CLASSIFICATION ESTIMATION DEVICE FOR ESTIMATING RATIO ON EACH CLASSIFICATION IN ENTIRE MEMBER OBJECTS IN A GROUP, PROGRAM AND METHOD

摘要

PROBLEM TO BE SOLVED: To provide a classification estimation device capable of calculating the ratio of each classification j without identifying the kind of class (attribute) of each member objects in estimation target group.SOLUTION: Each of member objects has a member identifier i and a feature amount fof each member in plural elements k(k=1-K). The classification estimation device includes: a target feature amount distribution calculation unit that calculates a probability distribution function Pof each element k from entire member objects of a group; a learning feature amount accumulation unit that accumulates a probability distribution function Pof a learning feature amount of each element in each class j; a weight determination unit that determines a weight wto the class learning feature amount distribution for each class j so that the inter-distribution distance d(&Sgr;wP, P) between a weight sum &Sgr;wPof the learning feature amount distribution of each class and the target feature amount distribution Pis the minimum under the condition of w≥0; and a class normalization unit that normalizes the weight wof each class j and calculates the ratio of the class j.