发明名称 |
Method for constructing a tree of linear classifiers to predict a quantitative variable |
摘要 |
A method is provided for predictive modeling of a physical process in order to determine and implement a solution to the physical process. The method includes storing representations of relevant physical relationships, observations, and outcome probabilities related to the physical process in a memory of a computer, recursively analyzing the stored representations by the computer, and generating at least one tree structure that models the physical process, the tree structure including at least one root node, a plurality of decision nodes, and a plurality of end nodes. The method also includes determining a solution to the physical process based on criterion variables, by the computer using the tree structure of the physical process to arrive at an end node, and implementing a physical solution to the physical process, the physical solution being based on the determined solution. |
申请公布号 |
US9111213(B2) |
申请公布日期 |
2015.08.18 |
申请号 |
US201414151131 |
申请日期 |
2014.01.09 |
申请人 |
AT&T INTELLECTUAL PROPERTY I, L.P. |
发明人 |
Rotondo John |
分类号 |
G06N7/00;G06N99/00;G06N5/04 |
主分类号 |
G06N7/00 |
代理机构 |
Greenblum & Bernstein, P.L.C. |
代理人 |
Greenblum & Bernstein, P.L.C. |
主权项 |
1. A method for using predictive modeling of a physical process in order to determine and implement a solution to the physical process, the method comprising:
storing representations of relevant physical relationships, observations, and outcome probabilities related to the physical process in a memory of a computer; recursively analyzing the stored representations by the computer, and generating at least one tree structure that models the physical process, the tree structure including at least one root node, a plurality of decision nodes, and a plurality of end nodes,
wherein the analyzing and generating include partitioning observations at each decision node into probable outcomes using target partitioning, and generating a plurality of other decision nodes based on the target partitioning, andwherein when the target partitioning of a decision node does not achieve a predetermined criteria, converting the decision node into an end node; determining a solution to the physical process based on criterion variables, by the computer using the tree structure of the physical process to arrive at an end node; and implementing a physical solution to the physical process, the physical solution being based on the determined solution. |
地址 |
Atlanta GA US |