Attend College

The following demonstration procedures provide information that will not fit

in the margin notes or is not appropriate for student notes

In this demonstration, you will create a data mining model by using a decision tree with relational data Specifically, you will create a decision tree that determines why students attend college

You will create a new OLAP database with a data source connecting to the

Module 17 relational database

1 In Analysis Manager, expand the Analysis Servers folder, right-click your

local server, and then click New Database

2 Enter Module 17 as the database name, and then click OK

3 Expand the Module 17 database, right-click the Data Sources folder, and then click New Data Source

4 On the Provider tab of the Data Link Properties dialog box, click

Microsoft OLE DB Provider for SQL Server Click Next

5 Type localhost in Step 1

6 In Step 2, click Use Windows NT Integrated security

7 In Step 3, click Module 17 from the list of databases Click OK

In this procedure, you will create the data mining model by selecting source, case table, data mining technique, and key column

1 In the Module 17 database, right-click the Mining Models folder, and then click New Mining Model

2 At the welcome page, click Next

3 From the Select source type step of the Mining Model Wizard, click

Relational data, and then click Next

Point out that either relational tables or OLAP cubes can be used as source data For this model, you are accessing relational data

4 From the Select case tables step, in the Available tables list, click College

Plans, and then click Next

5 From the Select data mining technique step, in the Technique list, click

Microsoft Decision Trees, and then click Next

Two algorithms ship with Analysis Services: Microsoft Decision Trees and Microsoft Clustering Use the Decision Trees algorithm for this

demonstration

6 From the Select the key column step, in the Case key column list, click

StudentID, and then click Next

Demonstration:

10 Minutes

! To select input and predictable columns for the mining model

1 From the Select input and predictable columns step of the Mining Model Wizard, in the Available columns list, click CollegePlans at the bottom of

the column list

2 Click the top arrow (>) to choose CollegePlans as a predictable column

3 In the Available columns list, click Gender, and then click the bottom arrow (>) to choose that column as an input column

4 In the Available columns list, click ParentIncome, and then click the bottom arrow (>) to choose that column as an input column

5 In the Available columns list, click IQ, and then click the bottom arrow (>)

to select that column as an input column

6 In the Available columns list, click ParentEncouragement, and then click the bottom arrow (>) to select that column as an input column Click Next

In this procedure, you name the model, initiate processing and then close the wizard

1 From the Finish the mining model wizard step, in the Model name box, type CollegePlans

2 Click Finish to create and process the model

3 When the model has completed processing, click Close to close the Process

dialog box

1 In the Relational Mining Model Editor, click the Content tab

2 In the Content Detail pane, click the All node

View the Totals tab of the Attributes pane, and point out that more than 67

percent of the students interviewed do not plan to attend college

3 Click the Parent Encouragement = Encouraged node

Point out to the students that parental encouragement is the most dominant attribute in this model More than 57 percent of students that are encouraged

by their parents plan to attend college

4 Click Parent Encouragement = Not Encouraged

Fewer than 7 percent of students who are not encouraged by their parents plan to attend college

5 Close the Relational Mining Model Editor

Module Strategy

Use the following strategy to present this module:

The structure of this module is multiple demonstrations showing students how

to build and browse various types of data mining models Except for the first example about students attending college, the demonstrations are documented directly in the student manual Integrate your lecture with live demonstration following the procedures included in the student notes Encourage students to follow along with your demonstrations on their computers Some students may choose to watch your demonstrations only, which is also acceptable

! Introducing Data Mining The case study introduces students to data mining Data mining may be new

to many students and should be described in very simple terms highlighting the business application and uses Emphasize to students why this

technology is useful and complementary to the other forms of analysis they have been exposed to Then describe the various data mining techniques that are available

! Training a Data Mining Model Describe the process required to create a data mining model Define training data and cases

! Building a Data Mining Model with OLAP Data Introduce students to the membership card scenario Use the membership card scenario to step students through the process of building a data mining model with OLAP data by using the Mining Model Wizard Describe each step in the process—selecting the data mining technique, selecting the case, selecting the training data, creating a dimension and virtual cube, and browsing the data mining model

! Browsing the Dependency Network Demonstrate how to browse the dependency network Explain that the Dependency Network Browser can be used to view all the relationships in your model

Overview

! Introducing Data Mining

! Training a Data Mining Model

! Building a Data Mining Model with OLAP Data

This module provides you with an introduction to Microsoft® SQL Server™

2000 Analysis Services Data Mining

The objective of the module is to introduce you to both data mining principles and applications while exploring the Analysis Services wizard-driven interface for creating data mining models

After completing this module, you will be able to:

! Describe data mining characteristics, applications, and modeling techniques

! Describe the process of training a model

! Use the online analytical processing (OLAP) Mining Model Wizard to edit, process, and explore the decision trees

! Analyze relational data relationships in the dependency network browser

! Describe the steps required to build a clustering model by using OLAP data

In this module, you will learn

about data mining, how data

mining can be used to

address business

application requirements,

and how to create data

mining models by using the

Analysis Manager

# Introducing Data Mining

! Defining Data Mining

! Data Mining Applications

! Data Mining Models

! Introductory Example

! Exploring the Decision Tree

This section introduces data mining concepts, including:

! Defining data mining

! Discussing how data mining can be applied to solve common business applications

! Describing what data mining models are available

! Presenting a simple example of how data mining can be used

! Exploring the decision tree

Topic Objective

To introduce the concept of

data mining

Lead-in

In this section, you will be

introduced to a simple case

study example In that

example, data mining will be

defined, common

applications and techniques

discussed, and its role in the

data warehouse explored

Defining Data Mining

! Is The Process of Deducing Meaningful Patterns and Rules from Large Quantities of Data

! Searches for Patterns in Data Rather than Answering Predefined Questions

! Is Used To:

$ Provide historical insights

$ Predict future values or outcomes

$ Close the loop for analysis

In many organizations, data volumes are so large that it is difficult, even for the most seasoned analyst, to identify the key information most relevant to

managing the business

Data mining is the automatic or semi-automatic process of deducing meaningful patterns and rules from large quantities of data These patterns provide valuable insights to business managers and offer information that may be overlooked by more traditional manual methods of analysis

Data mining programs search for patterns in data rather than answer predefined questions Because of this, they can be used for knowledge discovery in addition to hypothesis testing

Data mining is used to:

! Provide insight into historical data

! Predict future values or outcomes based on historical patterns

! Close the analysis loop by taking action based on the information derived from the analysis

Topic Objective

To provide a definition of

data mining

Lead-in

Data mining provides a

means by which the system

deduces knowledge from

the data by identifying

correlations and other

patterns in the data

Data Mining Applications

! Advertising on the Internet

$ “What banner will I display to this visitor?”

$ “What other products is this customer likely to buy?

! Detecting Fraud

$ “Is this insurance claim a fraud?”

! Pricing Insurance

$ “How much of a discount will I offer to this customer?”

! Managing Credit Risk

$ “Will I approve the loan for this customer?”

Data mining techniques are used in a variety of applications This section provides some interesting examples

Advertising on the Internet

You can use data mining to classify groups of customers with similar information into segments for targeting advertising or special offers

Following are two Internet customer examples:

! An e-commerce Web site sells sporting equipment When a customer registers, a database management system collects information about the customer, such as gender, marital status, favorite sport, and age

By using data mining techniques, the Web site displays a masculine banner

ad with a golfing motif for the male, golf-loving, 40-year-old who returns to the Web site after registering

! When you purchase merchandise on the Internet, you are sometimes offered additional merchandise that the Web site predicts you might be interested in—for example, a book similar to the one you are currently purchasing Such recommendations are based on data mining techniques that search out purchase patterns of customers who purchased the same book you are now buying The system recommends: “If you like xyz books, check out the additional books below.”

Detecting Fraud

You can use a data mining system to identify characteristics of suspicious insurance claims by analyzing characteristics of legitimate and fraudulent claims For example, specific types of injuries that are difficult to diagnose, such as neck and back injuries, may be more likely candidates for a fraudulent claim

applications We are now

going to talk about some

common uses

Delivery Tips

Incorporate your own

examples of how data

mining is used to solve

business problems Ask

students for examples from

their businesses

Point out that data mining is

no longer an art used by just

PhDs This technology is

available and useful to a

variety of businesses

Pricing Insurance

In the insurance industry, you use data mining techniques to analyze historical data such as age, marital status, gender, and driving history All these factors play a role in predicting the likelihood of a specific driver for getting into an automobile accident Data mining techniques help you to weigh and factor these data points into pricing for an individual insurance policy

Managing Credit Risk

When you apply for a loan, the bank collects a broad range of information about you—for example, income, years of employment at a current job, marital status, and credit standing

By using data mining techniques applied to historical loan application information, the bank can predict whether you are a good or bad credit risk and can use this information when deciding on loan approval

Data Mining Models

! Analysis Services Models

Analysis Services Models

Analysis Services includes two data mining techniques—Microsoft Clustering and Microsoft Decision Trees

Clustering

You use the clustering technique, sometimes called K-nearest neighbor, to group data records that are similar to each other You often use this common technique as the starting point for market or customer analysis

For example, you may want to segment your market so that you can offer customized programs and pricing to specific customer groups With clustering, you can segment your customers into groups with similar characteristics

Decision Trees

Decision trees are a popular method for both classifying and predicting By using a series of questions and rules to categorize data cases, you can predict the likelihood of certain types of cases having a specific outcome

For example, insurance companies use a decision tree to predict the likelihood

of high claims by analyzing statistical data organized by a set of rules that help predict the likelihood of high claims

Topic Objective

To describe different data

mining models and how they

apply to data analysis

Lead-in

A variety of data mining

models are available These

techniques represent

different approaches to

classification and prediction

Delivery Tip

Do not spend much time

describing the different

models Simply discuss that

various models are available

for analysis and that

Microsoft provides two of

the models in Analysis

Services

Other Models

Analysis Services provides two types of data mining models—clustering and decision trees However, users may define their own models or use other proprietary data mining algorithms Common data mining models include market basket analysis, memory-based reasoning, and neural networks

Market Basket Analysis (Affinity Grouping)

Market basket analysis, sometimes called affinity grouping, is used for finding groups of items that occur frequently together in a single transaction

For example, customers who buy gin may also purchase tonic water, which is a frequent accompaniment Customers who buy potato chips frequently buy potato chip dips on the same shopping trip Understanding when products sell together helps a retail store manage placement of items on shelves to maximize affinity group purchases

Memory-Based Reasoning

Memory-based reasoning (MBR) is a directed data mining technique that is used for prediction and classification MBR analyzes a collection of the known instances of the nearest neighbor and from that information makes predictions about unknown instances

For example, if a patient exhibits a series of symptoms, doctors apply their experience with similar patients to diagnose the current case The doctors perform their diagnoses by using a form of MBR

Suppose you want to sell your car Several factors affect the sales price, such as the age of the car, its condition, its manufacturer and model, and so forth Analyzing historical car prices, the neural network can create a series of input and output factors to predict the sales price

Summary of Models

The following table defines commonly used data mining models and their typical usages

Market basket analysis (affinity grouping)

Clustering or affinity grouping

Introductory Example

Why Do High School Students Attend College?

A survey was conducted recently in the United States asking high-school seniors to answer the following five questions:

1 What is your gender?

2 What is your parents' income?

3 What is your IQ?

4 Do your parents encourage or not encourage you to go to college?

5 Do you plan to attend college?

Data from the survey was compiled into a table shown in the preceding illustration

Glancing at the table, you cannot easily determine how many students plan to attend college and how many do not You can see that roughly 50 percent will attend based on the first 22 records of this file This result may or may not be representative of the whole set of 9,000 cases

To determine how many students plan to attend college, you can execute a query that counts students grouped by those planning on attending and those not planning on attending

Suppose you are interested in determining the attribute or combination of attributes that have the highest potential of predicting the likelihood of a student for attending college This is a more complex question and involves segmenting the data based on various attributes you collect

To answer the question, you can spend several hours exploring the data manually, or you can use data mining to explore the data automatically

Topic Objective

To introduce an example of

how data mining can be

used for prediction

Lead-in

What do you think is the

principal attribute for

predicting whether students

attend college? What, if

anything, can you conclude

from the information in the

table?

Delivery Tips

Browse the actual relational

table data when discussing

the case study You can find

the CollegePlans table in

the Module 17 SQL Server

2000 database

Ask students what they think

are the most dominant

attributes that will predict

whether a student plans to

attend college

Demonstration: Determining Why Students Attend College

In this demonstration, you will create a data mining model by using a decision tree with relational data Specifically, you will create a decision tree that determines what causes students to attend college

Topic Objective

To demonstrate how to

create a data mining model

by using a decision tree with

relational data

Lead-in

In this demonstration, you

will learn how to create a

decision tree that

determines what causes

students to attend college

Delivery Tips

The steps for this

demonstration are included

in the Instructor Notes

Encourage students to

follow your demonstration

on their computers

Exploring the Decision Tree

Attend College:

33% Yes 67% No

All Students

Parental Encouragement?

Attend College :

57% Yes 43% No

Parents Encourage = Yes

Attend College:

6% Yes 94% No

Parents Encourage = No

Attend College:

74% Yes 26% No

Attend College:

29% Yes 71% No

High IQ Low IQ

Attend College :

18% Yes 82% No

Attend College :

9% Yes 91% No

Attend College:

4% Yes 96% No

The most dominant attribute is always the first rule in the decision tree

! Students who received encouragement from their parents had a 57.27 percent probability of planning to attend This is much higher than the general population Of the students who were encouraged by their parents:

• Those with an IQ higher than 110.25 had more than a 74 percent probability of attending college

• Those who also had parents with a high income were even more likely to attend college—77 percent

! Students who did not receive encouragement had a very low probability, 6.22 percent, of planning to attend Of the students who were not encouraged by their parents:

• Those students with a very high IQ had a higher probability than those with a lower IQ Of students with an IQ higher than 118.25, 17.96 percent plan to attend versus 3.52 percent of students with an IQ lower than 99.25

• Parental income had no impact on the likelihood of planning to attend college if the student were exceptionally smart with an IQ higher than 118.25

Topic Objective

To demonstrate how data

mining is applied by using a

decision tree

Lead-in

Looking at all the students

interviewed, roughly 33

percent plan to attend and

the remaining do not plan to

attend

Delivery Tips

After switching to the slide,

ask students the following

question: Of the collected

attributes, which do you

think is most likely to have

an impact on a student’s

decision to attend college?

Then use the build slide to

step through the results

Switch to Analysis Manager

to show the same results in

the Relational Mining Model

Editor

Tip

This example demonstrates that data mining allows you to validate or discredit specific hypothesis Data mining also helps you identify patterns that you may not expect or notice by analyzing the data manually

Training a Data Mining Model

Mining Model

DM Engine

Data

To Predict

DM Engine Predicted Data

To create a model, you must assemble a set of data where the attributes to be

predicted are known Such a data set is called the training data During the

training process, data is inserted into the data mining model The data mining model analyzes the training data and looks for rules and patterns that can be used later to determine the predictive columns

You perform training by processing the data mining model in Analysis Manager

The training data has two characteristics:

! It is typically historical data

! It is statistically representative of the cases for which you are building a predictive model

The case is the basic unit for analysis in the mining model The case is the

element that is used for classifying and grouping the data

As depicted in the preceding illustration, the data mining engine evaluates the cases identified in the training data and creates the model based on the algorithm selected When the model is built, it can be applied to future data to predict outcomes or classify data

Topic Objective

To explain the methodology

for creating a mining model

and to define terminology

Lead-in

When creating a data

mining model, you need a

training data set This is

typically historical data

where the attributes to be

predicted are known

Delivery Tip

Use the build slide to

explain how Analysis Server

evaluates training data to

build a data mining model,

and then uses the model to

predict future outcomes

based on new data sets

# Building a Data Mining Model with OLAP Data

! Introducing the Membership Card Scenario

! Selecting the Data Mining Technique

! Selecting the Case

! Selecting Predicted Entity

! Selecting Training Data

! Creating a Dimension and Virtual Cube

! Browsing the Data Mining Model

You can use the Mining Model Wizard in Analysis Manager to create a data mining model This section uses the Membership Card scenario to demonstrate the creation of a data mining model

Building and reviewing a data mining model entails several steps:

1 Selecting the data mining technique

2 Identifying the case

3 Selecting the entity to be predicted

4 Identifying the training data

5 Optionally creating a dimension and virtual cube from the resulting model

6 Processing the model and browsing the results

Topic Objective

To describe the steps used

to build a data mining model

with OLAP data

Lead-in

These are a variety of steps

involved in building a data

mining model with OLAP

data

Introducing the Membership Card Scenario

$ Identify opportunities for enhancing services at each current card level

$ Market programs based on customer demographics

$ Find membership card selection patterns

$ Select Customer as the mined dimension

$ Select the Member Card property as the pattern identifier

$ Use Customer demographics to train the model

$ Browse the decision tree

The Vice President of Marketing of Foodmart wants to evaluate current

member card programs To improve customer retention and satisfaction, she specifically wants to identify opportunities for enhancing services provided at each card level:

! Golden

! Silver

! Bronze

! Normal Demographic information about customers is available The information includes:

! Gender

! Marital status

! Yearly income

! Education level

In this card membership scenario, you will learn how historical data in the

Foodmart 2000 Sales cube predicts the likelihood of customers applying for

different levels of membership cards based on a variety of attributes

wants to evaluate the

current member card

programs

Delivery Tip

Use this example to

describe each of the

following pages in this

section