-
In the past three decades of ICCBR, several CBR frameworks have been developed in several
research groups. However, it is not easy, especially for young researchers, to know which CBR
frameworks are being actively developed and what features they provide for their research.
For this purpose, the ICCBR community will highly benefit from a tutorial in which the most
prominent CBR frameworks will be presented in a short time and an overview of the framework
features will be given. We, researchers and developers from four diferent universities, intend
to give a tutorial for the CBR frameworks developed at our departments. For this purpose, the
CBR frameworks CloodCBR, exit*CBR, myCBR, and ProCAKE will be presented in a tutorial
and are shortly introduced in the following.
CloodCBR [
1
] is a case-based reasoning (CBR) framework that uses a microservices’
architecture to enable the creation of distributed and highly scalable applications. CloodCBR is
developed and maintained by the AI and Reasoning (AIR) group at Robert Gordon University,
Aberdeen, UK. The framework features a variety of similarity measures for retrieval, such as
ontology-based similarity and neural network language models. The tutorial will cover how
to deploy CloodCBR in a developer-friendly manner using Docker, as well as how to create
CBR applications, configure application options, and the facilities to support the CBR cycle
(mainly the retrieval phase). Moreover, the tutorial will demonstrate how to integrate CloodCBR
into other applications using its API and its extensibility for added functionality, e.g., custom
adaptation or new similarity measures. The framework is implemented in Python and comes
with an AngularJS-based client dashboard.
eXiT*CBR [
2
] is designed with the purpose of supporting the development of intelligent
decision-making systems based on case-based reasoning (CBR). The aim of eXiT*CBR was to
go further than a collection of methods to support CBR system development but to be a tool
that supports experiment repetition and the interpretation of results by users, responding to
the need of conducting a responsible research and innovation (RRI). Once the experiments fit
the desired model for users, a running module can be generated and integrated into any other
piece of software. eXiT*CBR has several versions, including a distributed approach, and allows
the plugin of diferent modules to complement the diferent steps of CBR (as genetic algorithms
for feature learning). The tool has been implemented in Java and applied mainly to Medicine
and Healthcare (cancer prognosis, premature baby monitoring, insulin dose recommendation)
but also in industry (fault detection, adjusting printing parameters).
myCBR [
3
] is a tool for developing CBR applications in Java. It has two main components: on
the one side, the typical SDK for developers, which can be integrated into any given Java project
by importing the SDK or the myCBR.jar file. On the other side, there is an additional tool called
the “myCBR workbench”. This tool ofers a graphical user interface, split into the ‘modeling’
and the ‘casebase’ view, which allows even non-programming afine people to rapidly create
their own first CBR prototype. Using the SDK, any application can be made, such as mobile
applications, web applications, or a typical Java project. myCBR has initially been developed
by the Competence Center for Case-Based Reasoning, DFKI, Kaiserslautern, Germany, and the
Centre for model-based software engineering and explanation-aware computing, University
of West London, UK. The former, and the University of Hildesheim, Germany, are currently
using, maintaining, and upgrading myCBR with further additional functionality, such as
improved maintenance capabilities, additional knowledge representations, such as taxonomies,
and explainability. Currently, the most common use cases are structured CBR applications
using attribute-value pairs. Regarding the CBR cycle, Retrieve, Reuse and Retain are covered.
Most prominently, the award-winning CookIIS application has been developed using myCBR.
Additionally, an Intrusion Detection System using myCBR is presented on the main conference
of ICCBR 2023.
ProCAKE [
4
] – the process-oriented case-based knowledge engine – is jointly developed
by the Department of Business Information Systems II at the University of Trier and the Trier
branch of the German Research Center for Artificial Intelligence (DFKI). It supports the first
two phases of the CBR cycle, namely, Retrieve and Reuse. ProCAKE supports a wide range of
case representations, such as textual and structural representations. It has a particular focus on
process-oriented cases represented as semantic graphs. Several syntactic and semantic similarity
measures are available for similarity assessment for various data types. In addition, it is also
possible to develop custom data classes and corresponding similarity measures for retrieval.
To support the adaptation of cases, a generic adaptation manager is provided with which own
adaptation algorithms can be developed. The ProCAKE framework is written in Java, using
XML for configuration and persistence. It has a pattern-driven architecture and relies heavily
on interfaces and factories. ProCAKE has been used in many domains, e.g., for generating
cooking recipes, supporting the modeling and execution of business and scientific workflows, or
processing IoT sensor data from smart factories, and has received some prizes at the Computer
Cooking Contest of ICCBR.
References