Search

FuzzyPred is an unsupervised-learning data-mining algorithm, which allows extracting fuzzy predicates in normal forms from the data. This method is used to solve a descriptive task where it is not known the kind of relationships that are to be found. The goal is to find patterns that describe the data and their relationships. Due to the large set of solutions or search space that may have, it was modeled as an optimization problem, where metaheuristics are applied to find good solutions. FuzzyPred provides as a result a set of predicates, evaluated in each of the quality measures, in spite of the fact that only one of the measures is optimized (truth value). This paper introduces a multiobjective approach for FuzzyPred by using two of the most known multi-objective optimization techniques: Pareto technique (or pure multi-objective) and weighted factors. An experimental study is presented in order to compare the efficacy of both techniques in this problem. The results in several international databases show that better results are obtained by the pure multi-objective technique. [ABSTRACT FROM AUTHOR]

In this paper, the design of a genetic algorithm that modifies and recalculates the values those describe the membership functions for a fuzzy controller, has showed. The controller is applied in a bio-film fermentative process to obtain an improvement in the dichromate's production (DCM). The membership functions act over the concentration of substrate in the aeration tank so that produces the homogenous double saturation kinetic of the DCM, and this is increased, obtaining an improvement in their production. This is shown by development of the numerical algorithm. [ABSTRACT FROM AUTHOR]