engleski

Possibilities of See5 software in forecasting of life expectancy not achieving

Results of application of See5 to epidemiological domain were presented. The aim of the paper was to find if See5 can be used for forecasting whether life expectancy will be achieved or not. The basis for forecasting are data - attributes, predictors (anthropological data, living habits, laboratory data, blood pressure measured in years 1970/71), and the status of examinee - class (alive or dead) in 1990. Results were given in the form of decision trees and rilesets, and tested through crossvalidation. The most interesting question - what are prediction candidates for not achieving life expectancy estimated in the year of examination - did not give a satisfactory results: the accuracy of classification was about 0.70. However, a very interesting fact is that each rule consists of at least one candidate predictor, usually considered as risk factor (e.g. high diastolic blood pressure and animal fat consumption).

See5; forecasting; epidemiology

nije evidentirano