[Prognosis for life expectancy using the inductive learning method]

The aim of this paper was to find out the possibility of life expectancy achievement (LEA) prognosis in open population by using epidemiological data, to improve it and to determine the differences between regions. We were using inductive learning software tool, ASSISTANT Professional, based on modified Quinlan's inductive learning method. Data from an epidemiological study of oil/fat consuming influence on diabetes incidence in different regions in Croatia, 1887 examinees, have been used. In spite of limited number of attributes that were available, an improvement in prognosis of LEA has been made by selecting the proper attributes and by changing the limiting values of attributes (values that change the meaning of attributes). We reached the absolute accuracy of 75.54%. Even after further pruning of decision tree that lowered this value, Assistant showed 13% better result than those of random selected outcome. Differences between regions were also established that could not been explained with attributes that were used. Expanding the list of attributes and analysis of their influence in particular region can make further improvement in prognosis of LEA.