RULE DISCOVERY SYSTEM, METHOD, APPARATUS AND PROGRAM

摘要

A system, an apparatus, a method and a program are provided which render it possible to obtain with high efficiency a set of rules useful in grasping or correcting contents of a database. The system includes a free itemset generation unit (21) that generates a set of free itemsets, each being made up of an attribute-value pair, a frequency of the free itemset in the database being greater than or equal to a predetermined threshold value, a rule candidate generation unit (22) that generates as a rule candidate, a rule having a conditional part set to the free itemset α, having a consequent part set to an item x not sharing an attribute with the free itemset, and finds a set of attributes of an antecedent part of the rule by depth first search, the attribute not included in neither α nor x, a validity decision unit (23) that collates the rule to the database to decide whether or not the rule is valid, and a rule minimality decision unit (24) that checks for minimality of the rule decided to be valid to output the rule to an output device (4), when the rule is minimal.

主权项

1. A rule discovery system comprising:
a storage device configured to store a database; a data processor; and an output device, wherein the data processor includes: a free itemset generation unit configured to generate a free itemset formed by an item that is an attribute-value pair in the database, a frequency of the free itemset in the database being not less than a predetermined threshold value; a rule candidate generation unit configured to generate, as a rule candidate, a rule having a conditional part set to the free itemset, having a consequent part thereof set to an item not sharing an attribute with the free itemset of the conditional part, and having an antecedent part thereof set to an attribute included neither in the free itemset of the conditional part nor in the item of the consequent part, the rule candidate generation unit retaining the rule generated in a storage unit; a rule validity decision unit configured to collate the rule generated by the rule candidate generation unit to the database to decide the rule to be valid when the rule is matched with a confidence greater than or equal to a pre-set threshold value of the confidence; and a rule minimality decision unit configured to decide whether or not the rule decided to be valid by the rule validity decision unit is minimal, the rule minimality decision unit outputting the rule to the output device when the rule is decided to be minimal.