Download presentation
Presentation is loading. Please wait.
Published byLorin Randall Modified over 9 years ago
1
Production Rule Representation Team Response Presentation to BEIDTF OMG Montreal Aug 2004 Ruleml.org
2
Agenda Introduction to PRR Introduction to PRR Team PRR Scope, Goals, Requirements PRR Positioning & Other Standards
3
What is a Production Rule? A statement of programming logic that specifies the execution of one or more actions in the case that its conditions are satisfied. –If Conditions then Actions –Executed by an inference engine NOT what is meant in OMG model transformations
4
Why a PRR Standard? Production Rules & Inference Engines are used in industry as a means of –Efficiently representing and executing business rules –100Ks – Ms of such rules in use across different organizations –Used in decision automation in financial svcs, government, insurance, healthcare etc Issues in PRR implementations include –How do I represent rules in my UML model? –How do I guarantee vendor neutrality in rule representation? Lack of standard results in –Rules being omitted from models –Rules being integrated into UML incorrectly –No rule exchange standards –No standardized training for modelers
5
State of the Technology Inferencing with Production Rules: 20 years old –Backward chaining and forward chaining are well established technologies Vendors in the industry –Commercial inference engines Fair Isaac: Blaze Advisor ILOG: JRules CA: Aion –Open source / source-available tools JESS DROOLS –More in development
6
PRR Team Members IBM: largest IT services and tools company Fair Isaac: Ilog: LibRT: vendor of rule verification technology RuleML: existing rule standards body the Top 2 Suppliers of inference engines & rule management technology
7
PRR Scope MDA Platform Independent Model: an “Inferencing platform” that is independent of specific inferencing product Excludes –Logic programming and unification –Constraint programming and constraint satisfaction –Event-Condition-Action (ECA) rules –Fuzzy rules and Bayesian inferencing
8
PRR Team Goals Opportunity for PRR standard that satisfies both OMG and RuleML.org needs Provides basis for rule interoperability & encourage further adoption of production rules in software systems Standardize the representation of rules within common UML diagrams NOTE: At this time, we are NOT concerned with theoretical MDA transformations to/from other metamodels HOWEVER: We do not preclude this capability being provided outside of this standard
9
PRR Requirements MOF 2.0 compliant Production Rule Metamodel –Abstract syntax for representing PR in UML models –Supports “patterns” across Conditions and Actions Supported inferencing algorithms –Forward or backward chaining –Support for Rete algorithm is required XML Schema for XMI support Rulesets –Properties: reasoning mode, identifier –Ruleset aggregation –Association with UML Activity Diagram Rule/ruleset naming Examples
10
PRR in UML Re-use existing UML languages as much as possible, with minimum extensions / changes required to achieve a PRR Conditions: OCL subset against a defined object model + extensions to support patterns + extensions to support variable / pattern sharing with actions - OR - RuleML constructs Actions: Action Language subset against a defined object model + extended to support scope defined through conditions’ context
11
Metamodel of Rules Possibly also consider “Procedural Rules”
12
Metamodel under development Investigating basics of representation structure as well as effects of OCL and AS; extended for rule scope
13
Comparison with Existing Stds Java Invocation JSR-094 Desire to include standard rule transport as well as invocation W3C / XML XPath etc Need for specific production rule representation RuleML Predicate-logic Requirement for more commercial usage Eg OORuleML separates rules from data
14
PRR positioning in MDA for Current Tools CIM PIM PSM Execution layer Fair Isaac & Ilog Rule Mgmt Technology IBM Tools and Solutions that define, embed production rules LibRT Rule Testing Technology PRR
15
PRR positioning in MDA for BEIDTF Models CIM PIM PSM Execution layer BPDM BSBR* Rule & Process Models Business Metamodel Business Object Model PRR BSBR Vocabulary
16
PRR positioning in MDA for BEIDTF Models CIM PIM PSM Execution layer BPDM BSBR* Rule & Process Models Business Metamodel Business Object Model PRR BSBR Vocabulary Appropriate engines / code generators
17
PRR Issues Is PRR still useful, given BSBR defines formal rules at business level? Will transformations be automatable? Or just traceable? Is OMG Metamodel structure too heavyweight? POSSIBLY: BSBR is an information model; not targeted at implementation of services UNKNOWN: Depends on customer interest. Note lack of success of CASE… POSSIBLY: Could derive a lightweight version in future
18
PRR Issues (2) Will RuleML be coverable in same standard? Use/extend of BOTH OCL & ASL? Why not extend just 1? PROBABLY: Working on examples to verify this assumption POSSIBLY: Might be that extensions are useful for other rules / uses in OMG?
19
Status & Summary Effective PRR team in place Current work: –Expanding metamodel to accommodate RFP requirements –Exploration of OCL + AL to provide behavior –Example-driven investigations comparing RuleML, LibRT RBML, Ilog + Fair Isaac rule implementations, and OCL + AL definitions Next: Diagramming PR Next Update: Nov mtg
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.