Mireia Castanys Homepage
N�ria Maci� » Research


Case-based reasoning (CBR) systems solve new problems through an analogical procedure based on experiences represented by a set of cases stored in a case memory. As the case memory feeds this process, its size and organization plays an important role in the CBR performance in terms of computational time and accuracy. This is especially critical in real world problems because most of the them deal with a huge amount of data, which are often unreliable and uncertain. This is due to the fact that reality is always complex and, moreover, it may be partially unknown. Therefore, the features of the aforementioned domains (e.g. melanoma cancer) may negatively affect the cases recovery in the case memory:

* The uncertain cases may confuse the system. This issue may be critical in applications in which the error cost is high. For instance, in medicine, the error cost may be a human life.
* The response time may be slow if there are a lot of cases. This aspect may be critical in real time applications in which a high response time in needed.


The Soft CBR area is a group of 12 members composed by Phds, Phd students and external collaborators. The research lines of the area divided in four interrelated goals:

1) Organization of the CBR case memory through Soft-Computing techniques like Self-Organizing Maps, Evolutionary Computation or Fuzzy Logic.
2) Helping experts to understand why a solution is proposed.
3) Combination of multiplesource experiences to improve the reliability of the solution.
4) Development of data mining tools like plugins for jColibri framework, ANALIA or for the management of the melanoma domain.

Heads of the area:
Dra. Elisabet Golobardes (elisabet@salle.url.edu)
Dr. Albert Fornells (afornells@salle.url.edu)

Dr. Carles Garriga (cgarriga@salle.url.edu)
Dra. Elisa Martínez (elisa@salle.url.edu)
Dra. Mireia Castanys (mcastanys@salle.url.edu)
Guiomar Corral (guiomar@salle.url.edu) (Phd presentation in September’09)
David Vernet (dave@salle.url.edu)
Rubén Nicolás (rnicolas@salle.url.edu)
Álvaro García (alvarog@salle.url.edu)
Marc Aguilar (maguilar@salle.url.edu)

External members:
Dr. Fernando de la Torre (ftorre@cs.cmu.edu) (Carlingne Mellon, EUA)
Dr. Josep Maria Martorell (j.martorell@fundacio.upc.edu) (UPC, Barcelona)


The research activity of the area is mainly focused in the MID-CBR project (TIN2006-15140-C03-03) leaded by Dr. Enric Plaza. The goal is to define an integrative framework for the development of case-based Reasoning systems. The team project is composed by researchers from IIIA research center belonging to the Spanish Council for Scientific Research (CSIC), from GAIA at Complutense University of Madrid (UCM) and from us.

The main objectives of the project can be summarized as follows:

1) new ways to use techniques of soft computing for CBR,
2) techniques for case reuse of a declarative and generic nature,
3) techniques for case retrieval in knowledge-intensive CBR systems,
4) integrating ontologies both in CBR systems and retrieval and reuse techniques,
5) maintenance techniques both for case bases and for CBR systems capable of dealing with issues arising from design, implementation, and deployment of industrial strength CBR systems,
6) the empirical evaluation of the developed techniques by means of CBR prototypes implemented for several experimental domains, and
7) developing component-based software platforms to support CBR systems development.


Our research is mainly focused on three application domains: Melanoma cancer diagnosis, detection of vulnerabilities in telematic networks and education.

Melanoma cancer diagnosis
Skin is mainly divided in three layers as picture shows. The lowest part is called dermis and it contains cells called melanocytes responsible of the production of melanin, which give the skin pigmentation. Melanoma cancer appears when melanocytes turn cancer cells, it means, they are able to invade other organs because they grow and reproduce in an uncontrolled way. Althoug it is not the most common skin cancer, it is which causes most deaths. This increase, caused by solar habits, makes crucial the early diagnosis, even more if we analyze that this cancer is mortal in approximately 20% of cases and prompt diagnosis permits practically a secure regain.

One of the most used techiques to the diagnosis of melanoma is the ABCD rule which considers four clinical features commonly observed in this kind of tumour: asymmetry, border irregularity, colour variegation, and a diameter larger than 5 mm. Although most of melanomas are correctly diagnosed following this rule, a variable proportion of melanomas does not comply with these criteria. The current procedure when a suspicious skin lesion appears is to excise and to analyse it by means of biopsy. Commonly, the result of the biopsy allows to determine the accurate malignity of the lesion. For this reason, we work with experts from the melanoma unit at the Hospital Clínici Provincial de Barcelona in order to help them to improve the early diagnosis and avoid unnecessary byopsis.

Detection of vulnerabilities in telematic networks
Nobody conceives a competitive organization without all its resources interconnected and continuously available. As organizations have become increasingly dependent on their networks and Internet, they have also opened themselves up to potentially immense risks and vulnerabilities. Consequently periodically audits and vulnerability assessments are needed to detect and eliminate these possible security holes. Vulnerability assessment is the process of identifying and quantifying vulnerabilities in a system or a network. However, time and cost can limit its depth. These limitations justify the automation of the processes involved in a vulnerability assessment, especially those related to the analysis of test results, as a thorough network security test generates large data quantities.

We collaborate with security experts from ISECOM and with experts from the telematic department of our university.We focus on helping experts to speed up and improve the reliability of their security diagnosis by tackling this problematic through the development of two crucial tasks:

1) Network testing. CONSENSUS is a tool developed for gathering data related to vulnerability aspects from computer.
2) Analysis of testing results. ANALIA is a tool developed for helping expert to analyse and diagnose the level of vulnerability of new computers.