Back to results
US 8,914,366 B1Granted
Evaluating clustering based on metrics
AssigneeGoogle Inc.
InventorXin Li
Life of this patent
Estimating the term…
Abstract
Clusters generated by a first clustering process are received, each cluster including one or more related records. A second clustering process is applied to the received clusters. Received clusters generated by the first clustering process are compared with results of the second clustering process. Information is output based on the comparison for evaluating the first clustering process. A tendency of the first clustering process to duplicate a record associated with a single entity in multiple clusters can be determined. A tendency of the first clustering process to cluster records associated with different entities in a single cluster can be determined.
Claims27
A method executed by a computer, comprising:
- receiving first clusters generated by a first clustering process, each first cluster including one or more records;
- applying a second clustering process to the received clusters to generate second clusters, where the second clustering process differs from the first clustering process;
- comparing the first clusters generated by the first clustering process to the second clusters generated by the second clustering process, the comparing including:
- determining a duplication metric, the duplication metric measuring mergers of records that belong to separate first clusters into a second cluster;
- and
- determining an overclustering metric, the overclustering metric measuring splits of a first cluster into two or more of the second clusters;
- and
- displaying results of the comparison for evaluating and modifying the first clustering process.
5 more elements
A method executed by a computer, comprising:
- receiving first clusters generated by a first clustering process, each cluster including one or more records;
- applying a second clustering process to the received clusters to generate second clusters, where the second clustering process differs from the first clustering process;
- comparing the first clusters generated by the first clustering process to the second clusters generated by the second clustering process, the comparing including:
- determining duplication metric, the duplication metric measuring mergers of records that belong to separate first clusters into a second cluster, including:
- calculating a first difference value that measures a difference between a first number of the first clusters and a second number of the second clusters;
- and
- determining the duplication metric based on the first difference value;
- determining an overclustering metric, the overclustering metric measuring splits of a first cluster into two or more of the second clusters, including:
- determining a third number of first records that split into multiple clusters generated from the first records by the second clustering process;
- determining a second difference value that measures a difference between the first number and the third number;
- and
- determining the overclustering metric based on the second difference value;
- and
- displaying results of the comparison for evaluating and modifying the first clustering process.
11 more elements
A system for evaluating a first clustering process, comprising:
- one or more computers configured to perform operations comprising:
- receiving first clusters generated by a first clustering process, each cluster including one or more records;
- applying a second clustering process to the received clusters to generate second clusters, where the second clustering process differs from the first clustering process;
- comparing the first clusters generated by the first clustering process to the second clusters generated by the second clustering process, the comparing including:
- determining a duplication metric, the duplication metric measuring mergers of records that belong to separate first clusters into a second cluster, including:
- calculating a first difference value that measures a difference between a first number of the first clusters and a second number of the second clusters;
- and
- determining the duplication metric based on the first difference value;
- determining an overclustering metric, the overclustering metric measuring splits of a first cluster into two or more of the second clusters, including:
- determining a third number of first records that split into multiple clusters generated from the first records by the second clustering process;
- determining a second difference value that measures a difference between the first number and the third number;
- and
- determining the overclustering metric based on the second difference value;
- and
- displaying results of the comparison for evaluating and modifying the first clustering process.
12 more elements
A system, comprising:
- one or more computers configured to perform operations comprising:
- receiving first clusters generated by a first clustering process, each first cluster including one or more records;
- applying a second clustering process to generate second clusters, where the second clustering process differs from the first clustering process;
- comparing the first clusters generated by the first clustering process to the second clusters generated by the second clustering process, the comparing including:
- determining a duplication metric, the duplication metric measuring mergers of records that belong to separate first clusters into a second cluster;
- and
- determining an overclustering metric, the overclustering metric measuring splits of a first cluster into two or more of the second clusters;
- and
- displaying results of the comparison for evaluating and modifying the first clustering process.
6 more elements
A computer readable medium storing a computer program, the computer program including instructions that, when executed, cause at least one processor to:
- receive first clusters generated by a first clustering process, each first cluster including one or more records;
- apply a second clustering process to the received clusters to generate second clusters, where the second clustering process differs from the first clustering process;
- compare the first clusters generated by the first clustering process to the second clusters generated by the second clustering process, the comparing including
- determining a duplication metric, the duplication metric measuring mergers of records that belong to separate first clusters into a second cluster;
- and
- determining an overclustering metric, the overclustering metric measuring splits of a first cluster into two or more of the second clusters;
- and
- display results of the comparison for evaluating the first clustering process.
5 more elements
Dependencies are inferred from the claim text.