Naslov
An inductive learning method versus the common statistical approach to outcome forecasting

Autori
Kern, Josipa ; Težak-Benčić, Marija ; Vuletić, Silvije ; Deželić, Gjuro ; Kujundžić, Mirjana

Izvornik
Collegium antropologicum (0350-6134) 15 (1991), 2; 283-289

Vrsta, podvrsta i kategorija rada
Radovi u časopisima, članak, znanstveni

Ključne riječi
Inductive learning method ; statistical approach ; outcome forecasting

Sažetak
Decision tree construction through inductive learning method and common statistical techniques, i.e. discriminant analysis and Bayesian approach, are compared. The problem example illustrating results of all the procedures was the gestationol age which had to be forecasted on the basis of variables describing not only the current pregnancy but also the previous ones. The best result, the highest prognostic accuracy (61.31% as absolute and 53.06% as relative one), was achieved by using the combination of two methods: decision tree construction and Bayesian approach as »First Tree Than Bayes« approach. This result can be considered as much more valuable, taking into account the fact that selected predictor system is not very close to the outcome which has to be predicted.

Izvorni jezik
Engleski

Znanstvena područja
Matematika, Interdisciplinarne prirodne znanosti, Javno zdravstvo i zdravstvena zaštita

POVEZANOST RADA