Intelligent Filtering with Genetic Algorithms and Fuzzy Logic
María-Amparo J. Martín-Bautista, M.A. Vila, Daniel Sánchez, Henrik Legind Larsen · Studies in fuzziness and soft computing · 2002
We present two different approaches combining fuzzy information retrieval of documents with genetic algorithms, and the pre-processing stage of classification called feature selection. The differences between these approaches lie basically in the target of the fitness function selected. In the first approach, the Term-Oriented Model, the fitness function is based on a measure to find the most discriminatory terms, by rewarding not only the terms from the good documents, but also those from the bad ones, if they are considered as good partial classifiers. However, the aim of the Document-Oriented Model, as traditionally, is to rank the documents by relevance. So, the best chromosome represents the optimal query. The fuzzy weighting scheme used in this model considers also the discriminatory terms by introducing the knowledge about the user preferences in the genes, but rewarding the genes belonging to the good documents. These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.