To provide the method for modeling variables of a data set through a probabilistic network including data nodes and casual links.

Variables of input data are registered, and groups of solid bodies consisting of plural genomes are generated and each have input data sets modeled individually and separately. Each genome has a chromosome representing a data node in the probabilistic network and a chromosome representing a cause and effect link between data nodes. Cross-recombining operation is performed between chromosome data of the parent genomes in the solid body groups to generate descendant genomes, which are added to the genome solid body groups. For the genomes in the solid body groups, scoring showing matches between the genomes and input data is performed, genomes are selected from the solid body groups, and cross-recombination, scoring, addition, and selection are carried out for plural genome generations, thereby selecting genomes from the final generation according to the most suitable scores.