Chapter 2:

Modeling in
Business Intelligence
Contents

1 Models and Modeling in BI

2 Logical and Algebraic Structures

3 Graph Structures

4 Analytical Structures

5 Models and Data

6 Summary and Outlook

W.Grossmann, S. Rinderle-Ma, University of Vienna – Chapter 2: Modeling in Business Intelligence 2
1 Models and Modeling in Business Intelligence

There are many different models used in BI

− Examples you know:

W.Grossmann, S. Rinderle-Ma, University of Vienna – Chapter 2: Modeling in Business Intelligence 3

1 Models and Modeling in Business Intelligence

− Model Definition: Models represent some part of

the business process and allow precise formulation
of interesting questions (Analytical Goals)
− How can we realize the representation?
(representation function)
− How should we formulate the representation?
(“model language”)

W.Grossmann, S. Rinderle-Ma, University of Vienna – Chapter 2: Modeling in Business Intelligence 4

1 Models and Modeling in Business Intelligence

Representation function of models

− Models of phenomena
− Phenomena: Features of the business process interesting
from an analytical point of view
− Models define a picture of the phenomena (caricatures)
• Idealized models, e.g., control flow of the business
process, a treatment process, a course design
• Analogical models: Overtake ideas from other sciences,
e.g., gravity model for relations between persons in
dependence of distance
• Phenomenological models: Statistics, e.g. regression

W.Grossmann, S. Rinderle-Ma, University of Vienna – Chapter 2: Modeling in Business Intelligence 5

1 Models and Modeling in Business Intelligence

Representation function of models (ctd.)

− Models of data
• We have no precise idea about the models, but only a
number of candidate models for the empirical data
• The task is to learn the most appropriate model (Machine
Learning, Data Mining)
• Simple example: Churn management:
 Which variables influence the churn behavior of a customer , e.g.,
age, sex, marital status, income?
 How should we define the relation between churn behavior and
theses variables?

W.Grossmann, S. Rinderle-Ma, University of Vienna – Chapter 2: Modeling in Business Intelligence 6

1 Models and Modeling in Business Intelligence

Representation function of models (ctd.)

− Models of Theories
• Each application domain of BI has specific domain
knowledge, usually defined by concepts and
relation (logical relations) between the concepts
• Concepts and logical relations define a formal
system (ontology)
• Understanding this formal system as a theory data
instances are models of this theory
→ Database models
W.Grossmann, S. Rinderle-Ma, University of Vienna – Chapter 2: Modeling in Business Intelligence 7
1 Models and Modeling in Business Intelligence

Languages for Models

− Corresponding to the multitude of models
there are different formulations (languages)
used, for example:
• UML or ER-modeling for data

• BPMN for modeling the control flow

• Statistics in case of modeling customer

behavior
• Connectedness (reachability) in a graph
W.Grossmann, S. Rinderle-Ma, University of Vienna – Chapter 2: Modeling in Business Intelligence 8
1 Models and Modeling in Business Intelligence

Formulation of models
− Each language has its own semantic allowing
definition of certain model elements an formulation
of generic questions
• Queries in a database

• Simultaneous occurrence of two events in a

business process
• Strength of association between two variables

• Graph models for social networks

W.Grossmann, S. Rinderle-Ma, University of Vienna – Chapter 2: Modeling in Business Intelligence 9

1 Models and Modeling in Business Intelligence

Formulation of models (ctd.)

− Generic questions can be formulated in different
languages
• Example: Relations between attributes
 Formulate a query in a data model and represent the
result as a table
 Define a regression model and formulate the relation as
an equation
 Use a graphical language and visualize the relation in a
scatterplot

W.Grossmann, S. Rinderle-Ma, University of Vienna – Chapter 2: Modeling in Business Intelligence 10

1 Models and Modeling in Business Intelligence

Model Structures
− Putting all these things together leads to the concept of a model
structure composed of:
• Model Language:

 Syntax defines basic elements and the rules how to compose

model elements
 Semantic defines the meaning of the elements in the language,
independent from any domain
 Notation for communication of the expressions in the language

− Model Elements: Certain expressions in the model language, useful for

describing facts about the business process
− Generic questions: Questions formulated in the semantic of the model
language about properties of model elements
• Generic questions can be answered by specific analysis techniques
W.Grossmann, S. Rinderle-Ma, University of Vienna – Chapter 2: Modeling in Business Intelligence 11
1 Models and Modeling in Business Intelligence

Modeling
− A mapping of some part of the domain semantic of a business process
into a certain model structure (“Conceptual Modeling”)
• Examples for domain concepts and relations:
 Health Care Use Case:
 Higher Education Use Case:
 CRM Use Case:
− Definition of a model configuration: admissible expression in a model
structure which allows formulation of the analytical goal in questions
about the model configuration
− Connection of model configuration with observations: data about the
instances of the business process have to fit to the model configuration,
i.e., views and perspectives
− Definition of model variability: Usually data are blurred due to noise or
statistical variability
W.Grossmann, S. Rinderle-Ma, University of Vienna – Chapter 2: Modeling in Business Intelligence 12
1 Models and Modeling in Business Intelligence

Model Assessment and Quality

− Quality criteria for business process models
• Correctness: model is syntactical correct and mapping of domain
semantic and model semantic is appropriate
• Relevance: model complies with intended function, i.e., explain past
observations and predict future observations
• Economic efficiency: trade-off between complexity and costs (Occams
razor)
• Clarity: model can be understood by users
• Comparability: model fits in the overall analysis framework of the
business process

W.Grossmann, S. Rinderle-Ma, University of Vienna – Chapter 2: Modeling in Business Intelligence 13

1 Models and Modeling in Business Intelligence

Model Assessment and Quality (ctd.)

− Quality criteria for empirical models
• Objectivity: Results are independent of the person using
the model
• Reliability: results of the model can be reproduced

• Validity: model is useful from a practical point of view

 Content validity: model represents phenomenon under
consideration
 Criterion validity: high correlation between model results and other
external properties
 Construct validity: new results can be derived from model

W.Grossmann, S. Rinderle-Ma, University of Vienna – Chapter 2: Modeling in Business Intelligence 14

1 Models and Modeling in Business Intelligence

Models and Patterns

− Patterns describe local behavior whereas models
describe global behavior
− Examples:
• Medical treatment process: a pattern of co-
occurrence of certain medications
• Customer relationship: A pattern of occurrence of
certain combination of variables like outliers

W.Grossmann, S. Rinderle-Ma, University of Vienna – Chapter 2: Modeling in Business Intelligence 15

Contents

1 Models and Modeling in BI

2 Logical and Algebraic Structures

3 Graph Structures

4 Analytical Structures

5 Models and Data

6 Summary and Outlook

W.Grossmann, S. Rinderle-Ma, University of Vienna – Chapter 2: Modeling in Business Intelligence 16
2 Logical and Algebraic Structures

Language: Propositional logic and predicate logic

− Individual constants (names), e.g., “John Dee”
− Variables: placeholders for constants, e.g., “Student”, “Course”
− Functions: operating on constants or variables, e.g., “grade(Student)
= passed”
− Predicates: define properties for the individual constants, e.g., “Attends
BI”
− Quantifiers (“for all (∀)”, “exists (∃)”)
− Definition of terms by individual constants, individual variables, and
functions
− Generate atomic formulas by a predicate symbol followed by a number
of terms for which the predicate is applicable, e.g., “AttendsBI[John
Dee]”
− Build well formed formulas using propositional calculus and quantifiers,
e.g., ∃(Student)(∀(Course) grade(Student, Course) = passed
W.Grossmann, S. Rinderle-Ma, University of Vienna – Chapter 2: Modeling in Business Intelligence 17
2 Logical and Algebraic Structures

Model elements and generic questions

− Building expressions according to predicate logic
− Assign truth values to the expressions (interpretation)
− If the interpretation results in truth values TRUE for all
possible assignments of the free variables we call the
interpretation a model
− Generic questions are whether a well formed formula is
true
− Modeling using logical structures tries to capture domain
knowledge in a logical form
− The simplest form are terminology systems like taxonomies
W.Grossmann, S. Rinderle-Ma, University of Vienna – Chapter 2: Modeling in Business Intelligence 18
2 Logical and Algebraic Structures

Ontologies: “A specification of a conceptualization”

− OWL:
• T-Box: Vocabulary of a domain as a logical theory

• A-Box: Assertion about the domain, which has to

be checked
• Uses the open world assumption, i.e., anything
can be entered in the T-Box unless it violates
constraints

W.Grossmann, S. Rinderle-Ma, University of Vienna – Chapter 2: Modeling in Business Intelligence 19

2 Logical and Algebraic Structures

© 2015 Springer-Verlag Berlin Heidelberg

W.Grossmann, S. Rinderle-Ma, University of Vienna – Chapter 2: Modeling in Business Intelligence 20

2 Logical and Algebraic Structures

Frames
− Representation in an object-oriented style
− For each object a number of slots are defined for
attributes of the objects
− Frames use the closed world assumption, i.e., a
statement is true if its negation cannot be proven
within the system
− Example:
• “All birds can fly” (closed world)

• “There exist non flying birds” (open world)

W.Grossmann, S. Rinderle-Ma, University of Vienna – Chapter 2: Modeling in Business Intelligence 21
Contents

1 Models and Modeling in BI

2 Logical and Algebraic Structures

3 Graph Structures

4 Analytical Structures

5 Models and Data

6 Summary and Outlook

W.Grossmann, S. Rinderle-Ma, University of Vienna – Chapter 2: Modeling in Business Intelligence 22
3 Graph Structures

Model Structure - Language:

− Syntactic elements:
• Nodes (vertices)
• Edges (directed, undirected)
• Labels for edges (e.g., “distance”) or nodes (e.g.,
“degree”)
− Notation:
• Numeric representation (adjacency matrix)
• Visual representation

W.Grossmann, S. Rinderle-Ma, University of Vienna – Chapter 2: Modeling in Business Intelligence 23

3 Graph Structures

Model Structure
− Model elements
• Special kinds of graphs, e.g., trees, series parallel networks, bipartite
graphs
• Connected graphs (path)

© 2015 Springer-Verlag Berlin Heidelberg

− Generic questions
• Generic questions refer to properties of the graph and can be
answered by well known algorithms like spanning tree, shortest path,
best matching of nodes

W.Grossmann, S. Rinderle-Ma, University of Vienna – Chapter 2: Modeling in Business Intelligence 24

3 Graph Structures

Modeling using graph structures, e.g.,

− Business process modeling and notation (BPMN)

© 2015 Springer-Verlag Berlin Heidelberg

W.Grossmann, S. Rinderle-Ma, University of Vienna – Chapter 2: Modeling in Business Intelligence 25
3 Graph Structures

− Modeling using graph structures, e.g., Petri Nets

© 2015 Springer-Verlag Berlin Heidelberg

W.Grossmann, S. Rinderle-Ma, University of Vienna – Chapter 2: Modeling in Business Intelligence 26

Contents

1 Models and Modeling in BI

2 Logical and Algebraic Structures

3 Graph Structures

4 Analytical Structures

5 Models and Data

6 Summary and Outlook

W.Grossmann, S. Rinderle-Ma, University of Vienna – Chapter 2: Modeling in Business Intelligence 27
4 Analytical structures

Calculus (ctd.)
− Model elements (ctd.)
− Inner product, given a vector w of coefficients:
𝑓𝑓 𝑥𝑥 = 𝑤𝑤 𝑇𝑇 𝑥𝑥 = w1 x1 + ⋯ + wp xp
− Linear functions in more than one variable (matrices): 𝑓𝑓 𝒙𝒙 = 𝐵𝐵𝒙𝒙
where B is a 𝑘𝑘 × 𝑝𝑝 matrix
− Projections: The orthogonal projection of a vector x onto another
′ 𝑤𝑤
vector w is defined by 𝑝𝑝𝑤𝑤 𝑥𝑥 = 𝑥𝑥 ∗
𝑤𝑤

W.Grossmann, S. Rinderle-Ma, University of Vienna – Chapter 2: Modeling in Business Intelligence 29

4 Analytical structures

Probability – Language
− Events, Calculus of events: E
𝑃𝑃(𝐸𝐸)
− Probability of events 𝑃𝑃 𝐸𝐸 , 𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜 𝐸𝐸 =
1−𝑃𝑃(𝐸𝐸)
− Random variables as model for measurement: X
− Probability Distribution:
• Distribution function: 𝐹𝐹 𝑥𝑥 = 𝑃𝑃(𝑋𝑋 ≤ 𝑥𝑥)
• Density function and probability function: 𝑝𝑝(𝑥𝑥)
• We interpret the density as likelihood of an observation

W.Grossmann, S. Rinderle-Ma, University of Vienna – Chapter 2: Modeling in Business Intelligence 31

4 Analytical structures

Probability – Language (ctd.)

− Conditional probability and independence:
𝑝𝑝(𝑥𝑥 |𝑦𝑦) = 𝑝𝑝(𝑥𝑥, 𝑦𝑦)/𝑝𝑝(𝑦𝑦)
− Two variables are independent if
𝑝𝑝 𝑥𝑥, 𝑦𝑦 = 𝑝𝑝 𝑥𝑥 ∗ 𝑝𝑝 𝑦𝑦
− Bayes Theorem: 𝑝𝑝(𝑥𝑥|𝑦𝑦) = 𝑝𝑝(𝑦𝑦|𝑥𝑥)/𝑝𝑝(𝑦𝑦)
− Interpretation of Bayes Theorem in the discrete
case: Compute column percentages from row
percentages

W.Grossmann, S. Rinderle-Ma, University of Vienna – Chapter 2: Modeling in Business Intelligence 32

4 Analytical structures

Probability – Example

© 2015 Springer-Verlag Berlin Heidelberg

W.Grossmann, S. Rinderle-Ma, University of Vienna – Chapter 2: Modeling in Business Intelligence 33

4 Analytical structures

Statistics – Language
− Statistical units (observation units)
− Population
− Observable variable
− Transfer the concepts of probability to observations, e.g., “distribution” to
“sample distribution” (“empirical distribution”)
− Model elements and generic questions:
• Descriptive methods
• Inferential methods
 Estimation
 Testing
 Confidence regions
− Modeling methods
• Regression

W.Grossmann, S. Rinderle-Ma, University of Vienna – Chapter 2: Modeling in Business Intelligence 34

Contents

1 Models and Modeling in BI

2 Logical and Algebraic Structures

3 Graph Structures

4 Analytical Structures

5 Models and Data

6 Summary and Outlook

W.Grossmann, S. Rinderle-Ma, University of Vienna – Chapter 2: Modeling in Business Intelligence 35
5 Models and data

Data Generation
− In BI we have usually secondary data, i.e., data which have been
collected for other purposes, e.g.,
• Transactional data
• Administrative data
• Web data
− An important question for interpretation of results is defining the
population which is represented by the data (e.g., tweets or
evaluations on portals)
− Measurement of the data

W.Grossmann, S. Rinderle-Ma, University of Vienna – Chapter 2: Modeling in Business Intelligence 36

5 Models and data

Elements of the knowledge based temporal abstraction method

− Time stamps 𝑇𝑇𝑖𝑖 are the basic primitives with a predefined granularity and
a well defined zero.
− Time intervals 𝑇𝑇 = 𝑇𝑇𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 , 𝑇𝑇𝑒𝑒𝑒𝑒𝑒𝑒 are defined as pairs of time stamps for
start and end. Time points are zero length intervals.
− An interpretation context 𝜉𝜉 is a proposition that can change the
interpretation of parameters within the scope of a time interval.
Interpretation contexts can be nested.
− A context interval 𝜉𝜉, 𝐼𝐼 defines time intervals for which the interpretation
context holds.
− An event proposition e represents the occurrence of an external
volitional action or process and has to be distinguished from a
measurable datum.
− An event interval 𝑒𝑒, 𝐼𝐼 represents the temporal duration of an event e.

W.Grossmann, S. Rinderle-Ma, University of Vienna – Chapter 2: Modeling in Business Intelligence 37

5 Models and data

Elements of the knowledge based temporal abstraction method (ctd.)

− A parameter schema 𝜋𝜋 is a measurable aspect of the state of the world (states
of a process) with values in some domain 𝑣𝑣 𝜖𝜖 𝑉𝑉𝜋𝜋 . Parameter schemas may be of
different type: primitive parameters (measurable data), abstract parameters
(concepts), constant parameters (instant specific or instant independent).
− A parameter proposition 𝜋𝜋, 𝑣𝑣, 𝜉𝜉 defines the values of parameters in a context.
− An abstraction function 𝜃𝜃 𝜖𝜖 Θ maps parameters into abstract parameters.
− A parameter interval 𝜋𝜋, 𝑣𝑣, 𝜉𝜉, 𝐼𝐼 denotes the value v of the parameter 𝜋𝜋 in the
context 𝜉𝜉 during time interval I.
− An abstraction is a parameter or a parameter interval.
− An abstraction goal 𝜓𝜓 𝜖𝜖 Ψ represent a specific intention or goal.
− An abstraction goal interval 𝜓𝜓, 𝐼𝐼 represents the idea that abstraction goal 𝜓𝜓
holds in interval I.
− Induction of context intervals allows the induction of events, parameters, or
abstraction goal propositions for some context interval.

W.Grossmann, S. Rinderle-Ma, University of Vienna – Chapter 2: Modeling in Business Intelligence 38

5 Models and data

Quality Dimensions for Data

− Relevance measures in how far the data are useful in the intended context.
− Accuracy is the degree of conformity of a measure to a standard or a true value.
− Completeness is a characteristic measuring the degree to which all required
data is known, with respect to depth, breath, and scope.
− Timeliness: Data coming early or at the right time, appropriate or adapted to the
times or the occasion.
− Consistency is expressed as the degree to which a set of data is equivalent in
redundant or distributed databases.
− Coherence refers to the adequacy of the data to be reliable combined in different
ways and for various uses.
− Reliability is a characteristic of an information infrastructure to store and retrieve
information in an accessible, secure, maintainable, and fast manner.

W.Grossmann, S. Rinderle-Ma, University of Vienna – Chapter 2: Modeling in Business Intelligence 39

Contents

1 Models and Modeling in BI

2 Logical and Algebraic Structures

3 Graph Structures

4 Analytical Structures

5 Models and Data

6 Summary and Outlook

W.Grossmann, S. Rinderle-Ma, University of Vienna – Chapter 2: Modeling in Business Intelligence 40
6 Summary and outlook

− Modeling is a rather intricate activity in BI

− Different approaches for model representation and model
presentation
− Key tasks:
• Definition of a model configuration

• Connection of the model configuration with the

observations from business process instances.
• Formulation of a model component for capturing the
variability of the different process instances.

W.Grossmann, S. Rinderle-Ma, University of Vienna – Chapter 2: Modeling in Business Intelligence 41

6 Summary & outlook

W.Grossmann, S. Rinderle-Ma, University of Vienna – Chapter 2: Modeling in Business Intelligence 42