A genetic fuzzy classifier to adaptive user interest profiles with feature selection.
Marı́a J. Martı́n-Bautista, M.A. Vila, Henrik Legind Larsen · 1999
We propose a method for guiding genetic algorithms for information retrieval by fuzzy classification and a genetic feature selection process of terms from documents evaluated by the user. The fuzzy classifier implements an inductive derivation of the current, experience based, interest profile in terms of an importance weighted conjunction of genes. A gene is defined by a symbol and a fuzzy number of occurrences of the symbol in documents belonging to the class of documents that satisfy the user’s information need. Once the classification of the documents is made, a genetic selection from the most discriminatory terms is carried out. In this way, the terms that allow the system to discern between good and bad documents are selected and stored as a part of the user’s profile to be used in future queries to the system. The fuzzy classification and term selection processes provide a better utilization of valuable knowledge for genetic algorithms in order to get an improvement of the quality of the estimates of the current and near future information needs in the areas of interest to the user.