DEFICIT VALUE PREDICTION DEVICE AND MERCHANDISE ESTIMATION DEVICE AND METHOD AND PROGRAM

摘要

PROBLEM TO BE SOLVED: To improve predicting accuracy by supplementing unobserved value data (deficit value) on a tensor.;SOLUTION: The deficit value prediction device includes: relevancy calculation means for inputting observed data showing a relation of multiple dimensions and taxonomy as background knowledge on each dimension, and for calculating relevancy between the observed data and unobserved data; tensor construction means for constructing a tensor expressing the relation of the multiple dimensions by using the calculated relevancy with respect to the unobserved data as well as the observed data; and a deficit value estimation means for estimating a deficit value with respect to the unobserved data by decomposing and restoring the tensor by a tensor decomposition method.;COPYRIGHT: (C)2014,JPO&INPIT