摘要:
Disclosed is a method for assigning features to nodes of a tree structured classifier and for determining terminal nodes in response to a training set of objects, each of such objects being determined by a plurality of features. The method comprises the steps at each node of the tree of: (1) determining a selected characteristic, such as a cost function based on the minimum description length, of the plurality of features unused at prior nodes along the path from the root to the present node; (2) assigning a feature to the node having a preferred value for the selected characteristic relative to the other features; (3) creating child nodes in response to the assigned feature; (4) for each child node, determining the selected characteristic for the plurality of features unused at prior nodes and assigning a feature to the child node having a preferred value for the selected characteristic relative to the other features; (5) generating a combination of the values for the selected characteristics of the assigned features for the child nodes of the node; and (6) classifying the node as a terminal node in response to a comparison of the combination of values for the features assigned to the child nodes and the value for the feature assigned to the node.
摘要:
Disclosed is a method for assigning features to nodes of a tree structured classifier and for determining terminal nodes in response to a training set of objects, each of such objects being determined by a plurality of features. The method comprises the steps at each node of the tree of:
(1) determining a selected characteristic, such as a cost function based on the minimum description length, of the plurality of features unused at prior nodes along the path from the root to the present node; (2) assigning a feature to the node having a preferred value for the selected characteristic relative to the other features; (3) creating child nodes in response to the assigned feature; (4) for each child node, determining the selected characteristic for the plurality of features unused at prior nodes and assigning a feature to the child node having a preferred value for the selected characteristic relative to the other features; (5) generating a combination of the values for the selected characteristics of the assigned features for the child nodes of the node; and (6) classifying the node as a terminal node in response to a comparison of the combination of values for the features assigned to the child nodes and the value for the feature assigned to the node.