Classification method of data pattern and classification system of data pattern

Abstract

To easily find columns containing information to be shared between databases found when the databases (DBs) are combined. A data-pattern classification method for a database containing character strings regularly overlapping between the input values of mixed columns and information inputted in another specific column, the mixed column including types of information, the method including: extracting means for extracting a branch column for changing the type of information to be inputted to the mixed column, the branch column being extracted by one of a heuristic first technique and a second technique, the first technique using timing to change the pattern of the mixed column in the database including the mixed column, the pattern referring to a column where a character string overlaps the input value of the mixed column, the second technique using a statistical technique of a likelihood test; and classification to obtain the number of types of information stored in the mixed columns, by grouping, according to information indicated by the patterns, the patterns obtained from the mixed columns based on the extracted branch column.