Decision Modelling

Barbara  Re

Dr.   Barbara  Re

[email protected] 1
Decision Logic and Decision Task

(Ross  2011,  p.  152f)

Decision  Modeling 2
Modeling   Decision Logic

• There are two well-­‐known modeling notations for Decision Logic

• The  Decision Model
• Based on  a  book from Barbara  von  Halle  and Larry  Goldberg
• Decision Model  and Notation  DMN
• A  new standard from OMG

Decision  Modeling 3
Decision Model   and Notation   (DMN)
Decision  Modeling 4
 Two  Different   Perspectives

• Decision-­‐making  is  addressed  from  two  different  perspectives  by  existing  

modeling standards:  
• Business  process  models  (e.g.,  BPMN)   can  describe  the  coordination  of  decision-­‐
making  within  business  processes  by  defining  specific  tasks  or  activities  within  
which  the  decision-­‐making  takes  place
• Decision  logic  (e.g.,  PRR,  PMML)   can  define  the  specific  logic  used  to  make  
individual  decisions,  for  example  as  business  rules,  decision  tables,  or  executable  
analytic  models.

5
 Decision   Requirements   Diagram

DMN  will  provide  a  third  perspective  – the  Decision  Requirements  

Diagram – forming  a  bridge  between  business  process  models  and  
decision  logic  models:  
• Business  process  models  will  define  tasks  within  business  processes  
where  decision-­‐making  is  required  to  occur
• Decision  Requirements  Diagrams  will  define  the  decisions  to  be  made  in  
those  tasks,  their  interrelationships,  and  their  requirements  for  decision  
logic
• Decision  logic  will  define  the  required  decisions  in  sufficient  detail  to  
allow  validation  and/or  automation

6
Aspects  of  Modelling

7
Decision Model  and Notation   (DMN)
• The  Decision Model  and Notation  is a  
new standard from the OMG
• It is currently published in  its version 1.0
• Purpose  of  DMN:  provide  the  constructs  
that  are  needed  to  model  decision,  so  
that  organizational  decision-­‐making  can  
be
• readily  depicted  in  diagrams
• accurately  defined  by  business  
analysts
• (optionaly)  automated

Decision  Modeling 8
 DMN
It  defines  the  concept  of  a  Decision  
Requirements  Graph  (DRG)  
comprising  a  set  of  elements  and  
their  connection  rules,  and  a  
corresponding  notation:  the  Decision  
Requirements  Diagram  (DRD)

It  defines  Decision  Logic  via  

Value  expression  and/or  
decision  table

It  provides  a  language  called  

FEEL  for  defining  and  
assembling  decision  tables,  
calculations,  if/then/else  logic,  
simple  data  structures,  and  
externally  defined  logic  from  
Java  and  PMML  into  executable  
expressions  with  formally  
defined  semantic

9
 Scope  and  Uses  of  DMN

• Decision  modeling is  carried  out  by  business  analysts  in  order  to  
understand  and  define  the  decisions  used  in  a  business  or  organization.  
• Such  decisions  are  typically  operational  decisions  made  in  day-­‐to-­‐day  
business  processes,  rather  than  the  strategic  decision-­‐making  for  which  
fewer  rules  and  representations  exist.  
• Three  uses  of  DMN  can  be  discerned  in  this  context:  
• For  modeling human  decision-­‐making  
• For  modeling the  requirements  for  automated  decision-­‐making  
• For  implementing  automated  decision-­‐making  

10
 Main  Concepts of DMN

Business Process
More  details

Decision Requirements

Decision Logic

(Coenen  2013)
Decision  Modeling 11
 Basic  Concepts  – Decision   Requirements  
Level
• A  decision  is  the  act  of  determining  an  output  value  (the  chosen  option),  from  a  
number  of  input  values,   using  logic  defining  how  the  output  is  determined   from  the  
inputs  
• Decision  logic  may  include  one  or  more  business   knowledge  models  which  
encapsulate  business  know-­‐how  
• A  decision  may  require  multiple  business  knowledge  models,  and  a  business  
knowledge  model  may  require   multiple  other  business  knowledge  models

Decision  Modeling 12
 Basic  Concepts  – Decision   Requirements  
Level
• Decisions  can  be  decomposed  into  sub-­‐decisions.   Top  level  decisions  can  be  thought  
of  as  selecting  an  answer  from  a  range  of  possible  answers.  Lower  level  decisions  
often  will  simply  provide  input  to  other  decisions
• Decisions  may  therefore  be  connected  in  a  network  called  a  Decision  Requirements  
Graph  (DRG),  which  may  be  drawn  as  a  Decision  Requirements  Diagram  (DRD)

13
 Basic  Concepts  – Decision   Requirements  
Level
• A  decision  may  require  multiple  business  knowledge  models,  and  a  
business  knowledge  model  may  require  multiple  other  business  
knowledge  models

14
 Basic  Concepts  – Decision   Logic  Level

• Using  decision  logic,  the  same  components  described  at  decision  

requirements  level  may  be  specified  in  greater  detail,  to  capture  :
• a  complete  set  of  business  rules  and  calculations
• (if  desired)  to  allow  the  decision-­‐making   to  be  fully  automated

15
 Decision   and  corresponding   value  expression  
• At  the  decision  logic  level,  every  decision  in  a  DRG  is  defined  using  a  value  
expression  which  specifies   how  the  decision’s  output  is  determined   from  its  
inputs
• The  decision  is  considered   to  be  the  evaluation  of  the  expression
• The  value  expression   may  be  notated  using  a  boxed  expression

16
 Business   knowledge   model  and  
corresponding   value  expression  
• At  the  decision  logic  level,  a  business  knowledge  model  is  defined  using  
a  value  expression  that  specifies  how  an  output  is  determined  from  a  set  
of  inputs
• Value  expressions  may  be  encapsulated  as  functions,  which  may  be  
invoked  from  decisions’  value  expressions

17
 Business   knowledge   model  and  
corresponding   decision   table  
• A  business  knowledge  model  may  contain  any  decision  logic  which  is  capable  of  being  represented  as  a  
function
• This  will  allow  the  import  of  many  existing  decision  logic  modeling standards  (e.g.,  for  business  rules  and  
analytic  models)  into  DMN
• An  important  format  of  business  knowledge,  specifically  supported  in  DMN,  is  the  Decision  Table
• Such  a  business  knowledge  model  may  be  notated  using  a  Decision  Table

18
 Main  concepts  – Decision  Requirements   Level
• Business   concepts  only
• Business   decisions
• Areas   of  business   knowledge
• Sources   of  business   knowledge Decision Requirements

(Coenen  2013)
Decision  Modeling 19
DRD  Elements

Decision  Modeling 20
Constructs   of  a  Decision   Requirements   Model

Decision  Modeling 21
 Decision
• A  decision  is  the  act  of  determining  an  output  value  (the  chosen  option),  from  a  
number  of  input  values,   using  logic  defining  how  the  output  is  determined   from  the  
inputs  
• Two  properties   should  be  captured  for  every  decision:
• Question:  A  natural  language  statement  that  represents   the  decision  in  the  form  
of  a  question.  This  should  be  specific  and  detailed.
• Allowed  Answers:  A  natural  language  description   of  the  possible  answers  to  this  
question.  
• For  action-­‐oriented   decisions,  the  allowed  answers  represent   the  responses   that  
the  process  must  handle  when  the  decision  model  is  invoked  by  a  business   rule  task

Decision  Modeling 22
 Business   Knowledge   Models

• Business  knowledge  models  represent  functions  that  encapsulate  

reusable  decision  making  logic.  The  logic  they  encapsulate  might  be  a  set  
of  business  rules,  a  decision  tree,  a  decision  table,  or  an  analytic  model.
• The  specifics  of  knowledge  representation  involved  need  not  be  
displayed  on  the  diagram  but  could  be.
• The  decision  logic  that  can  be  specified  in  a  business  knowledge  model  
can  also  be  linked  directly  to  a  decision,  but  encapsulating  it  in  a  
business  knowledge  model  allows  it  to  be  reused,  parameterized  and  
displayed  on  a  Decision  Requirements  Diagram

Decision  Modeling 23
 Input  Data

• Decisions  require  inputs,  and  many  of  these  are  input  data,  which  is  data  
that  is  input  to  the  decision  making  from  outside  the  decision  context.  
• Input  data  elements  typically  represent  business  entities  that  are  being  
used  in  the  decision  making,  such  as  Policy  or  Customer.  However,  
sometimes  they  can  represent  any  information  element  at  any  level  of  
detail.  
• Each  input  data  element  can  be  described  in  terms  of  a  hierarchical  
information  model  that  specifies  exactly  what  information  elements  
comprise  the  input  data.  

Decision  Modeling 24
Decision   with  the   Listed  Input  Data  option  

25
 Knowledge   Source

• Knowledge  sources  represent  the  source  of  know-­‐how  for  making  a  

decision
• This  could  be:
• Regulations   or  policies  about  how  a  decision  must  be  made,  best  practices  or  
expertise  on  how  it  should  be  made
• Analytic   knowledge  on  how  it  might  be  made  more  accurate

Decision  Modeling 26
 Knowledge   Sources  representing   authorities  

• Knowledge  sources  are  the  authorities  for  a  decision  and  typically  refer  
to  some  external  document  or  source  that  contains  detailed  guidance

27
 Knowledge   source  representing   predictive  
analytics  
• They  may  be  drawn  from  Input  Data  and Decisions  to  Knowledge  
Sources,  where,  in  conjunction  with  use  (a),  they  represent  the  
derivation  of  Business  Knowledge  Models  from  instances  of  Input  Data  
and  Decision  results,  using  analytics

28
DRD  Requirements

Decision  Modeling 29
 Elements  and  Allowed  Relationships  of  the  
Requirements  Graph
Input data Decision
Decision 1 Decision 2
Input  data  is  used  as  input  for  decision
Decision   1  is  used  as  input  for  decision   2

Knowledge
Knowledge Input data Source
Decision Source
Input  data  d epends  o n  Knowledge   Source
Decision   depends  o n  Knowledge   Source
Knowledge
Source Decision
Business Decision
Knowledge Knowledge   Source   depends  o n  Decision

Business   Knowledge   invokes a  Decision

Knowledge Business
Source Knowledge
Business Business
Knowledge 1 Knowledge 2 Knowledge   Source   depends  o n  Business   Knowledge

Business   Knowledge   1  invokes Business   Knowledge   2 Knowledge Knowledge

Source 1 Source 2

(Coenen  2013) Knowledge   Source1   depends  o n  Knowledge   Source2

Decision  Modeling 30
Decision  Modeling 31
 Main  concepts  – Decision  Logic
• Greater  detail
• Business  rules
• Calculations Decision Logic
• Automated
• Display

(Coenen  2013)
Decision  Modeling 32
 Modeling   Decision   Logic

Text/SBVR

Supported  by  DMN

“This will allow the import of many existing decision logic modeling
standards (e.g. for business rules and analytic models) into DMN”
(Coenen  2013)
Decision  Modeling 33
Decision  Modeling 34
Boxed  Expression
The  notation  for  decision  logic  is  boxed  expressions which  decompose  the  decision  
logic  model  into  small  pieces  that  are  associated  with  elements  of  Decision  
Requirements  Diagram

Decision  Modeling 35
Structure  of  a  Decision   Table  in  DMN

Decision  Modeling 36
Decision   Tables

Decision  Modeling 37
Hit  Policies  (1)
• The   hit  policy   specifies  what  the  result   of  the  decision  table  is,  if   there  are  
multiple  matches  for  a  given   set  of  inputs.  
• The   hit  policy   indication   is  mandatory   and  is   summarized  using   a  single  
character  in   a  particular  decision  table  cell.

Single  Hit  Policies:

Decision  Modeling 38
Hit  Policies  (2)

Multiple  Hits  Policies  for  Single  Output

Aggregation  for  Multiple  Hits  Policy

Decision  Modeling 39
FEEL  =  Friendly  Enough   Expression   Language  
FEEL  is  a  script  language  for  decision  tables

Decision  Modeling 40
 Orientation   of  Rules  in  a  DMN  Decision   Table
Rules  as  Rows:

Rules  as  Columns: Rules  as  Crosstabs:

Decision  Modeling 41
The  Decision Model

Von  Halle,  B.,  &  Goldberg,  L.  (2010).  The  Decision  Model:  A  Business  Logic  
Framework  Linking  Business  and  Technology.  CRC  Press  Auerbach
Publications.
Decision  Modeling 42
The  Decision Model
• The  Decision Model  was  developed by
Barbara  von  Halle  and Larry  Goldberg  (2010)
• Objective:  
• a  rigorous,  repeatable,  and  technology-­‐independent   model   of  business   logic   that   is  
simple  to   create,  interpret,   modify,  and  automate
• The  Decision Model  is  a  template  for  perceiving,  organizing,  and  managing  
the  business  logic  behind  a  business  decision.
• It  is  a  declarative  representation  of  decision  logic
• specifies the conditions on  which a   decision is made
• does not   specific how the conditions are tested,   in   particular it does not   specify the
order in   which conditions are tested

Decision  Modeling 43
Decision Model  Elements
A  Decision  Model  has  two  different  kinds  of  diagrams:

Decision Model  Diagram Rule Family  Table

Decision  Modeling 44
Decision Model  Diagrams
• The  root  of  a  Decision  Model  diagram  
(its  start)  is  an  octagonal  shape  that  
represents  the  entire  business  
decision
• It  is  this  shape  that  relates  to  tasks  
within  business  process  models.
• The  other  nodes  in  the  Decision  
Model  diagram  represent  Rule  
Families

(von  Halle  &  Goldberg  2010,  p.  26f)

Decision  Modeling 45
 Decision Model  Diagram
The   Decision   Model   Diagram  represents   Rule  Family   Tables.  

Decision

Rule Family Rule Family  Table

(von  Halle  &  Goldberg  2010,  p.  29)

Decision  Modeling 46
Rule Family:  Basic  Element  of the Decision
Model
• Rule Family  is  a  two-­‐dimensional  table  relating  conditions  to  one—and  
only  one—corresponding  conclusion.

name of conclusion
Column headings:  names being reached
of facts being tested

operator value of
condition column heading
(von  Halle  &  Goldberg  2010,  p.  18f)
Decision  Modeling 47
 Rule Family  Tables are Decision Tables

• A  Rule Family  table is a  kind of decision table

• In  a  Rule Family  Table  each row represents a  rule
• In  a  decision table each column represents a  rule
• A  Decision Model  is a  structured collection of decision tables
• There are some specialties:
• A  Rule Family  must  only have one conclusion column
• Inferential  relationships  between  Rule  Family  are  made  explicit  in  a  Decision  
Model diagram

(von  Halle  &  Goldberg  2010,  p.  25)

Decision  Modeling 48
Rule Family:  Basic  Element  of the Decision
Model
• A  Rule  Family  node  has  three  parts:
• The   name  is   the   conclusion  of  the  Rule   Family
• Inferred  conditions:  There   are   Rule  Families  with   these  names
• Basic  conditions:  There  are  no   Rules  Families  with   theses   names

Person Likelihood
of Defaulting on a
Loan

Person
Employment
History
Person Mortgage
Situation
Person
Miscellaneous
Loans Assessment

(von  Halle  &  Goldberg  2010,  p.  18f)

Decision  Modeling 49
 (von  Halle  &  Goldberg  2010,  p.  29)
Decision  Modeling 50
Translating a  Rule Family  into Natural  
Language
• It  is  possible  to  convert  each  row  in  a  Rule  Family  into  a  sentence  that  sounds  
natural  to  a  business  audience

■ Possible  Conversions
♦ If/when   Person  Employment  History   is  Poor  and   Person   Mortgage   Situation  is   Poor  
and   Person  Miscellaneous   Loans  Assessment   is   High,  then  the   Person   Likelihood  of  
Defaulting   on   a  Loan   is  High.
♦ A  Person   with   Poor  Employment  History   and   Poor  Mortgage  Situation  and  High  
Miscellaneous   Loans  Assessment   has   a   High  Likelihood   of  Defaulting   on   a  Loan.
♦ It is obligatory that the Person  Likelihood  of  Defaulting  on  a  Loan  is High if  
the  Person  Employment  History  is Poor and  the  Person  Mortgage  Situation is
Poor and  the  Person  Miscellaneous  Loans  Assessment  is High
Decision  Modeling 51
(von  Halle  &  Goldberg  2010,  p.  20)
 A  Rule Family  represents all  Rules  for one
Conclusion
The   Decision   Model   has  only   one   Rule  Family   for  each   type  of   conclusion  
column,  i.e.   all   rules   for  a  conclusion  are   in  one   table.

(von  Halle  &  Goldberg  2010,  p.  29)

Decision  Modeling 52
Rule Pattern
• A  set  of  Rule  Family  rows  with  a  common  set  of  populated  condition  
cells  is  called  a  Rule  Pattern.
• The  following  Rule  Family  represents  two  rule  patterns

1. The   1st  and 3rd    

rule have conditions
for the fact types
"Person   Student  
Loans"   and "Person  
Customer   Status"
2. The   2nd   and 4th  
rule have conditions
for the fact types
"Person   Business  
Loans"   and "Person  
Customer   Status"

(von  Halle  &  Goldberg  2010,  p.  24)

Decision  Modeling 53
Two dependend Rule Families
• Conditions  of  one  rule  family  can  depend  on  another  rule  family
• Example:  Person  Employment  History  in  the  first  rule  family  depends  on
• Person  Years   at   Current   employer   &  
• Person  Number  of  Jobs   in   Past   Five  Years

Person
Employment
History

Person Employment
History

(von  Halle  &  Goldberg  2010,  p.  23)

Decision  Modeling 54
Decision Model  Diagram (1/3)

• The  Decision  Model  diagram  shows    

the  inferential  relationships  between  
Rule  Families
• Solid   lines   between   Rule  Family  
nodes   represent   inferential  
relationships:  
• The   name  of  the   node   at   the  end  
with   the   dot   occurs   as   condition  in  
the   other  node
• Leave   nodes   in  a  Decision  Model  
diagram  to  not  have   inferred  
conditions

(von  Halle  &  Goldberg  2010,  p.  26f)

Decision  Modeling 55
 A  Decision   Model   diagram  begins   with  an  
octagonal  shape   that  represents   the  entire  

Decision Model  Diagram (2/3) business   decision

The  other  shapes   in  the  Decision   Model  
The  R ule  family  directly  connected  to  the   diagram  represent   Rule   Families.
business   decision  shape  is  called  the  “   This   diagram  has  6  Rule   Families.
Decision  R ule  Family”,  its  conclusion  is  the  
Inferring   value  for  the  
conclusion  sought  by  the  entire  Decision  
Second   Condition   of  
Model. Decision   Rule   Family

Conclusion
Inferred
Conditions
Conditions
Inferring   value  for  
the  first  Condition  
Conditions   based  
of  the  Decision   The  Decision   Rule   on  facts
Rule   Family family  

The  name  of  each  

Rule   Family  is  its  
conclusion   column  
heading.

Decision  Modeling 56
(von  Halle  &  Goldberg  2010,  p.  28)
 The  Decision   Model  Diagram   (3/3)
The  solid  line  terminated  by  the  dot connects   Rule  
Families  that  have  an  inferential  r elationship:  The  
conclusion  of  one  Rule  Family  is  used  as  a  condition  
in  another.  

The  solid   line:

The  conditional   columns   are  
shown   below   the  solid   line  and  
above  the  dotted  line.

Inferred  
Inferred   Conditions
Conditions

The  dotted  line: Conditions  based  

The  labels  below   the  dotted  line  denote   condition  
on  facts
column   headings   that  do   not  serve  as  a  conclusion  
column   heading   in  another  Rule   Family.

Decision  Modeling (von   Halle   &  Goldberg   2010,  

57p.  28)
(von  Halle  &  Goldberg  2010,  p.  28)
Decision  Modeling 58
The  Decision Model  vs.  DMN:   Diagrams
• On  the graphical level,  the Decision Model  Diagram is a  subset of DMN's
Decision Requirements Diagram
• Decision is in  both diagrams
• Rule Family  corresponds to Business  Knowledge  Model
• DMN  is more expressive;  compared to the Decision Model  Diagram
contains Input  data and Knowledge  Sources

Decision

Rule
Family

Decision  Modeling 59
The  Decision Model  vs.  DMN:   Decision Tables
• Decision Model  and DMN  use decision tables to represent the
decision logic.
• The  main  structural  differences  are  down  to  the  split  cell  versus  
single  cell  convention  for  the  operator  and  operand.
• The  semantics  of  decision  tables  in  DMN  is  more  expressive:  It  
can  return  multiple  values  and  can  specify,  how  multiple  values  
are  aggregated

Source:  http://blog.maxconsilium.com/2014/09/introduction-­to-­decision-­model-­notation.html

Decision  Modeling 60
Decision   Tables  in  DMN  and  TDM

Decision  Modeling 61
Literatur
• Von  Halle,  B.,  &  Goldberg,   L.  (2010).  The  Decision  Model:  A  Business  Logic  Framework  Linking  
Business  and  Technology.  CRC  Press  Auerbach Publications.
• BPM  Professional,   2014,  Introduction   to  Decision  Model   &  Notation  (DMN),  
http://blog.maxconsilium.com/2014/09/introduction-­‐to-­‐decision-­‐model-­‐notation.html
• Ross,  R.  G.,  &  Lam,  G.  S.  W.  (2011).  Building  Business  Solutions:  Business  Analysis  with  Business  
Rules.  Business  Rule  Solutions   Inc.
• Alcedo  Coenen  (2013).  Decision Model  &  Notation  (DMN)  1.0  – a  new OMG  Standard.  
http://de.slideshare.net/alcedocoenen/intro-­‐dmn-­‐10
• OMG  (2014).  Decision  Model   and  Notation,  Beta  1,  http://www.omg.org/spec/DMN/
• Broom,   N.,  2014.  The  Decision  Model  and  Notation  (  DMN  )  standard  -­‐ A  worked  example.  A  
web  version  is  on  http://visionalysis.co.uk/2014/01/08/going-­‐dmn-­‐tal/
• Broom,   N.,  2014,  Next  stop,  [DMN/TDM]  Interchange,  
http://visionalysis.co.uk/2014/11/24/next-­‐stop-­‐dmntdm-­‐interchange/

Decision  Modeling 62