Automated observational decision tree classifier
AssigneeSap SE
InventorFlorian Hoffmann
Life of this patent
Estimating the term…
Abstract
Various embodiments of systems and methods for automatic classification of objects in a computer system are described herein. A class decision is received, where the class decision is a classification of an object from a number of objects to a class from a number of classes. The class decision is classification of the object by a role model. An exploration tree is expanded based on the class decision. A decision tree is constructed based on the exploration tree. Objects are classified based on said decision tree.
Claims17
A computer implemented method to automatically classify a plurality of objects in a computer system by observing a role model, the method comprising:
- receiving an exploration tree, wherein:
- said exploration tree represents a plurality of view sequences associated with the role model,
- a plurality of nodes of the exploration tree represents a plurality of views, a view represents features from a plurality of features of said plurality of objects, and
- a plurality of leaves of the exploration tree represents a plurality of classes;
- assessing an object in a view sequence of said plurality of view sequences by said role model to submit a class decision, wherein said view sequence is succession of one or more views of said plurality of views, said one or more views are assessments of one or more subsets of features of said object in order of assessment;
- receiving said class decision, wherein said class decision is a classification of said object of the plurality of objects to a class of said plurality of classes;
- the computer expanding said exploration tree based on said class decision by:
- merging said view sequence from said plurality of view sequences to said exploration tree;
- the computer constructing a decision tree based on said exploration tree;
- and
- classifying said plurality of objects to said plurality of classes based on said decision tree.
An automated classifier computer system to automatically classify a plurality of objects in the computer system by observing a role model, the system including:
- at least one processor and memory to execute program code related to:
- receive an exploration tree, wherein:
- said exploration tree represents a plurality of view sequences associated with the role model,
- a plurality of nodes of the exploration tree represents a plurality of views, a view represents features from a plurality of features of said plurality of objects, and
- a plurality of leaves of the exploration tree represents a plurality of classes;
- assess an object in a view sequence of said plurality of view sequences by said role model to submit a class decision, wherein said view sequence is succession of one or more views of said plurality of views, said one or more views are assessments of one or more subsets of features of said object in order of assessment;
- receive said class decision, wherein said class decision is a classification of said object of the plurality of objects to a class of said plurality of classes;
- expand said exploration tree based on said class decision by merging said view sequence from said plurality of view sequences to said exploration tree;
- construct a decision tree based on said exploration tree;
- and
- classify said plurality of objects to said plurality of classes based on said decision tree.
A non-transitory computer readable medium storing instructions thereon, which when executed by a processor cause a computer system to:
- receive an exploration tree, wherein:
- said exploration tree represents a plurality of view sequences associated with the role model,
- a plurality of nodes of the exploration tree represents a plurality of views, a view represents features from a plurality of features of a plurality of objects, and
- a plurality of leaves of the exploration tree represents a plurality of classes;
- assess an object from said plurality of objects in a view sequence of said plurality of view sequences by said role model to submit a class decision, wherein said view sequence is succession of one or more views of said plurality of views, said one or more views are assessments of one or more subsets of features of said object in order of assessment;
- receive said class decision, wherein said class decision is a classification of said object of the plurality of objects to a class of said plurality of classes;
- expand said exploration tree based on said class decision by:
- merging said view sequence from said plurality of view sequences to said exploration tree;
- construct a decision tree based on said exploration tree;
- and
- classify said plurality of objects to said plurality of classes based on said decision tree.
Dependencies are inferred from the claim text.