摘要 |
PROBLEM TO BE SOLVED: To complement the missing value of a matrix when there is a missing value in non-negative value matrix decomposition, and make cluster extraction possible.SOLUTION: When there is a missing value in the elements of a non-negative matrix X indicating the number of commodities purchased by a user (commodity purchase matrix X) and a non-negative matrix Y indicating the number of commodity categories purchased by the user (category purchase matrix Y), the present invention decomposes each of the commodity purchase matrix X and the category purchase matrix Y into the product of a non-negative matrix A (user feature matrix A) indicating the relationship between the user and a cluster and a non-negative matrix B indicating the relationship between the cluster and the commodity (commodity feature matrix B) and the product of the user feature matrix A and a non-negative matrix C indicating the relationship between the cluster and the commodity (category feature matrix C), and extracts a cluster that is characterized by the user, commodity, and commodity category.SELECTED DRAWING: Figure 10 |