Moss L.T., Atre S. Business intelligence roadmap: The complete project lifecycle for decision-support applications

Подождите немного. Документ загружается.

Retest the application

programs and scripts

from beginning to end

until they perform as

expected

Performance test those

programs that have

many JOINs, require

complicated calculations,

and read high-volume

tables

Use full-volume data for

performance testing

If using a stress test

simulation tool, define

the test components

(programs, databases,

tables, volumes, and so

on) to the simulation tool

and run a simulation test

before testing with full-

volume data

Move databases,

programs, and scripts

into the QA environment

QA test the entire

application from

beginning to end with the

operations staff

Obtain approval from the

operations staff to move

the application programs

into production

Acceptance test the

entire application from

beginning to end with the

subject matter expert

and business

representative

Obtain certification for

the application from the

business representative

北斗成功社区 BeiDouWeb.com 教育音视频/电子书/实用资料文档/励志音乐影视仅供免费试用/版权原著所有

541/631

5. Provide data access and

analysis training

Identify help desk staff to

be trained

Identify "power users" or

other business liaison

personnel to be trained

Identify business people

to be trained

Create training

materials:

- Presentation

slides and

instructor notes

- Student

workbooks with

exercises

- Exercise solutions

and other pertinent

handouts

Schedule training

sessions

Conduct training sessions

Measure training

effectiveness

13. Data Mining

1. State the business

problem

Define the business

problem

Obtain commitment for a

data mining solution

Set realistic expectations

for the data mining tool

Identify preliminary

algorithms relevant to

the business problem

北斗成功社区 BeiDouWeb.com 教育音视频/电子书/实用资料文档/励志音乐影视仅供免费试用/版权原著所有

542/631

2. Collect the data

Identify available data

sources (operational as

well as BI)

Extract pertinent data

from various internal

data sources

Acquire (purchase)

pertinent data from

external data sources

3. Consolidate and cleanse

the data

Merge data from various

internal data sources

Match and merge

internal data with

external data

Review the structure of

the merged data

Select a sample of data

for each analytical data

model

Select related meta data

from the meta data

repository

Review the data domains

(content) and measure

the quality and

reasonability of the data

values

Validate domain

reasonability across

active variables

4. Prepare the data

Review the frequency

distribution of categorical

variables

Review maximum,

minimum, mean, mode,

and median for

quantitative variables

Use statistical

distribution parameters

to filter noise in the data

Eliminate variables with

missing values or replace

北斗成功社区 BeiDouWeb.com 教育音视频/电子书/实用资料文档/励志音乐影视仅供免费试用/版权原著所有

543/631

the missing values with

"most likely" values

Convert data formats to

suit the particular data

mining algorithm used

Derive new variables

from original input data,

where appropriate

Consolidate customers by

assigning a household

number to related

customers

Relate customers with

products and services

Apply data reduction,

where appropriate

Apply data mining

transformation

techniques, where

appropriate:

- "Discretization"

technique to

convert

quantitative

variables into

categorical

variables

- "One-of-N"

technique to

convert a

categorical variable

to a numeric

representation for

input to a neural

network

北斗成功社区 BeiDouWeb.com 教育音视频/电子书/实用资料文档/励志音乐影视仅供免费试用/版权原著所有

544/631

5. Build the analytical data

model

Create the analytical

(informational) data

model

Select data mining

operations with the

appropriate algorithms

Test accuracy of the

analytical data model

using confusion matrices

and input sensitivity

analyses

Repeat prior steps (if

necessary) to train and

retrain the model

(beware of overtraining

the model)

6. Interpret the data

mining results

Note

This is an

ongoing activity.

Review the data mining

results

Look for results that are

interesting, valid, and

actionable

Present the new findings

in a convincing,

business-oriented way

using visualization

technology

Formulate ways in which

the new information can

best be exploited

7. Perform external

validation of the results

Note

This is an

ongoing activity.

Compare data mining

results to published

industry statistics

Validate the selection of

variables and time frame

of your data against the

variables and time frame

of the industry statistics

Identify the variations

between your analysis

results and the industry

statistics

Determine the reasons

for the variations

北斗成功社区 BeiDouWeb.com 教育音视频/电子书/实用资料文档/励志音乐影视仅供免费试用/版权原著所有

545/631

8. Monitor the analytical

data model over time

Note

This is an

ongoing activity.

Keep validating your

analytical data model

against industry statistics

at regular time intervals

When industry statistics

change, change your

analytical data model

and retrain it

Research the data mining

capabilities of your

competitors

Monitor your

competitors' market

share and adjust your

analytical data model

accordingly

14. Meta Data

Repository

Development

1. Build the meta data

repository database

Run the DDL to create

the physical meta data

repository database

structures

Run the DCL to grant

CRUD authority on the

meta data repository

database structures

If licensing a meta data

repository product, set

up CRUD authority on

the meta data repository

product

Test all meta data

repository product

components, especially

the meta data repository

database

2. Build and unit test the

meta data migration

process

Code the tool interface

programs or use the

export facility of the

various tools; (for

example, ETL tool, CASE

tool)

Code the meta data

transformation programs

Code the meta data load

programs or use the

import facility of the

北斗成功社区 BeiDouWeb.com 教育音视频/电子书/实用资料文档/励志音乐影视仅供免费试用/版权原著所有

546/631

meta data repository

product or the DBMS

load utility

Code the meta data

programs that will run

during the ETL process to

capture:

- Load statistics

- Reconciliation

totals (record

counts, domain

counts, amount

counts)

- Data-cleansing

(reliability) metrics

- Data rejection

counts and reasons

for rejections

Unit test the meta data

migration programs (or

meta data repository

product modules):

- Tool interface

programs

- Meta data

transformation

programs

- Meta data load

programs

Unit test the meta data

programs that will run

during the ETL process

北斗成功社区 BeiDouWeb.com 教育音视频/电子书/实用资料文档/励志音乐影视仅供免费试用/版权原著所有

547/631

3. Build and unit test the

meta data application

Code the access interface

programs (GUI or Web

front end)

Code the meta data

report programs

Code the meta data

query scripts

Code the meta data

repository online help

function programs

Unit test the meta data

application programs (or

meta data repository

product modules):

- Access interface

programs

- Report programs

- Query scripts

- Online help

function programs

4. Test the meta data

repository programs or

product functions

Create a test plan with

test cases for:

- Meta data

migration process

- Meta data

repository

application

programs or

product modules

- Meta data

programs that run

during the ETL

process

Create test data for meta

data repository testing:

- Meta data

migration process

- Meta data

北斗成功社区 BeiDouWeb.com 教育音视频/电子书/实用资料文档/励志音乐影视仅供免费试用/版权原著所有

548/631

repository

application or

product modules

- Meta data

programs that run

during the ETL

process

Integration or regression

test the meta data

repository:

- Meta data

migration process

- Meta data

repository

application or

product modules

- Meta data

programs that run

during the ETL

process

Log the actual test

results and document

any test issues

Compare actual test

results with expected test

results

Revise the meta data

repository programs

Retest the meta data

repository programs

from beginning to end

until they perform as

expected

Conduct QA testing with

operations staff

Conduct acceptance

testing with the subject

matter expert and the

business representative

(QA and acceptance

testing may be

conducted at the same

time)

北斗成功社区 BeiDouWeb.com 教育音视频/电子书/实用资料文档/励志音乐影视仅供免费试用/版权原著所有

549/631

5. Prepare the meta data

repository for production

Install and test the

server platform for the

production meta data

repository

Create DDL and DCL for

the production meta data

repository database

Write operating

procedures for the

operations staff with

instructions for running

meta data repository

reports at predetermined

dates and times

Write a reference guide

for the help desk staff

and the business people

with instructions on how

to use the meta data

repository

Develop performance

monitoring and tuning

procedures for the meta

data repository database

Develop meta data

repository usage

monitoring procedures

6. Provide meta data

repository training

Identify help desk staff to

be trained on the content

and use of the meta data

repository

Identify "power users" or

other business liaison

personnel to be trained

Identify business people

to be trained

Create meta data

repository training

materials:

- Presentation

slides and

instructor notes

- Student

北斗成功社区 BeiDouWeb.com 教育音视频/电子书/实用资料文档/励志音乐影视仅供免费试用/版权原著所有

550/631