Decision Analytics Closing Symposium
Decision Analytics focusses on complex day-to-day decisions in companies and their underlying decision process. The quantity and variety of decisions and their importance calls for transparent, reproducible, complete and future proof decision processes. A well-structured approach is required to ensure these features. In the Decision Analytics research project state-of-the-art results from the scientific domains of business analytics and knowledge representations and reasoning are used.
The combination of both approaches provides the ability to develop powerful systems that 1) can achieve advanced behavior and 2) is based on domain knowledge provided directly by business users without the intervention of IT staff.
The closing symposium of the project focuses on the practical application of decision analytics in companies. More specific, attention will be paid to knowledge elicitation methods to create a knowledge base, the tools to use this knowledge, and the way end users can adjust the knowledge themselves.
24 September 2019
Registration with sandwiches and coffee
Location: Main Building
Welcome and Introduction to the Research Project
Prof. Joost Vennekens
Gathering relevant domain knowledge
The Support of Decision Modeling Features and Concepts in Tooling
Prof. Jan Vanthienen
Constraint solving with Z3
Dr. Sebastijan Dumancic
Natural Language Processing applied to decision Intelligence for investment strategy mandates
Bart Coppens & Nuno Comenda - Coppens & Partners Consulting
DMN+ : adding constraint reasoning to DMN
ing. Bram Aerts
Closing and Reception
Prof. Joost Vennekens
Content of the presentations
Gathering relevant domain knowledge (Marjolein Deryck)
The construction of a Knowledge Base that contains the necessary information to be able to provide an answer to your operational questions, can be a tedious task. This presentation gives an overview of different sources of information , and discusses different ways of extracting information from them.
The support of Decision Modeling features and concepts in tooling (Prof. dr. Jan Vanthienen)
This presentation examines to which extent some of the important Decision Model and Notation (DMN) features and concepts are supported by tooling. It is not a tool comparison, and no product names are revealed, but the analysis tries to give an indication of which elements of decision requirements diagrams, decision logic specifications and the (S)FEEL expression language are commonly present in current decision modeling (and execution) tools.
Constraint solving with Z3 (dr. Sebastijan Dumancic)
Despite machine learning being the most prominently covered subfield of AI in media, not every practical problem can be solved with machine learning. One category of such problems includes tasks for which an extensive domain knowledge exists, and thus does not have to be discovered from big data. Think, for instance, about problems in financial domains which have to respect certain restrictions and tax calculations imposed by the law; such tax regulation is an example of domain knowledge. The available knowledge can be used to answer many different questions and perform simulations of various financial scenarios. One technology that can be used to solve such knowledge-intensive problems is constraints solving. This talk will summarise the main insight we have obtained while applying the constraints solving technology on real-world financial and tax regulation problems.
Natural Language Processing applied to decision Intelligence for investment strategy mandates (Bart Coppens & Nuno Comenda)
nvestment strategy mandates are formulated contractually through a set of unstructured rules. How can a bank use this information to determine which securities it is allowed to invest in? We propose an NLP interface which supports the creation of logic rules used as input into a knowledge base system (IDP). Given a set of input rules the system infers whether or not a specific security is eligible for investment or not. It provides further insight into the reasons of its conclusion (i.e. explainability on the individual consequence of each rule).
DMN+: adding constraint reasoning to DMN (Bram Aerts)
The DMN standard allows users to build declarative models of their decision knowledge. The standard is deliberately kept simple enough to allow business users to construct these models themselves, without help from IT staff. It achieves this by combining simple decision tables with a clear visual notation. However, for real-life applications, the representation sometimes proves too restrictive. In this presentation, we propose an extension to the decision table notation of DMN, that we call DMN+. It aims at allowing more knowledge to be expressed, while retaining the appealing visual simplicity of DMN.