We describe early machine learning research in Ljubljana, motivated by medical diagnostic problems, in the areas of building decision trees with Assistant, the development of Naïve and Semi-Naïve Bayesian classifier and its explanations of individual predictions, and the development of ReliefF and RReliefF algorithms for non-myopic evaluation of attributes in classification and regression, respectively. Povzetek: V članku opišemo zgodnje raziskave na področju strojnega učenja v Ljubljani, ki so

more »

... bile motivirane z medicinskimi diagnostičnimi problemi. Razvili smo sistem Asistent za gradnjo odločitvenih dreves, naivni in delno naivni Bayesov klasifikator in metodo razlage njunih napovedi ter algoritma ReliefF in RReliefF za nekratkovidno ocenjevanje atributov v klasifikaciji in regresiji.