Optimisation par algorithmes de clustering de la construction automatique de bases de connaissances floues

This master's thesis presents several applications of clustering algorithms for the automatic generation of fuzzy knowledge bases using genetic algorithms. Four main objectives are targeted: execution time, robustness, performance and autonomy.

Automatic knowledge base generation techniques such as genetic algorithms tend to be highly dependent on the quality and size of the learning data. First of all, large data sets can lead to unnecessary time loss, when smaller data sets could describe the problem as well. To address this issue, data are compressed by reducing similar and redundant information. Second of all, the presence of noise and outliers can lead the learning algorithm to degenerate. Clustering techniques allow the filtering of the data, thus making the generation of fuzzy knowledge bases more accurate.

The third and fourth objectives are accomplished by addressing a well known issue surrounding fuzzy knowledge base generation using genetic algorithms: finding an optimal number of fuzzy sets for each premise. Some of the current genetic algorithm This work proposes solutions based on cluster analysis and validation indices for the numbers of clusters used to predefine the numbers of fuzzy sets. (Abstract shortened by UMI.)