Research project:
Methodologies based on Data Mining for the design of Intelligent Hybrid Systems (M2D2-SHI)


                                                                                                                                                 REF: TIN2008-06872-C04-03

                                                                                                                                                 Ministerio de Ciencia e Innovación      Fondo Europeo de Desarrollo Regional


This project was conceived to delve deeper into the integrated use of various intelligent components in a useful and effective aid system in decision-making. Intelligent systems for aided decision-making need to take advantage of Soft Computing methodologies to integrate knowledge extraction techniques with intelligent optimization strategies in the assessment and management of alternatives using information which is characterized by some degree of uncertainty. Methods for assessing and managing solutions, guidelines for the design of intelligent optimization strategies, models for managing information with uncertainty and data mining techniques to extract relevant information from each process are all important in this context. Thus, the project aims at deepening knowledge in four clearly defined research areas (corresponding to teams of the universities of Granada (UGR), La Laguna (ULL), Murcia (UMU) and Valencia (UV)) and the integration of the products from each one of them.

  • 1) Interactive methods for the assessment and ranking of alternatives with linguistic valuations (UGR)
  • 2) Integration of data mining methodologies in hybrid optimization systems (UMU)
  • 3) Cooperative hyper-heuristics based on heterogeneous search agents (ULL)
  • 4) Integration of information with uncertainty in the optimization and assessment of alternatives (UV)
  • 5) Integration of Intelligent Systems (Coordinated Project)
However, each one of these fields must not only act in conjunction with the others in the design of Intelligent Systems for aiding in Decision Making in real contexts but each of them benefits by the integration of advances in any of the other fields in the project. Thus, in the design and implementation of interactive Decision Support Systems it is important to include data mining tools. These tools allow relevant information to be extracted and used, even if this information is characterized by some uncertainty, in the optimization process for the assessment of available alternatives. The hybridization of data mining techniques and optimization systems incorporates the approaches found in solution ranking methods using information from the uncertainty with a cooperative approach within the optimization process. On the other hand, one of the most advanced tendencies found in the optimization of intelligent systems is the design of hyper-heuristics as an evolution of cooperative meta-heuristics strategies. These hyper-heuristics are based on a system of heterogeneous search agents where machine learning techniques for the selection and the use of this exchanged information among agents, in addition to the ranking of heuristic methods for the selection in the search process and the management of uncertainty, are essential issues when seeking to improve their performance. The models for managing uncertainty are especially important in association with interactive decision support systems, incorporating data mining techniques in optimization systems and the implementation of cooperative hyper-heuristic strategies.

The performance of the designed tools and effectiveness of the respective advances will be studied and applied in important fields related to the efficient use of resources. These fields of activity include: financial investment, personnel management, evaluation of university quality and logistic planning. Portfolio selection and similar economical and financial problems are treated in the area of investment. We will also tackle the management of human resources by competencies, such as templates and optimal staff selection. Studies in logistic planning will take into account transport problems and the design of distribution networks by solving location, routing and load problems.