DecisionCAMP 2016 Program Rules, Decision Modeling and Decision Management Technology

Best Practices, Standards, Real-World Business Cases, and Supporting Products

  Sponsored By: 

The DecisionCAMP accepted the following presentations from leading decision management authorities, vendors, and practitioners: 

Authors Title with Presentation Link Abstract

Charles Forgy

(Production Systems Technologies)

Factors Affecting Rule Performance
Conventional Java, C, etc. programs have a fairly direct mapping onto computers, so it is usually easy to understand why certain parts of a system are slow. Such is not the case for rule-based systems.

Fortunately, there are some very powerful tools available today for analyzing our systems. In this talk we will use a few of them to examine a rule set running in a Java rule engine. We will first determine which rules are taking excessive time, then determine which parts of the rules are responsible. Then we can look at those parts to determine where the computer is spending its time, and why. Finally, we will look at what can be done to ameliorate the performance problems.

Bruce Silver

DMN as a Decision Modeling Language
Decision Model and Notation (DMN) is a relatively new decision modeling standard maintained by the Object Management Group. Based on a formal metamodel, it combines a business-oriented graphical notation with precise rule-based decision logic semantics. As such, DMN tools allow non-technical users to define, validate, and maintain executable decision logic themselves, as opposed to the traditional error-prone approach of writing business requirements for programmers.
In the notation, the dependencies of a complex decision on other supporting decisions and input data are represented graphically by a Decision Requirements Diagram (DRD). The decision logic of each decision node in the DRD is defined by a variety of tabular formats called boxed expressions, and DMN also specifies a new expression language, FEEL, used in the boxed expressions. In combination, the DRD, boxed expressions, and FEEL constitute a powerful decision modeling language standard. In fact, the XML serialization of a DMN model captures all the essential semantic details of the notation, so that it can be validated for completeness and consistency, and – supplied with input data values – directly executed on a suitable engine.
This keynote talk reviews the structure and key features of DMN 1.1 as a decision modeling language.

Alan Fish


Modeling Decision-Making Processes: Melding Process Models and Decision Models

Organisational decision-making is often extended over time, and involves a number of different activities which depend upon each other. Most business analysts would agree that one should model the logical structure of a domain of decision-making separately from any sequence flows between the surrounding process activities, but often the selection and ordering of those activities is one of the objects of the decision-making. This can give rise to an interesting tension between the process model and the decision model. I compare different approaches to resolving this issue, using real examples from recent decision automation projects using DMN and BPMN.

Gary Hallmark, 
Alvin To


Oracle Decision Modeling Service
Oracle's Decision Modeling Service (ODMS) is a cloud-based service for modeling, automating, and executing business decisions. ODMS supports all standard OMG DMN FEEL expressions, including decision tables, boxed expressions, and all FEEL expressions allowed at DMN Conformance Level 3. 

We will demonstrate ODMS by modeling, testing, and running a decision model abstracted from Oracle Application's Approvals Engine. This is a workflow engine that has both human-powered and automated decision making. We will focus on a small automated component - given a purchase order with multiple lines, each line having potentially multiple cost centers, return a list of approvers who must sign off on this PO. 

Our DMN standard solution uses functions to encapsulate decision tables that must be executed for each line and for each cost center, and FEEL for-loops to invoke the functions for each line and for each cost center. 

Decision models can be tested in two ways. The conventional method is to encapsulate all or part of a model in one or more decision services, and use a service testing framework. For more immediate feedback, ODMS allows you to attach sample data directly to the model inputs and see the resulting outputs as you build the logic. 

Decision services created in a DMN Model can be executed as RESTful web services by posting values for the inputs and then receiving values for the outputs.

Jan Vanthienen

(K.U. Leuven)

DMN: how to satisfy multiple objectives?
Decision modeling with DMN allows to model decisions and offers a standard notation and expression for decision requirements and decision logic. Is there only one model for a (set of) decision(s) or can you have multiple models?

Based on a recent Decision Management Community challenge, this session investigates various objectives of decision modeling: providing an overview for business, verification of business logic, traceability to knowledge sources, maintainability, model-driven execution. The solutions to the challenge showed three classes of models:
- A model to clearly capture the original specification, with advantages of traceability and maintenance.
- A model to show the overview and analyze the business concerns.
- A model that tries to combine both views.

All three classes are perfectly possible in DMN, and are immediately executable from the model. So which one to choose, and how to satisfy the multiple objectives?

Dan Selman


Decision Management at the Speed of Events
Learn how IBM is applying business rules and predictive models to event streams to detect business situations and decide on an appropriate response. 

Dan will present customer use cases that benefit from near realtime situation detection, the capabilities of the underlying platform and the key elements of the agent-based programming model.

Igor Elbert (GILT) 
Jacob Feldman (OpenRules)
Using Machine Learning, Business Rules, and Optimization for Flash Sale Pricing In this presentation we will share our experience in applying a combination of Machine Learning, Business Rules, and Multi-Criteria Optimization to build a pricing system for an online retailer selling curated collections of fashion products via flash sales. The system is capable of predicting demand and recommending optimal prices for ever-changing assortments of thousands of products. It decouples prediction from applying business rules from optimization to allow replacing each of these component with the best of breed solution. The system collaborates with business domain experts to set up optimization criteria and human-supplied hard and soft constraints to quickly generate prices that can be immediately go live on site

Tom Debevoise, 
Will Thomas


Welcome to Method for Parsing Regulations into DMN
The OMG specification for DMN defines the components for an executable decision; however, it does not suggest a path for transitioning from a plain-text specification to a model. Besides, It is rare that a company complies with the exact wording of regulation—these must be framed in the context of business operations.

In this discussion, we describe how businesses are meeting compliance needs with decision modeling. We suggest a method for identifying sub decisions, business rules, and data elements within compliance documents. Given the simplicity of DMN and the comprehensive nature of the decision model, business analysts, and legal experts can be trained in this method.

Merely drawing diagrams is not adequate for developing an executable compliance model. Also, the models need to be simulated so that details of logic can be clarified and test cases collected and documentation. A final benefit of using DMN is an added layer of compliance traceability. If needed, every section of regulation can be connected to a graphical model and corresponding element of the logic.

Larry Goldberg

(Sapiens DECISION)

Solving the "Last Mile" in model based development
Industry tools may gracefully transform our process and decision diagrams and their shapes into executable objects, but that may really be the simple part of the task at hand: the greater challenge may lie in the technical integration tasks; Neal Ward Dutton aptly calls this the “dirty secret” in Process and Decision Management. 

To really achieve the higher levels of maturity, we have to be able abstract into business manageable terms the data connections and APIs necessary to implement business-generated capabilities. 

This presentation will demonstrate such an abstraction layer and discuss the positive consequences of treating Big Data in business modeling terms.

Benjamin Grosof,
Janine Bloomfield

(Coherent Knowledge)

We present a case study and latest engineering extensions of our approach to advanced analytics based on Rulelog knowledge representation and reasoning (KRR). Our system Ergo is a platform component for Rulelog KRR that powers applications built with IT partners. Ergo flexibly represents complex knowledge sentences and then executes deep (multi-step) reasoning. The reasoning provides, for each answer, a fully detailed logical explanation (proof via the logical chain of reasoning). The explanation can be viewed in English and interactively navigated (by an end user) in order to drill down selectively. One recent engineering extension is ErgoText: templates which hybridize logical syntax with natural language syntax. Other recent engineering extensions are to add federation connectors that speedily import enterprise data from diverse forms and sources, including RDF/SPARQL graph databases, CSV and spreadsheets, SQL relational databases, and OWL ontologies. The resulting overall reasoning scales up to millions of inferred data items per processor, and can utilize extremely large data sources by importing on demand (i.e., dynamically during reasoning). The case study, in the health care and life sciences domain, focuses on decision support for disease and drug treatment guidance by answering queries about relationships between patients, symptoms, diseases, drug and non-drug treatments, risks, and contraindications. Complex knowledge, such as treatment policies and biomedical causal pathways, are fully represented by sentences in Ergo’s highly expressive syntax. Business benefits, illustrated by the case study, include agility, scope of automation, and verifiability/transparency, together with cost-effectiveness, accuracy, and reusability.

James Owen (KBS) Charles Forgy (Production Systems Technologies)

Improving BRMS Efficiency and Performance and Using Conflict Resolution
This presentation deals with some extremely important, but often misunderstood, aspects of BRMS: efficiency and performance. This can be achieved through Goal-Directed techniques, Daemon Rules, elimination of the number of elements in the LHS of the rule, avoiding large cross-products, re-ordering condition elements, merging element classes, avoiding changes to matched conditions, putting change elements in the LHS, avoiding excessive changes to control elements and deleting any elements that are not absolutely necessary. This, combined with the importance of conflict resolution (CR) techniques that have been removed from “modern” systems should dramatically improve both efficiency and performance. We will show in this paper how making CR parallel to improve time as well as comparing MEA (Means Ends Analysis) with LEX, Breadth and Width with different rule sets.

Mark Proctor
(Red Hat)

Learning Rule Base Programming with Classic Computer Games Writing classic computer games is a fun and challenging way to learn rule based programming, especially execution flow control. This talk will develop games live, while explaining the mechanics of how those things work.

Jacob Feldman

What-If Analyzer for DMN-based Decision Models
In this presentation we will demonstrate a new web-based graphical tool “What-If Analyzer for Decision Modeling” that supports what-analysis of different DMN-based decision models. It allows a user to activate some business rules and deactivate others showing the immediate changes in the decision variables. With a simple click it may produce multiple decisions that satisfy all active rules within the same decision model. If the decision model specifies a business objective that depends on other decision variables, then What-If Analyzer may find a decision that minimizes or maximizes this objective.

Shenghui Cheng,
Klaus Mueller
(Stony Brook University)

The Decision Boundary Map: An Interactive Visual Interface to Make Informed Decisions and Selections in the Presence of Tradeoffs Decisions must often be rendered in the presence of tradeoffs. However, tradeoffs are often difficult to recognize and balance when there are many factors playing a role in the decision process. In these scenarios, human decision makers are easily overwhelmed and render decisions that they are not fully sure about, only to recognize later that better choices could have been made. We propose an interactive visual interface, called the Decision Boundary Map, which exposes tradeoffs in the context of the decision factors and the available data. It allows users to interactively specify and modify acceptable intervals of the various factors and see how possible solutions comply with these settings.

QnA Panel
DMN Standard from OMG, Vendor, and Practitioner Perspectives
 Moderated by Bruce Silver