Specified text in a DMN model that describes something noteworthy that may not be obvious. It may be placed on a DRG with or without a relationship to something else. During workshops or early in the discovery phases of a project, the annotation is exceptionally helpful to use for notes and can be replaced later by more formal elements. Rows in decision tables may also use annotations to describe important information.

API management is a platform that governs the lifecycle of private and public APIs in the enterprise including capabilities for discovery, subscription, change management, proxy/redirection, transformation, reporting, billing and custom user portals. Decision management naturally aligns with API management when decision endpoints are exposed as either public or provide APIs.

In the context of enterprise software, an ‘Architect’ is a role that conceptualizes the whole system using visual modeling techniques capable of communicating the composition of software, hardware, connectivity, topologies, patterns and best practices.

An automated pipeline refers to a series of automated steps for automating testing, validation, and deployment of decisions within an organization. The goal is to streamline change processes and ensure consistency and efficiency.

Portion of a defined decision model that can be automated and orchestrated. Unstructured activities and human interactions typically fall outside this boundary.

A notation serving to decompose decision logic into small pieces which may be associated graphically with elements of a DRD.

Boxed logic is a visual means of representing common patterns of logic in an expression language such as conditional logic, filters and iteration. Rather than rely on text to convey the meaning of an expression, the common parts are accentuated by using boxes. Thus instead of determining a list of the discounted line items values of a shopping cart as

for Item in Shopping Cart
return Item.price*(100-Item.discount)

boxed logic represents it as:

Diagrammatic notation describing the order and flow of elements of a business process including orchestration of human interaction, systems interaction and automation tasks. Business Process Models can be informal sketches or use a formal standard such as OMG’s BPMN.

A Business Rules Management System is a platform that governs the lifecycle of rules and their optional deployment as executable endpoints, that are made available for execution through various protocols.

Business logic or domain logic is the encoding of real-world business rules that determine how decision outcomes can be created, stored, and changed.

https://en.wikipedia.org/wiki/Business_logic

A formal OMG standard. Motivation models allow an organization to articulate its strategic initiatives including mission, strategy, tactic, influencers, goals and objectives. It is well positioned as a framework for linking high-level thinking to lower-level automation activities.

See A/B testing.

Comparison of Incumbent approach to decision making to one or more new alternative approaches. Can be used in two contexts experimentation (testing) and simulation. In the case of simulation, the decision varies not the data in the case of testing the data varies.

A collection is a list of items. In the general case, items are not necessarily ordered, unique, or of the same type. There are more specific kinds of collections, such as sets (items are unique), ordered sets, maps (key-value pairs) and more.

The Conformance Scope is one of the documents published by the DMN On-Ramp Committee. It describes the scope of conformance being considered by the DMN On-Ramp Committee.

It identifies the features of the DMN standard that are in scope – those that are being considered for conformance – as well as those that are not. In addition, it identifies some areas outside the 1.3 version of the DMN standard that are being considered for conformance.

It outlines some principles for conformance in this context and identifies areas of best practice that the committee will consider.

An incremental method of application development and delivery characterized by automation of a pipeline to integrate, test, deploy and monitor every change made to the application.

CI/CD is the antithesis of risky, ‘big bang’ development: every change submitted by developers is integrated, checked to ensure it does not break the build and then fully tested prior to merging it into a central repository and deploying it to production.

CI/CD is typically the cornerstone of agile software delivery.

A person practicing data science. Data science is an interdisciplinary field that uses statistical methods, processes, algorithms and systems to extract knowledge and insights from noisy, structured and unstructured and apply knowledge from data across a broad range of application domains (Wikipedia)

A data type is a set of possible values that constrain what an Information Item can hold. For example, a bank balance would be numeric and a birthday would be a date.

Complex data types combine a number of properties each with its own data type. For example, a money amount might require both a currency and a number.

A collection of records, each containing multiple fields or data elements, used as input to test or simulate the impact of a decision. The records may be real historical data, anonymized or otherwise “cleaned” data, or synthetic data.

A business analyst, IT professional or subject matter expert who analyzes, creates and maintains the structure of a decision requirements model.

They may help multiple SMEs maintain the structure of decisions in multiple domains or they may maintain the structure of decisions they know well in a specific domain.

They may or may not also be a decision logic analyst.

An enterprise or application architect responsible for the stewardship of a decision management capability and the deployment and integration of decision services into the overall IT environment.

A DMN diagram is a “dependency diagram”: it depicts how decisions depend on information which would be provided by making other, dependent decisions or by using supplied data.

A record or log of how a decision was made about a particular transaction. Specifically, an account of the inputs and outcome of the decision and a description of how the decision model for that decision executed in the particular circumstances of that transaction.

A collection of all the decisions that need to be made in a particular domain or area of the business. These decisions may be described only at a high level or they may be part of an overall decision requirements model.

A decision inventory often includes mapping the decisions to KPIs or metrics, organizational structure and business processes.

The logic used to make decisions, defined in DMN as the value expressions of decisions and business knowledge models and represented visually as boxed expressions.

A business analyst, IT professional or subject matter expert who analyzes, creates, maintains and tests the logic of decisions. They may help multiple SMEs maintain the logic of decisions in multiple domains or they may maintain the logic of decisions they know well in a specific domain.

A formal model of an area of decision-making, expressed in DMN as decision requirements and decision logic.

The person or people who can approve changes to decision-making and who can choose between alternatives being considered. The person who gives final sign off to a decision requirements model and a set of decision logic that represents a specific business decision.

A set of requirements linking Decisions, Input Data, Knowledge Sources, Decision Services and Business Knowledge Models represented by one or more Decision Requirements Diagrams.

The model or graph is the complete set of these requirements not simply those represented on any specific diagram. Elements in the model may be shown on many diagrams each of which depicts a subset of the graph.

A software component encapsulating a decision model and exposing it as a service, which might be consumed (for example) by a task in a BPMN process model.

This is the role borne by a person who will decompose a decision and identify its sub-decisions, Input Data and Knowledge Sources, ultimately leading to a decision requirements diagram.

A tabular representation of a set of related input and output expressions, organized into decision rules indicating which output entry applies to a specific set of input entries.

A tree-like depiction of a set of rules which can serve as the implementation of a decision. Each node in the tree represents a test on an information item, and each leaf the decision made after traversing to that point.

Short for software development (“Dev”) and IT operations (“Ops”). It is a set of practices and tools which aims to shorten the development life cycle and allow for continuous delivery.

Shapes in a DMN diagram have an associated position which indicates how they are laid out. Typically, DMN tools keep these positions across sessions, and even provide one or several automated layouts.

 A DMN tool can provide some specific styling (extending that of the DMN standard), but could also allow users to provide custom styles, thus providing additional custom semantics.

The diagram layout and style can be interchanged between DMN tools if they support the DMNDI (DMN Diagram Interchange).

The primary goal of DMN is to provide a common notation that is readily understandable by all business users, from the business analysts needing to create initial decision requirements and then more detailed decision models, to the technical developers responsible for automating the decisions in processes, and finally, to the business people who will manage and monitor those decisions.

DMN creates a standardized bridge for the gap between the business decision design and decision implementation.

The purpose of DMN is to provide the constructs that are needed to model decisions, so that organizational decision-making can be readily depicted in diagrams, accurately defined by business analysts, and (optionally) automated.

A DMN product is any software application that allows users to deal with DMN (such as creating, viewing or editing models)

The Guide to DMN Product Capabilities is one of the documents published by the DMN On-Ramp Committee. It is a systematic means of assessing decision modeling and management tool functionality.  

The Guide is a disciplined means of comparing DMN tools, but more importantly of establishing a recommended minimum tool functionality for specific business scenarios (defined in the User Scenarios and Roles document).

A diagram presenting a (possibly filtered) view of a DRG.  A DRD visually depicts the Input Data, Decisions, Business Knowledge Models and Knowledge Sources pertinent to a specific aspect of decision making and the dependencies between them.

A graph of DRG elements (decisions, business knowledge models, knowledge sources and input data) connected by requirements and visually represented by one or more DRDs.

An enumeration is a list of allowed values. For example, States: CA, NM, NY, …

If the logic of a decision is fully defined, a computer is able to read and act on provided instructions to determine the outcome of a decision given specific inputs. The machine executing the instructions produces specific effects in accordance with the semantics of those instructions. On way of doing this in DMN is to execut the FEEL language.

A literal expression, decision table, invocation, list, context, function definition, or relation used to define part of the decision logic for a decision model in DMN. Returns a single value when interpreted.

A formal language used to express (or calculate) a value. Expression languages are a means of creating new values from existing names each of which are bound to a value. These new values can then be assigned to a name also. Expression languages have some of the properties of many programming languages (e.g., traversal of hierarchical data structures, inequities, filtering, projection, function calls, function definition). However, as with functional programming languages, expressions can only be create new values, not revise existing ones. As such, expression languages are often used to specify values in a readable way that can also be executed. Example expression language definitions might read:

Driver Type: if driver age < 18 then minor else adult
Insurance Price: goldPolicyCharge + minorUplift

The “Friendly Enough Expression Language” which is the default expression language for DMN.

A group of logic that can be used repeatedly, that can accept inputs and return output. Functions allow for the breakdown and reuse of logic.

Indicates how overlapping decision rules have to be interpreted. A single hit table returns the output of one rule only; a multiple hit table may return the output of multiple rules or an aggregation of the outputs.

A DMN element used to model either a variable or a parameter at the decision logic level in DMN decision models. These are used to express the inputs, interim results and outputs of a decision.

The dependency of a decision on an input data element or another decision.

The property of an information item that describes the kind of data it can represent is called the information type. Information types constrain the content of information items to make them more useful for consumers.

The robustness of an executable decision model can be improved by checking the type of an nformation item when it is created. Explicit typing of information items also improves readability. Typical types include date, number, string, Boolean, list and types derived from these.

Used to model the structure and the range of values of input data and the outcome of decisions, using a type language such as FEEL or XML Schema.

Key performance indicators are metrics used to measure business initiatives, actions or goals. A quantifiable measure of performance over time. Good KPIs are specific, measurable, attainable, relevant and time-bound.

A structured or semi-structured staged creation or change cycle from ideation to production deployment. There are many available methodologies and approaches that help to frame lifecycle management allowing for multiple groups of actors to participate in a lifecycle process.

An open-source tool to allow for Local Interpretable Model-Agnostic Explanations.

LIME Overview

Paper: “Why Should I Trust You?”: Explaining the Predictions of Any Classifier

LIME Open-soure Project

Through the use of data science models in computer systems, algorithms are trained to make classifications, clusters, associations or predictions. This allows for learning and adaptation without explicit instruction with the goal of analyzing and drawing inferences from patterns in data.

Consolidation of two different artifact versions, e.g., a decision table. This is useful when multiple groups or individuals work on the same artifact in parallel and require consolidated changes. Tools commonly provide visual comparisons to aid with the merge process. Merge results often fall into multiple states with visualization helping the user understand conflicts and problems.

Data that provides information about data e.g., author, date created, date modified.

By analogy to DevOps, MLOps is a set of practices and automation that combines machine learning and IT operations. It aims to shorten and improve the robustness of the machine learning development life cycle and its integration into a development system that delivers business value. MLOps defines an iterative plan-develop-test-deploy-operate lifecycle focusing on continuous testing, integration and delivery. An associated software stack supports and enforces these practices.

In Decision Table design, normalization refers to the technique of simplifying decision tables (for example by refactoring tables that have dependent conditions). This is analogous to normalization in relational databases, and is a best practice for decision table design. Using normalization prevents the needless repetition of conditions in decision tables, making them smaller, more transparent and more robust. This results in easier change and maintenance because there is no redundancy.

A way to define and discover entry points and input/output contracts for web services and automatically generate clients for communication. This allows for quick integration of decision services.

Perturbment analysis is a technique of determining the relative importance of a set of inputs to a decision in reaching an outcome. The technique measures, for each input, the relative adjustment required to change the decision’s outcome. For example, perturbment analysis might discover that a decision to grant or deny a mortgage is more sensitive to an applicant’s income than to their age or location.

It is an XML-based interchange format defined by the Data Mining Group (DMG) that supports the sharing of predictive (machine learning) models produced by data mining and other techniques. PMML is used in DMN to interface decision models with predictive and machine learning models such as regression and neural networks.

A practitioner is an individual skilled in the practice of decision modelling in the course of their job.

A pseudo (programming) language is a semi-formal dialect of natural language used to express an approach to solving a problem or to define an item of data. Such languages usually support some of the key ideas of programming (e.g., variables, conditional statements, iteration, function calls) without the attendant technical clutter of programming languages. An example of psuedo code follows:

If the driver is younger than 18 then classify the driver as a minor

Simplifying a table by splitting it into two or by combining two related tables into one.

In general, the process of restructuring without changing the intent of the original functionality. This is common when migrating from legacy programming languages, versions or systems.

Insert link here

A concept in cooperative game theory that allows one to measure the relative contribution of a group of collaborators to an outcome. Originally applied to economics, it is relevant both to decision models and machine learning as a method of determining the relative significance of each input to making a decision or predicting a value.

https://en.wikipedia.org/wiki/Shapley_value

An expert in a field or domain, with the necessary knowledge to build or help build a model for the decisions to be made in that field or domain.

This is the prediction of the value of a variable, given a set of inputs. Such a prediction is made by providing large data sets to algorithms which will build a statistical model close enough to the data sets, but not too close (the output of such algorithms could be visualized with a decision tree). Decision tree classifier is an example of such an algorithm. A decision tree is built from a ‘training’ data set using recursive binary descent. When new inputs are supplied, the decision tree is traversed to ultimately reach the conclusion – the prediction of the value.

When a decision is linked to another decision with an Information Requirement, the decision at the source of the information requirement is a sub-decision or dependent decision of the decision at the target (arrow) end of the requirement. Sub-decisions or dependent decisions must be made before the decision(s) that require them can be made.

A test case is a specific set of inputs and optionally expected results that validate a decision outcome in a particular business scenario.

Graphical representations of the logic of a specific decision in a decision model. For example, representation of the logic as a decision table, decision tree or boxed expression. In contrast, traditional code or script is not visual decision logic.

Insert link here