Decision tree generating apparatus, decision tree generating method, non-transitory computer-readable recording medium, and inquiry system

A decision tree generating method of a decision tree generating apparatus, comprising:

• when a classification target data set including a plurality of pieces of classification target data respectively having mutually different attributes with attribute values assigned thereto is hierarchically segmented into a plurality of subsets in a form of a decision tree, calculating an amount of reduction in entropy of the pre-segmentation classification target data set caused by segmentation on each attribute of each piece of the classification target data included in a pre-segmentation data set, and calculating an information gain when the pre-segmentation data set is segmented in accordance with the attribute value of each attribute, based on the amount of reduction in the entropy and reliability that is an index representing correctness or incorrectness of a user's answer responsive to an inquiry asking about the attribute, and
• successively determining an attribute having a maximum information gain to be a node of the decision tree by recursively iterating the segmentation of the pre-segmentation data set in accordance with the attribute value of the attribute having the maximum information gain from among the information gains calculated on the attributes, and generating the decision tree that is to be used to determine an order of inquiries asking about the attributes in order to classify the pieces of classification target data by successively assigning the attribute value of the attribute having the maximum information gain to an edge of the node.

A non-transitory computer-readable recording medium storing a program causing a computer to execute a process for generating a decision tree, the process comprising:

• when a classification target data set including a plurality of pieces of classification target data respectively having mutually different attributes with attribute values assigned thereto is hierarchically segmented into a plurality of subsets in a form of a decision tree, calculating an amount of reduction in entropy of the pre-segmentation classification target data set caused by segmentation on each attribute of each piece of the classification target data included in a pre-segmentation data set, and calculating an information gain when the pre-segmentation data set is segmented in accordance with the attribute value of each attribute, based on the amount of reduction in the entropy and reliability that is an index representing correctness or incorrectness of a user's answer responsive to an inquiry asking about the attribute, and
• successively determining an attribute having a maximum information gain to be a node of the decision tree by recursively iterating the segmentation of the pre-segmentation data set in accordance with the attribute value of the attribute having the maximum information gain from among the information gains calculated on the attributes, and generating the decision tree that is to be used to determine an order of inquiries asking about the attributes in order to classify the pieces of classification target data by successively assigning the attribute value of the attribute having the maximum information gain to an edge of the node.