Symbolic data analysis proposes a general framework to extend usual data an
alysis methods to more complex data called symbolic objects. The prediction
problem for symbolic objects is defined: it is seen to be a generalization
of the prediction for standard data. An algorithm of tree-growing is devel
oped for probabilistically imprecise data. The new algorithm is presented a
s a procedure for extracting knowledge from data of a more general type tha
n standard data. Two data sets, respectively, based on categorical and cont
inuous variables, are treated in detail. (C) 2000 Elsevier Science B.V. All
rights reserved.