Predictive cluster analytics optimization

Patent metadata

Field	Value
Publication number	US-9477745-B2
Application number	US-201414325233-A
Country	US
Kind code	B2
Filing date	Jul 7, 2014
Priority date	Jul 7, 2014
Publication date	Oct 25, 2016
Grant date	Oct 25, 2016

Abstract

Cluster analysis of data points in a data set can be optimized by identification of a preferred cluster analysis method. This identification can be based on indexing the data using a Hilbert curve and determining whether the data points are predominantly in spherical or non-spherical clusters. Methods, systems, and articles of manufacture are described.

First claim

What is claimed is:

1. A computer program product comprising a non-transitory machine-readable medium storing instructions that, when executed by at least one programmable processor, cause the at least one programmable processor to perform operations comprising:
indexing a data set using a Hilbert curve, the data set comprising a plurality of data points, the indexing comprising determining a distance metric for each data point along the Hilbert curve;
detecting whether the data points are predominantly in spherical or non-spherical clusters, the detecting comprising comparing an amount of the data points designated as being in dominant sets based on the indexing to a threshold value, such that when the amount exceeds the threshold, the data points are detected to be predominantly in spherical clusters and when the amount does not exceed the threshold, the data points are detected to be predominantly in non-spherical clusters;
identifying a centroid model as a preferred cluster analysis method when the data points are detected to be predominantly in spherical clusters and a density model as the preferred cluster analysis method when the data points are detected to be predominantly in non-spherical clusters; and
promoting the preferred cluster analysis method.

2. The computer program product of claim 1 , wherein the density model comprises a DBSCAN algorithm and the centroid model comprises a K-means algorithm.

3. The computer program product of claim 1 , wherein the operations further comprise designating subsets of the data points as dominant sets such that a first overall similarity among internal members of a subset is higher than a second overall similarity between external data points and the internal data points.

4. The computer program product of claim 3 , wherein the first and second overall similarities are based on the distance metrics resulting from the indexing.

5. The computer program product of claim 1 , wherein the promoting comprises at least one of presenting the preferred cluster analysis method as a suggestion to a user and automatically performing a cluster analysis on the data set using the preferred cluster analysis method.

6. A system comprising:
computer hardware configured to perform operations comprising:
indexing a data set using a Hilbert curve, the data set comprising a plurality of data points, the indexing comprising determining a distance metric for each data point along the Hilbert curve;
detecting whether the data points are predominantly in spherical or non-spherical clusters, the detecting comprising comparing an amount of the data points designated as being in dominant sets based on the indexing to a threshold value, such that when the amount exceeds the threshold, the data points are detected to be predominantly in spherical clusters and when the amount does not exceed the threshold, the data points are detected to be predominantly in non-spherical clusters;
identifying a centroid model as a preferred cluster analysis method when the data points are detected to be predominantly in spherical clusters and a density model as the preferred cluster analysis method when the data points are detected to be predominantly in non-spherical clusters; and
promoting the preferred cluster analysis method.

7. The system of claim 6 , wherein the density model comprises a DBSCAN algorithm and the centroid model comprises a K-means algorithm.

8. The system of claim 6 , wherein the operations further comprise designating subsets of the data points as dominant sets such that a first overall similarity among internal members of a subset is higher than a second overall similarity between external data points and the internal data points.

9. The system of claim 8 , wherein the first and second overall similarities are based on the distance metrics resulting from the indexing.

10. The system of claim 6 , wherein the promoting comprises at least one of presenting the preferred cluster analysis method as a suggestion to a user and automatically performing a cluster analysis on the data set using the preferred cluster analysis method.

11. The system of claim 6 , wherein the computer hardware comprises:
a programmable processor; and
a machine-readable medium storing instructions that, when executed by the at least one processor, cause the at least one programmable processor to perform at least some of the operations.

12. A computer-implemented method comprising:
indexing a data set using a Hilbert curve, the data set comprising a plurality of data points, the indexing comprising determining a distance metric for each data point along the Hilbert curve;
detecting whether the data points are predominantly in spherical or non-spherical clusters, the detecting comprising comparing an amount of the data points designated as being in dominant sets based on the indexing to a threshold value, such that when the amount exceeds the threshold, the data points are detected to be predominantly in spherical clusters and when the amount does not exceed the threshold, the data points are detected to be predominantly in non-spherical clusters;
identifying a centroid model as a preferred cluster analysis method when the data points are detected to be predominantly in spherical clusters and a density model as the preferred cluster analysis method when the data points are detected to be predominantly in non-spherical clusters; and
promoting the preferred cluster analysis method,
wherein the indexing, the detecting, the identifying, and the promoting is performed by at least one system comprising computer hardware.

13. The computer-implemented method of claim 12 , wherein the density model comprises a DBSCAN algorithm and the centroid model comprises a K-means algorithm.

14. The computer-implemented method of claim 12 , further comprising designating subsets of the data points as dominant sets such that a first overall similarity among internal members of a subset is higher than a second overall similarity between external data points and the internal data points.

15. The computer-implemented method of claim 14 , wherein the first and second overall similarities are based on the distance metrics resulting from the indexing.

16. The computer-implemented method of claim 12 , wherein the promoting comprises at least one of presenting the preferred cluster analysis method as a suggestion to a user and automatically performing a cluster analysis on the data set using the preferred cluster analysis method.

Assignees

• Tyercha Edward-Robert
• Kazmaier Gerrit Simon
• Gildhoff Hinnerk
• Pekel Isil
• Volker Lars
• Grouisborn Tim
• Sap Se

Inventors

• Tyercha Edward-Robert
• Kazmaier Gerrit Simon
• Gildhoff Hinnerk
• Pekel Isil
• Volker Lars
• Grouisborn Tim

Classifications

• G06F16/287 Primary

  Visualization; Browsing · CPC title

• G06F16/901

  Indexing; Data structures therefor; Storage structures (for retrieval from the web G06F16/951) · CPC title

• G06F16/22

  Indexing; Data structures therefor; Storage structures · CPC title

• G06F16/2465

  Query processing support for facilitating data mining operations in structured databases · CPC title

• G06F16/2228

  Indexing structures · CPC title