Extensions of Logical Analysis of Data for Medical Applications
Résumé
Logical Analysis of Data (LAD) has already been applied on several biomedical applications [Bonates and Hammer, 2006]. In this talk we shall present two developments designed to cope with new requirements. For a problem of height-loss due to irradiation during childhood, we shall show how Combinatorial regression - an extension of LAD to numerical outputs (instead of the classical binary ones) - allowed to precisely characterize the influences of several variables for a problem of height-loss due to irradiation during childhood. For the diagnosis of growth hormone deficiency, there is a need for very simple (e.g., straightforwardly usable by a GP) and accurate classifiers. We shall show an the concept of Functional patterns allowed to meet this requirement by aggregating many Boolean patterns into one single (non-binary) condition.