Research projects

Semantic Concept Modeling for Digital Image Characterization:
Application to Linguistic Description and Visual Information Retrieval

Modelado de conceptos sem�nticos para la caracterizaci�n de im�genes digitales:
Aplicaci�n a la descripci�n lingu�stica y recuperaci�n de informaci�n visual

Lazy Objects in Data Integration

Representaci�n y manipulaci�n de objetos imperfectos en problemas de integraci�n de datos

Anomaly detection and subgroup discovery in Databases:
Application to scientific experimentation

Detecci�n de anomal�as y subgrupos en bases de datos: Aplicaci�n a la experimentaci�n cient�fica

Semantic management of document data with Soft Computing techniques:
Application to authority control

Gesti�n sem�ntica de datos documentales mediante t�cnicas de Soft Computing: Aplicaci�n al control de autoridades

MINDAT-PLUS:
Data Mining for users in different application domains

MINDAT-PLUS: MINeria de DATos Para Los USuarios en diferentes �reas de aplicaci�n

Information and Knowledge Fusion
in Hospital and Medical Emergencies Environments

"Fusi�n de Informaci�n y Conocimiento en el �mbito Hospitalario y de Urgencias M�dicas"

Soft Data Server:
Soft-computing-based Object-Relational Database Server Model

"Soft Data Server: Modelo de servidor de bases de datos objeto-relacionales basado en soft computing"

An Intelligent System for Decision Making
in the Agricultural Environment of the Spanish Southeast

"Sistema Inteligente para la Toma de Decisiones en el entorno Agr�cola del Sudeste Espa�ol"

MSc. Theses

Frequent Tree Pattern Mining

Identificaci�n de Patrones Frecuentes en �rboles

Class-Complexity-Sensitive Models

Modelos Sensibles a la Complejidad por Clase

Analysis of Train Traffic Flow Using Non-Supervised Data Mining Techniques

Model-Based Language Specification

Especificaci�n de Lenguajes basada en Metamodelos

Software Development Projects

TMiner

TMiner collects some Data Mining algorithms which might prove useful to interpret huge datasets and discover associations among the data stored in databases. TMiner is implemented in Java and works on relational databases through JDBC, the Java standards CLI [call-level interface].

TMiner allows the user to perform standard SQL quesries, import and export data, build classification models, and extract association rules using Apriori or TBAR (Fernando Berzal, Juan Carlos Cubero, Nicol�s Mar�n & Jos� Mar�a Serrano: "TBAR: An efficient method for association rule mining in relational databases", Data & Knowledge Engineering, vol. 37, pp. 47-64, 2001). TMiner lets you build classification models such as decision trees (as those built using ID3 or C4.5). It also incorporates other machine learning techniques such as rule inducers (AQ and CN2) and algorithms to build partial classification models using association rules.

Numerical Cruncher

Numerical Cruncher [NC]] is a Java 2 applet/application which collects some of the most important algorithms used in Pattern Recognition. This software package includes several classification methods, both parametric (linear and quadratic, for example) and non-parametric (k-NN, LVQ, or DSM). NC also implements some clustering techniques (such as the Batchelor-Wilkins' algorithm, the ubiquitous K-Means, a GRASP algorithm which improves K-Means, and the more complex ISODATA). Numerical Cruncher is bilingual (English and Spanish) and includes HTML help for both English and Spanish.

NC is able to read ASCII files (as C4.5 or LVQ_PAK), as well as raw images. It can also access to local or remote databases using JDBC JDBC [Java DataBase Connectivity], the standard database access interface for the Java platform. When NC is run as an applet, it can obtain input data only through JDBC (due to Java security constraints). Numerical Cruncher seamlessly integrated with TMiner Personal Edition.

SEA: Software Engineer Assistant

Integrated version control, defect tracking, and time accounting system aimed to facilitate software development project management. Originally developed in C++Builder, it stores all relevant data in a relational database (InterBase) and is able to work with zipped files.

SEA version control or software configuration management subsystem allows, among other things, to categorize software componentes (in order to foster reuse), specify relationships and dependencies among components, and automatically restore the project state at a given date, as well as the usual check-in and check-out operations.

SEA defect tracking subsystem manages software bugs and error reports, maintenance requests and software-related risks. Each SPR [Software Problem Report] can be linked to particular components and versions stored by the version control subsystem, thus speeding up the management of all the information related to software development projects.

Finally, SEA human resource subsystem can be used to keep a time log, which is an unvaluable resource for time accounting, invoicing, and future project cost and effort estimation.

I²R: Intelligent Information Retrieval

Information retrieval system based on SIRE [Scalable Information Retrieval Engine]. It uses an inverted index to answer users' queries and allows users to experiment with different criteria to match their queries against the documents in the database. It can also employ stemming and stop words. I2R is able to index files (in ASCII or HTML format) as well as the content of any relational database.

A set of Java servlets lets I2R integrate with existing document databases and deploy a complete information retrieval system for digital libraries, where documents can be found by search (using keywords and boolean queries) or exploration (browsing category hierarchies). The I2R system is even able to suggest related categories without human intervention.

Berzal @ Imagen

Berzal@Imagen is an open-source application which collects some of the most-widely used algorithms in Image Processing and Computer Vision: image editing, smoothing, sharpening, histogram manipulation, equalization, contrast enhancement, morphology, segmentation, edge and line detection, several filters, Fourier and Hough transforms, wavelets, etc.. All the algorithms are implemented in ANSI C and a graphical user interface was developed using Borland C++Builder. Additional information regarding to the collected algorithms can be found in the following references:

• A. K. JAIN: "Fundamentals of Digital Image Processing", Prentice Hall, 1989.
• J. R. PARKER: "Algorithms for Image Processing and Computer Vision", John Wiley & Sons, 1997

DOS DBMS

Small dBase-like program for MS-DOS which allows you to manage your own databases (creation, updates, efficient search, even a restricted form of referential integrity). The libraries which manage the data files are written in ANSI C and have been successfully used in several projects. The program, as it is freely available, was compiled using the Watcom C/C++ compiler.