5 views

Uploaded by rohit sharma

- Ansys Advanced Analysis Techniques Guide
- ad07
- The CLUSTER Procedure
- Trade-Off Theory or Pecking Order Theory With a State-Ownership Structure- The Vietnam Case
- Guidelines for Collaborative Study Procedures to Validate Charasteristics for a Methhod of Analysis
- Data Mining and Business Intelligence
- PARAMETRICS AND OPTIMIZATION USING ANSOFT HFSS_10.1.1.169.1598[1]
- AQR05-MeanStandardDeviation
- Data Mining and Data Warehousing
- IJETTCS-2013-11-25-024
- Salmon Review
- Optimal Process Design Under Uncertainty
- pemodelan sistem
- Limit order
- An Income Strategy Approach to the Positive Theory of Accounting Standard Setting Choice
- data mining dan bigdata
- Salmon Review
- XERprojects
- An Improved Analysis of Data Mining Techniques on Medical Data
- Predictive Modelling Analytics through Data Mining

You are on page 1of 17

5

-2- 07.10.2004

Copyright

No part of this publication may be reproduced or transmitted in any form or for any purpose without the

express permission of SAP AG. The information contained herein may be changed without prior

notice.

Some software products marketed by SAP AG and its distributors contain proprietary software

components of other software vendors.

Microsoft, WINDOWS, NT, EXCEL, Word, PowerPoint and SQL Server are registered

trademarks of Microsoft Corporation.

IBM, DB2, DB2 Universal Database, OS/2, Parallel Sysplex, MVS/ESA, AIX, S/390,

AS/400, OS/390, OS/400, iSeries, pSeries, xSeries, zSeries, z/OS, AFP, Intelligent Miner,

WebSphere, Netfinity, Tivoli, Informix and Informix Dynamic ServerTM are trademarks of IBM

Corporation in USA and/or other countries.

UNIX, X/Open, OSF/1, and Motif are registered trademarks of the Open Group.

Citrix, the Citrix logo, ICA, Program Neighborhood, MetaFrame, WinFrame, VideoFrame,

MultiWin and other Citrix product names referenced herein are trademarks of Citrix Systems, Inc.

HTML, DHTML, XML, XHTML are trademarks or registered trademarks of W 3C, World Wide Web

Consortium, Massachusetts Institute of Technology.

JAVASCRIPT is a registered trademark of Sun Microsystems, Inc., used under license for

technology invented and implemented by Netscape.

SAP, SAP Logo, R/2, RIVA, R/3, SAP ArchiveLink, SAP Business Workflow, WebFlow,

SAP EarlyWatch, BAPI, SAPPHIRE, Management Cockpit, mySAP, mySAP.com, and other

SAP products and services mentioned herein as well as their respective logos are

trademarks or registered trademarks of SAP AG in Germany and in several other countries

all over the world. MarketSet and Enterprise Buyer are jointly owned trademarks of

SAP Markets and Commerce One. All other product and service names mentioned are the

trademarks of their respective owners.

Scoring BW 3.5

-3- 07.10.2004

Contents

Use and Applications ............................................................................................................................ 5

Typical Input........................................................................................................................................... 6

Typical Output........................................................................................................................................ 7

Data Mining Functions in the Analysis Process Designer ( APD)..................................................... 8

Settings for Weighted Score Tables .................................................................................................... 9

Model Fields for Weighted Score Tables ..................................................................................... 9

Field Parameters for Outlier Treatment ..................................................................................... 13

Field Parameters for Treating Missing Values .......................................................................... 14

Model Parameters for Weighted Score Tables .......................................................................... 15

Algorithm .............................................................................................................................................. 16

Recommendations for Weighted Score Tables ............................... Fehler! Textmarke nicht definiert.

Scoring BW 3.5

-4- 07.10.2004

A weighted score table is a method of evaluating alternatives when the importance of each criteria

differs. In a weighted score table, each alternative is given a score for each criteria. These scores are

then weighted by the importance of each criterion. All of an alternative's weighted scores are then

added together to calculate that alternative's total weighted score. The alternative with the highest total

score should be the best alternative

You can use weighted score tables to make predictions about future customer behavior. You create a

model in the data mining application to make predictions. After a model has been created based on

historical data, it can then be applied to new data to make prediction s. The prediction, that is, the

output of the model is called a Score. You can create a single score for your customers by taking into

account different dimensions.

SAPs Weighted Score Tables method allows you to define your own valuation function by first

assigning weights to the individual model fields and then creating a weighted total from these model

fields.

Scoring BW 3.5

-5- 07.10.2004

Weighted Score Table

A beverage outlet wants to attract the younger end of the market by introducing a product from a

higher price category into its product range. To determine potential customers, customer data

(including attributes like age, income, and drink expenditure) is valuated directly using the weighted

score tables. The age group 10-19 is valuated with 15, 20-29 with 10, 30-39 with 5, and so forth. The

customer incomes are valuated continuously by taking the respective figure as the value in each case

(that is, an income of 50,000 is valuated with 50,000). A weight is then assigned to each attribute: 2 for

age and 0.0001 for income. Thus, the score for a 25-year-old customer with an income of 40,000 is

calculated as follows: (2 x 10) + (0.0001 x 40,000) = 24.

Advertising Measures

You can use Weighted Score Tables to define a customer valuation that is dependent on the

characteristics and key figures of the customers. You can use valuation to support advertising

measures, to make service offers or rebates, or to select customers of interest for other purposes.

Scoring BW 3.5

-6- 07.10.2004

Typical Input

The following table contains data that could make up part of the typical input data for weighted score

tables. The customer id is the tables key, and state, salary and status are discrete fields. The

remaining fields are continuous.

Customer No of

ID State Salary Children Status

1 WA USD50K - USD70K 1 BRONZE

2 BC USD70K - USD90K 1 BRONZE

3 WA USD50K - USD70K 1 BRONZE

4 BC USD10K - USD30K 4 NORMAL

5 CA USD30K - USD50K 3 SILVER

6 WA USD70K - USD90K 3 BRONZE

7 OX USD30K - USD50K 2 BRONZE

8 DF USD50K - USD70K 2 BRONZE

9 BC USD10K - USD30K 5 NORMAL

10 OR USD30K - USD50K 4 GOLDEN

11 CA USD50K - USD70K 4 BRONZE

12 WA USD50K - USD70K 1 BRONZE

Scoring BW 3.5

-7- 07.10.2004

Typical Output

The weighted score tables results can be displayed in a tabular form showing the individual records

and their result values.

In the above example, the partial score for each model field, that is, Number of Children, State, Salary

and Status, is displayed.

Scoring BW 3.5

-8- 07.10.2004

Designer (APD)

The Analysis Process Designer (APD) is the application environmen t for the SAP data mining solution.

From SAP BW Release 3.5, data mining functions are fully integrated into the APD. You can perform

the following functions in the APD:

Training data mining models with SAP BW data (data mining model as data target in the analysis

process)

Execution of data mining methods such as prediction with decision tree, with cluster model and

integration of data mining models from third parties (data mining model as a transformation in the

analysis process)

Visualization of data mining models

For more information, see SAP Library at help.sap.com under SAP NetWeaver -> Release 04 ->

Information Integration -> SAP Business Information Warehouse -> BI Platform -> Analysis Process

Designer / Data Mining

Scoring BW 3.5

-9- 07.10.2004

The input data for SAPs Weighted Score Table is divided into two parts:

Model Fields

Model Parameters

1 2

Model fields are the attributes that define the object and the predictable field is the class label. In

Model Fields screen, you can add the fields that are required for creating decision trees. You must

define the content type for each model field.

It defines the data in a model field. There are 4 content types for model fields used in decision tree

classification.

Key field: The key field acts as the record identifier. This field does not have any influence on the

outcome.

Discrete: Also referred to as categorical, the data in the model fiel d for this content type contains a

finite set of values. For example, a model field Gender has two values - Male and Female. Attributes

like Color, Gender, and Status are examples of discrete attributes.

Continuous: Continuous data can have any value in an interval of real numbers. This implies that the

value does not have to be an integer. Attributes having infinite set of possible real values are called

Continuous. Typically, they have a Minimum and Maximum value and attribute values could be

anything within this interval. Attributes like Salary, Sales Revenue, Quantity sold etc are examples of

Continuous attributes. You can discretize a Continuous attribute by defining fixed intervals. For

example, if the salary ranges from $100 to $20000, then we can form intervals like $0 2000, $2000

$4000, $4000 $6000. $18000 $20000. An attribute value will fall into any one of these intervals.

You can specify the parameter values for each model field, except t he Key Field. You must specify the

partial weights for the model fields depending on the type of the model field.

Scoring BW 3.5

- 10 - 07.10.2004

Discrete: For model fields of type discrete, you specify the individual values of the field. As described

below, you can define a common partial weight for some of the remaining values. This weight is

applied only if you have set the Treat as separate instance indicator in the Outlier treatment tab page.

For more detailed information about handling outliers, see the section Parameters of the Model Fields

for Treating Outliers.

For a model field of type Discrete, the field parameters are as follows:

Weight of Model Field: You can define the weight of a model field.

Value: You enter the values for the model field in the column.

Partial Score: For each model field value that you enter, you must specify a model field weight. The

weights for the model fields determine the share of partial weights that the score has.

Partial Score for remaining values: You can define a single weight for all the remaining values in the

dataset.

The score is calculated as follows:

Score (Field1, Field2 ...) = Weight1 x Partial Weight1 (Field1) + Weight2 x Partial Weight2 (Field2) + ...

Scoring BW 3.5

- 11 - 07.10.2004

In the case of continuous model fields, you specify partial weights for individual threshold values. You

must also specify how to deal with values of partial weights between the threshold values. You have

the following options:

Function is piecewise constant: You have to specify the function of the partial weights between the

threshold values. Check this option to specify the function of the partial weight as piecewise constant.

If you check this option, then the function is constant between each pair of threshold values . That is,

the partial weight of the left threshold and of the right threshold is considered for each setting.

Alternatively, if you do not set the function is piecewise constant indicator, linear interpolation is

applied to calculate the partial weight continuously between each pair of threshold values.

In each case, you have to specify at least two threshold values because the value range used for

outlier treatment lies above the largest threshold and beneath the smallest threshold.

The score value to be defined is dependent on the discrete model field Status and the

continuous model field Salary. The weighting of these two model fields should be 3 and 1

respectively. In the model field Status, the data to be processed takes the values gold,

silver, bronze, copper, and iron. The following partial weightings could then be specified:

Value Partial Weighting

Scoring BW 3.5

- 12 - 07.10.2004

Gold 10

Silver 6

Bronze 4

The partial weighting 2 can be assigned to the remaining values. For the model field

Salary, the threshold values and corresponding partial weightings could be assigned as

follows:

Threshold Value Partial Weighting

0 0

10 000 10

25 000 20

50 000 30

The partial weightings function should be piecewise constant and take the partial

weighting of the left threshold value in the interval between two threshold values. In this

way, the score value (silver, 40 000) = 3 x 6 + 1 x 20 = 38 is obtained. If the partial

weightings function for the income should be continuous instead of piecewise constant,

then it produces the score value (silver, 40 000) = 3 x 6 + 1 x 26 = 44.

If the Treat as separate instance option was selected in outlier handling for the model field

Status, then the function produces the score value (iron, 10 000) = 3 x 2 + 1 x 10 = 16.

If the Constant extrapolation option was chosen in outlier handling for the model field

Income, then the function produces the score value (silver, 60 000) = 3 x 6 + 1 x 30 = 48.

If the Extrapolation option is chosen, this produces the score value (silver, 60 000) = 3 x 6

+ 1 x 34 = 52.

For more detailed information about how to treat outliers, see the section Parameters of the Model

Fields for Treating Outliers.

Scoring BW 3.5

- 13 - 07.10.2004

The parameters for the model fields offer control options for treating outliers.

Which values are treated as outliers?

For discrete model fields, outliers are values that do not belong to the values specified explicitly or to

the most frequently occurring values.

For continuous model fields, outliers are values falling outside of the outer borders that are

determined during the definition of the value ranges, either explicitly or automatically.

You can make an outlier treatment setting to decide whether processing is stopped, the record is

ignored, or the default score is set when a record occurs containing an outlier.

For continuous model fields, you can specify that an extrapolation is applied.

Extrapolation: Whenever a linear regression model is fit to a group of data, the range of the data

should be carefully specified. Using the regression method, it is possible to make predictions for

values outside this specified range of data. This process is known as extrapolation. If you choose the

option Extrapolation, the outliers, that is, the values lying outside the specified range of values will not

be treated separately.

Constant Extrapolation: If you select this option, then the function is constantly extrapolated beyond

the external borders.

Scoring BW 3.5

- 14 - 07.10.2004

Treat as separate instance: This option is valid only for discrete model fields. By setting this option,

all outliers are treated as a single, common remainder.

The parameters for the model fields offer control options for handling missing values.

Scoring BW 3.5

- 15 - 07.10.2004

For treating missing values, you first have to set the appropriate indicator and identify a missing value.

If, for example, the size of family is denoted by a numeric value and NA has been used to denote a

value that is unknown, you can enter NA as the Missing Value. You define a separate treatment for

this value accordingly.

You can make a setting to decide whether processing is stopped, the record is ignored, or the default

score is set when a value defined in this way occurs. Using the option Replace by value, you can

substitute the missing value with another value.

Default Score

You use this parameter to specify a default output value for weighted score tables. If required, this

value is always set whenever a record does not fulfill certain conditions (for example, it has missing

data or outliers). The default value for this field is 0 (zero).

Scoring BW 3.5

- 16 - 07.10.2004

Algorithm

Weighted Score Tables

A function f that is defined by weighted score tables is a linear combination of functions of a variable.

f ( x1 ,..., x n ) = w1 f 1 ( x1 ) + ... + wn f n ( x n )

The weights w1 ,..., wn are arbitrary numbers. Each of the functions f1 ,..., f n is mapped to exactly one

model field. The arguments x1 ,..., xn of these functions are those values that the model fields can

take.

For discrete model fields, the score table of the model field is used to directly assign a function value

f i ( xi ) to individual values xi of the model field. A common function value can be assigned to values

that are not listed explicitly in the table.

For continuous model fields, the score table of the model field is also used to d irectly assign a function

value xi to individual values f i ( xi ) of the model field. Either a linear interpolation is made between

two points, or the function value from the left or right point is taken. Respectively, either a polygon line

or a piecewise constant function is defined. Depending on the option selected by the user, the function

is continued as linear or continuous beyond the outer points.

Let us assume that you would like to valuate your customer data on the basis of the fields Occupation

and Age. For this, you could define a weighted score table function as follows:

Score (Occupation, Age) = w 1 x ps1(Occupation) + w2 x ps2 (Age)

w1 and w2 stand for the weights you give the two fields, such as:

ps1 and ps2 stand for the functions with which you define partial scores for both fields, as in the

following table:

Group Group)

Employee 5

Civil servant 7

Self-employed 10

Other 2

0 0

20 6

Scoring BW 3.5

- 17 - 07.10.2004

30 10

50 4

65 2

Ages falling between those specified above should be interpolated. This then gives you, for example:

Score (Employee,25) = 2 x ps 1(Employee) + 5 x ps 2(25) = 2 x 5 + 5 x 8 = 50

Start with small models containing few model fields

For continuous model fields, you can choose to perform valuations as continuous or in stages. If

you opt for valuation in stages, you must specify whether the stages you choose contain the left

interval border or the right interval border.

Do not use any model fields with the weight 0. Remove all such fields from the model.

The structure on which the function is based, is defined as the weighted sum of the functions in

one individual model field. This produces certain dependencies. If, for example, the function f is to

be defined for gender and age, then male and female customers have profiles that only differ in

one additive constant. If f(male,20)=15 and f(female,20)=10, then the following applies for every

age x f(male, x)f(female, x) = f(male, 20)f(female, 20) = 5. If f(male, 50)=20, then it also applies

that f(female, 50)=15.

Scoring BW 3.5

- Ansys Advanced Analysis Techniques GuideUploaded bySimion Laurentiu
- ad07Uploaded byJuni Ardi
- The CLUSTER ProcedureUploaded byHefu Liu
- Trade-Off Theory or Pecking Order Theory With a State-Ownership Structure- The Vietnam CaseUploaded byangelcomputer2
- Guidelines for Collaborative Study Procedures to Validate Charasteristics for a Methhod of AnalysisUploaded byCarlos Roque
- Data Mining and Business IntelligenceUploaded byDreamtech Press
- PARAMETRICS AND OPTIMIZATION USING ANSOFT HFSS_10.1.1.169.1598[1]Uploaded bydhfsi7496
- AQR05-MeanStandardDeviationUploaded bysoharoma
- Data Mining and Data WarehousingUploaded bymycatalysts
- IJETTCS-2013-11-25-024Uploaded byAnonymous vQrJlEN
- Salmon ReviewUploaded byThe Oregonian
- Optimal Process Design Under UncertaintyUploaded byLads
- pemodelan sistemUploaded bySyaiful Mansyur
- Limit orderUploaded byOrlando Santa Olalla
- An Income Strategy Approach to the Positive Theory of Accounting Standard Setting ChoiceUploaded byZhang Peilin
- data mining dan bigdataUploaded byMuhammad Iman Santoso
- Salmon ReviewUploaded byThe Oregonian
- XERprojectsUploaded bybobmailestrada5679
- An Improved Analysis of Data Mining Techniques on Medical DataUploaded byDaniel Das
- Predictive Modelling Analytics through Data MiningUploaded byIRJET Journal
- GM150 Listenhandbuch LH3 0112 EngUploaded bysandeep_bcpl
- IPC2012-90491Uploaded byMarcelo Varejão Casarin
- A Brief Study of Data Mining Techniques to Study Patterns in Medical SciencesUploaded byIRJET Journal
- quiz- ksa3Uploaded byapi-265310722
- Softimage User Guide_ Sampling and Bracketing ParametersUploaded byJeremy George
- EdTech 503 Part 6 and 7 Formal EvaluationUploaded byAngie Kruzich
- Tabulation Coding and EditingUploaded byKritika Jaiswal
- Articulo Lab Saneamiento.pdfUploaded byGina Cote
- Study of different Data Mining system & PlatformUploaded byIRJET Journal
- Week-02.pdfUploaded bynadeemajeedch

- 869-plain-variant-configuration.pdfUploaded byrohit sharma
- 755-optimizing-ibm-netfinity-servers-for-sap-r3-and-windows-nt.pdfUploaded byrohit sharma
- 993-sap-bw-and-hr-human-resources.docUploaded byrohit sharma
- 955-universal-data-access-via-sap-bw-java-integration-functionality.pdfUploaded byrohit sharma
- c EssentialsUploaded byREXTERYX
- Essential PerlUploaded bydogstory
- 957-xml-analysis-for-sap-bw.pdfUploaded byrohit sharma
- 948-handling-balance-sheets-and-pal-statements-in-sap-bw-30.pdfUploaded byrohit sharma
- 865-sap-mm-t-codes-by-business-roles.pdfUploaded byrohit sharma
- 893-introduction-to-webobjects-j2ee-programming-guide.pdfUploaded byrohit sharma
- C++ Quick ReferenceUploaded byGlenn Fabia
- 874-step-by-step-procedure-to-schedule-a-background-job-in-sap.pdfUploaded byjarmsj
- An Introduction to the Java EE PlatformUploaded byapi-27156242
- 840-crm-analytics-analysis-process-designer-apd.pptUploaded byrohit sharma
- Pointers and MemoryUploaded byblack Snow
- Unix Programming ToolsUploaded byblack Snow
- Tree List Recurs IonUploaded byblack Snow
- 872-how-to-update-sap-kernal-patches.pdfUploaded byrohit sharma
- LinkedList ProblemsUploaded byGowreesan
- Linked List BasicsUploaded bygbland
- C EssentialUploaded byVijay kumar Gupta .C
- Binary TreesUploaded byblack Snow
- 863-ias-prelims-2008-general-studies-question-paper.pdfUploaded byrohit sharma
- 870-sap-pp-production-planning-flow.pdfUploaded byrohit sharma
- 868-personal-administration-infotypes.pdfUploaded byrohit sharma
- mocktestpaper2eceUploaded byKiran Bhawsar
- 867-payroll-year-end-preparation-checklist.pdfUploaded byrohit sharma
- 864-material-master-field-selection-attribute.pdfUploaded byrohit sharma
- 795-a-z-guide-to-sap-bi-netweaver-project-part-2.pptUploaded byrohit sharma
- General Studies Pre 2008 Ans SheetUploaded byJan Abi

- Soil InvestigationUploaded byMohan Prasad
- (Ceramic Engineering and Science Proceedings) -Applications of Refractories_ Ceramic Engineering and Science Proceedings, Volume 7, Issue 1_2-The American Ceramic Society, Inc. (1986)Uploaded byJuan Bustamante
- Seminar ReportUploaded bySanth Blr
- Deriving Rainfall Intensity-duration-frequency RelUploaded byOpata Opata
- Report Project 1.docxUploaded byNurul Nadiah
- 07A70503-DATAWAREHOUSINGANDDATAMININGUploaded bySravani Sravz
- 1-s2.0-S0921800916307613-main.pdfUploaded byrobert
- Tutorial_on_PT_Stats__(03-12.pdfUploaded byWahyu Pradana Arsitika
- Isatis Case Studies Oil GasUploaded byrenzoquim
- Chapter 1 4811 Fund ConceptsUploaded byJamriKallun
- Normality Testing in ExcelUploaded byAlina Constantin
- An Efficient Detection of Outliers and Hubs Using Minimum Spanning TreeUploaded byJournal of Computing
- CO1607 Final ReportUploaded byDiany M. Murillo Gomez
- VW10130_MFU_englisch_01.pdfUploaded byviniciusafonso
- Islamic Banks and Financial Stability: An Empirical AnalysisUploaded byabduljeli
- AP Chemistry Summer Assignment 2017Uploaded byJohn Smith
- Statistics for Data Analysis Lec 2 Measure of Center TendancyUploaded byNikesh Bajaj
- Asheesh PaperUploaded bymy_khan20027195
- 0f317536ed3e376505000000Uploaded byAyush53
- Analytical ChemistryUploaded byS J
- math project reportUploaded byapi-224466708
- #Academic Journal Quality GuideUploaded byΑγγέλα Καλότυχου
- finalpaper-firstsubmission seniorUploaded byapi-310605560
- 135147 18383 EGM2008validation SriLanka RevisedUploaded byMahesh Chathurange
- Creswell Mixed MethodsUploaded bySaif Ullah Qureshi
- AP Statistics Problems #13Uploaded byldlewis
- Data Camp - Cleaning DataUploaded byRahulChaudhary
- Group 8 - Business Stats Project - Installment IUploaded byAbhee Raj
- Productivity and Unit Labor Cost in IndiaUploaded byAshish Kumar
- Fault Detection of Fuel InjectorsUploaded byMaikol Pupiales