The immense popularity of the Internet and recent technological innovations in storage technology have caused a true tsunami of data. The need to build intelligent information systems, targeted at learning patterns from data and subsequently deploying them in key business processes, is now stronger than ever before.  In this research, we will first address some theoretical research objectives related to how to build intelligent information systems by developing new predictive algorithms.  In a second step, the theoretical contributions will be validated in several practical, real-life settings.
It will be noted that the suggested research is multi-disciplinary in nature, having a core management informatics focus, but with clear links to other fields.  Hence, we plan to collaborate with a world-wide team of experts in each of the respective domains. 

 

Today many companies rely on third party applications and application services for (part of) their information systems. When applications from different parties are used together, an integration problem arises. Similarly, cross-organisational application integration requires the coordination of distributed processing across several autonomous applications. In this project, we propose the architectural elaboration and formal definition of an integration approach based on an event-based coordination paradigm. Interaction is based on atomic units of interaction called "business events". Each business event mirrors some event in the real world that requires the coordination of actions within a nuber of components. The coordination between applications is achieved by separating business process management aspects from event notification. The business process management aspects are realised by having applications specify preconditions for business events. As a result, a business event becomes a small scale contract between involved applications: each application can insert its own cleuses into the contract by specifying preconditions. The formalisation allows the elaboration of a method for contract analysis, to verify whether the contract is free from contradictions and inconsistencies. In addition to its contracting aspect, the event based wommunication paradigm entails a notification and coordination mechanisms. The project aims at the definition of a platform independent model of this coordination framework, the formal definition of the semantics by means of process algebra, the definition of formal verification algorithms and the elaboration of semantic preserving model transformations to platform specific models.

In the context of the recently put forward Basel II requirements, financial institutions are being faced with important challenges concerning the need for adequate risk profiles of their customer portfolios. This need is getting more and more emphasized by the severe financial and competitive climate companies operate in nowadays (cf. the Enron case).  Using these risk profiles would allow to make better financial decisions, e.g. regarding the minimum required safety capital that needs to be put aside in order to cover unexpected losses.  They could also be effectively used in order to better customise products to the individual needs and characteristics of each customer.  A very popular example of this is risk based pricing, which aims at adjusting the price of a product (e.g. a mortgage loan) according to the financial risk of the applicant. The risk profiles can be built both for individual customers and companies, as well as for macro-economic entities such as countries.  Two important approaches can be distinguished to infer profiles: one can make use of the information provided by external agencies and companies (e.g. FairIsaac, Experian, Moody’s and S&Ps), or one can adopt the internal ratings based approach (IRB).  The latter is the focus of this project.  In the IRB approach, companies try to build risk profiles themselves using their own knowledge and expertise.  Using data from the past and domain specific knowledge from the financial experts, they try to infer patterns which reflect the risk of the particular entities, as adequately as possible.  It is precisely in this context that Business Intelligence (BI) can be successfully adopted.  Business intelligence provides a whole range of techniques and tools aimed at extracting patterns from data.  Knowledge discovery in data (KDD) and data mining are two fundamental constituting technologies of BI.  In this project, we investigate how all these technologies can be successfully used to build risk profiles for a variety of different economic entities.  We will especially study the power and potential of business rules as a means to represent the risk profiles.

The problem of customer scoring is a very challenging and important management science problem of which financial credit scoring and customer retention scoring are the most well-known examples.  Recently, researchers have found that neural networks and support vector machines perform very well for this complex and unstructured problem when compared to more traditional statistical approaches.  However, a major drawback associated with the use of these techniques for business decision making is their lack of explanation capability.  While they can achieve a high predictive accuracy rate, the reasoning behind how they reach their decisions is not readily available.

In this project, we wish to investigate how both neural networks and support vector machines may be adapted in order to explain their reasoning behaviour by e.g. a set of classical prepositional if-then rules.  Clarifying the neural network and support vector machine decisions by explanatory rules that capture the learned knowledge embedded in these techniques can help the human experts in explaining why a particular decision is made.   Furthermore, we will also investigate how these rules can be represented in alternative ways using e.g. decision trees, decision tables en decision diagrams.  This research will be conducted using larger data sets (typically > 10.000 observations) obtained from several Benelux financial companies and one retail company.  The ultimate purpose of our research is to develop fully deployable intelligent systems that aid the scoring expert in making his daily decisions.

De doelstelling van dit project is het bevorderen van de integratie van kennis die studenten doorheen hun opleidingstraject verwerven in aanverwante opleidingsonderdelen.

Binnen de opleidingen Handelingenieur, Toegepaste Economische Wetenschappen en Beleidsinformatica zijn diverse opleidingsonderdelen aanwezig die de student toelaten om informatiesystemen in organisaties te leren beoordelen, modelleren, ontwerpen of implementeren. Gezien de omvang en de complexiteit van een volledig gefaseerd systeemontwikkelingstraject, bestudeert de student in de meeste van deze opleidingsonderdelen slechts een deel van het gehele traject ten gronde. Dit leidt tot een sterke kennisfragmentatie waardoor de integratie van kennis over opleidingsonderdelen heen voor de student sterk wordt bemoeilijkt. Van de student wordt echter verwacht dat hij/zij de kennis uit de verschillende opleidingsonderdelen met elkaar integreert om zo te komen tot inzicht in het volledige traject.

Dit MIRO-project biedt een oplossing door enerzijds een raamwerk te definiëren waarbinnen de aangeboden kennis ten opzichte van de kennis van andere opleidingsonderdelen overzichtelijk kan gepositioneerd worden. Anderzijds worden een aantal omvattende real-life probleembeschrijvingen in de vorm van cases aangeboden, gestructureerd volgens het raamwerk. In praktijk zal dit materiaal aangeboden worden op een site die gebaseerd is op WIKI-technologie, wat actieve participatie van Studenten en docenten aanmoedigt en het groeien van de site verzekert, ook na afloop van het project.

Doel van het project is de de ontwikkeling van een databank die onderzoeksmatig bevraagbaar is.  De databank zal uiteindelijk gegevens bevatten die relevant en te combineren zijn vanuit verschillende invalshoeken en onderzoeksdomeinen.  De databank moet bruikbaar zijn voor zowel innovatief wettenschappelijk onderzoek als voor dienstverlening. 

Een brede waaier aan beschikbare databanken - in electronische en gedrukte vorm - wordt samengebracht in één grote electronische databank, met als achterliggend platform een Enterprise Content Management Systeem.  Essentieel in het opzet van het Impulsproject is de ontwikkeling van een handige en gebruiksvriendelijke 'tool' voor het raadplegen van de verzamelde data.  Gezien de multidisciplinaire samenwerking en de talrijke perifere doelgroepen lijkt een webtoepassing ter aanbieding van informatie een zeer practische oplossing.  De gecollecteerde data moeten op een consistente wijze worden toegevoegd aan de databank zodat een permanente en systematische opvolging noodzakelijk is.

The aim of the INFO-NS project is to provide an objective study to the applicability of mining and decision support tools for the Belgian Federal Police (BFP). More specifically, it is studied how information retrieval, extraction and processing tools might leverage intelligence and decision support by exploiting and linking the information that is contained in vast amounts of free text material with any coexisting, structured but concise data sets that are currently in use.  The project will achieve its objective by a thorough evaluation of existing (commercial) products of leading players in the field of text and data mining applications.  Envisioning a large-scale deployment of the software, necessary and recommended changes to the existing data management infrastructure will be formulated as well.

This project fits into the course of “Object-oriented Business Modelling” and the future course “Architectures and Models of Enterprise Systems”.  The goal of this course is to familiarise students with modern software engineering methods and techniques and to let them gain insight in the interplay between management information systems and organisational business processes.  At the end of the course, the student should be capable of analysing an enterprise’s information systems requirements using the afore-mentioned methods and techniques.  The goal of the project is to develop a didactical tool that will help students when making exercises for this course.  The resulting CASE-tool should be able to provide students with interactive feedback and help them to generate a prototype from the specifications.