Notes about DecisionCAMP-2016
Blog: Decision Management Community
DecisionCAMP/RuleML 2016 is now a history, and I’d say a very good history with a lot of very interesting mini-events. I will try to briefly describe some of them in these notes. This year the Decision Model and Notation (DMN) standard was in focus not only at DecisionCAMP and surprisingly at the RuleML as well (see its proceedings).
- representatives from the leading BR&DM vendors including IBM, FICO, Oracle, Signavio, Red Hat, OpenRules, Sapiens, Camunda, Sparkling Logic,…
- well-known authorities in the world of business rules and decision management from the “godfather” of rule engines Dr. Charles Forgy and to young and bright architects of the modern DM products
- scientists and famous book authors working in the area of business rules and semantic web
- current and potential DM customers from Wall Street and other corporations.
For me personally the informal discussions with DM experts and many bright individuals during presentations, coffee/lunch breaks, and evenings became one of the most important parts of the event.
On July 6 we hosted the OMG meeting of the DMN 1.2 Revision Task Force (RTF). It was the whole day of pretty hot discussions about features to be added and/or modified in the new DMN Release 1.2. All participating members of the RTF found this meeting very productive. Without going into specific details I’d just say that DMN authors are really incentivized by the current DMN acceptance and are trying to make the standard simpler for vendors to implement and for business users to apply in their operational decisioning practice.
All DecisionCAMP presentations are available for free downloads from here. The majority of presentations were delivered by people who really know what they are talking about and who brought interesting perspectives of what is going on in the decisioning domain now and what we may expect in the future. I’d provide just a few comments to presentations that generated the most interest.
In the very first presentation called “Melding process and decision models” Dr. Alan Fish from FICO discussed overlapping between BPMN and DMN. In spite of Alan’s carefully worded attempts to avoid any potential conflicts and recommending to “use DMN to specify services called by BPMN”, his presentation brought explosive discussions whether”BPMN” will survive. One extreme comment was: if we remove the majority of diamonds from a BPMN diagram, what will remain, a sequence of rectangles? Will not DRDs cover the entire process modeling by supporting end-to-end decisions? If business events are supported by a message bus, then what will be a role of BPMN? The fact that people got used to business process diagrams will hardly protect BPMN in a long run. Earlier this year at bpmNEXT-2016 people also talked about future consolidation of DMN, BPMN, and CMMN standards. I’d say, that all 3 standards will live and cooperate well under the roof of Decision Management that goes far beyond decision modeling adding to the mix other important technologies such as predictive analytics, event management, and decision optimization.
Gary Hallmark, the inventor of DMN FEEL, presented the first release of “Oracle Decision Modeling Service“. Most importantly Gary demonstrated that a complete FEEL is implementable. At the same time, the friendliness of the “Friendly Enough Expression Language” was put in question several times during his presentation. For me, the most questionable part of the current DMN is not FEEL expressions even with long names containing spaces but rather the notion of the “boxed expressions and invocations”. Gary expects that the official Oracle release will be available by the end of this year.
One of the most impressive presentations was “Decision Management at the Speed of Events” given by Dan Selman, a chief architect of IBM Decision Server. I really like the way how IBM incorporated event/entity constructs including temporal and spatial relationships into IBM ODM product. While the term “CEP” disappears, the new “Unified Business-Friendly Language” looks really good. Putting the Situation Detection in focus empowers the Decision Management today and proactively prepares us for the problems of tomorrow (such as those brought by blockchain). I expect the future versions of DMN will include similar event management constructs.
Tom Debevoise, a chief DM evangelist from Signavio, shared their Wall Street experience of applying DMN to an enormously complex government regulation compliance problem – see “Welcome to Method for Parsing Regulations into DMN“. I know from OpenRules own experience how difficult it is to deal with legal documents oriented to lawyers, never tested and containing internal conflicts and uncovered situations. So, it is even more interesting to learn how Signavio managed to provide DMN-based templates to tackle these problems.
A real-world use case for a combination of DM technologies was presented at “Using Machine Learning, Business Rules, and Optimization for Flash Sale Pricing” by Igor Elbert, a Distinguished Data Scientist from GILT.com. Igor shared a quite ingenious approach involving R, OpenRules, and Linear Solvers to deliver an efficient price optimization system for GILT, an inventor of the online flash sales model. Being intimately involved in this development, I may confirm that the biggest challenge for us was a switch from a POC model with 20K decision variables to the real production model with 300-400K variables. The critical solution we came with was a switch from one giant decision model to thousands of automatically generate smaller models easily solvable by JSR-331 compliant linear solvers. The presentation generated a lot of follow up questions from the audience proving that real-life applications are usually more interesting than any theoretical considerations.
The keynote presentation “DMN as a Decision Modeling Language” by Bruce Silver summarized the current state of the DMN standards showing in a very precise and demystifying way what it is all about. Bruce candidly shared his initial pessimism about vendor support for DMN and his current optimism that is based on the recent announcements of DMN FEEL support by Oracle, Trisotech, Signavio, OpenRules, Camunda, Red Hat, and Sapiens. Bruce, being a member of the DMN Task Force, criticized several “not so friendly” features of the current DMN, leaving all of us with an impression that DMN will be improved, supported by many vendors, and has a bright future. You may read a transcript of Bruce’s keynote here.
I didn’t have time to attend many RuleML presentations but heard they were really good. I listened the tutorial “Practical Knowledge Representation and Reasoning in Ergo” delivered by Michael Kifer – it looks now we finally have a quite practical reasoning tool. Adrian Paschke described a working translator from DMN to RuleML – interesting, but I believe it will be much more productive to go in an opposite way, from RuleML to DMN. Afterwards, I spoke with Michael Kifer and several other Semantic Web experts. It seems today it is quite possible to go from an ontology to an automatically generated decision model that may provide decision logic to address various problems defined in a knowledge repository. Hopefully, the collaboration started at DecisionCAMP/RuleML 2016 will lead us to some practical development using RuleML and DMN together.
There were many more interesting presentations:
- Charles Forgy and James Owen shared experience how to write and execute rules more efficiently – see Factors Affecting Rule Performance and Improving BRMS Efficiency and Performance and Using Conflict Resolution
- I presented “What-If Analyzer for DMN-based Decision Models” demonstrating how optimization can benefits DMN-based decision model. In particular, I suggested to extend DMN decision models to expect not one output decision for a given input but rather to consider alternative and optimal decisions. It will also exclude an unrealistic completeness requirement from DMN.
- Jan Vanthienen’s presentation DMN: how to satisfy multiple objectives? was devoted to what constitute a good decision model using the experince of this DMCommunity Challenge
- Mark Proctor demonstrated the use of Drools for Learning Rule Base Programming with Classic Computer Games
- Larry Goldberg spoke about Solving the “Last Mile” in model based development using the real-world experience of applying TDM to large financial applications
- Benjamin Grosof talked about Advanced Decision Analytics via Deep Reasoning on Diverse Data: For Health Care and More showing capabilities of their semantic web system “ERGO” already described by Michael Kifer
- Klaus Muller presented his The Decision Boundary Map: An Interactive Visual Interface to Make Informed Decisions and Selections in the Presence of Tradeoffs
As a summary, the DecsionCAMP/RuleML 2016 became a very important Decision Management event and we are looking forward to the next event in 2017 in London.
As a chair of the DecisionCAMP, I want to thank all attendees, presenters and organizers!