METHOD FOR DETERMINING LEAF NODE RANKING FOR SPATIAL PREDICTION IN DECISION TREE AND RECORDING MEDIUM FOR RECORDING SAME

摘要

Disclosed are a method for determining a leaf node ranking for a spatial prediction in a decision tree and a recording medium for recording the same. The method for determining a leaf node ranking in a decision tree includes: learning (S110) in a decision tree form having a plurality of child node relationships with one root node and a sub-node by using a training data set required for a spatial prediction; and determining (S120) a leaf node ranking from the decision tree completing the learning, wherein the each node of the decision tree in the step of learning (S110) stores a class distribution number of the training data and includes a structure for storing the class distribution number, and in the step of determining (S120) a leaf node ranking, the leaf node ranking is determined by using the class distribution number stored in each node through combining a series of nodes from the root node with the leaf node. According to the present invention, a separate process of finding an optimal parameter value is not required compared to other leaf node ranking methods, and because the ranking of the sub-node is used again in the sub-node ranking determination process, a compensation between child nodes separated from the sub-node is performed.;COPYRIGHT KIPO 2015