IDENTIFYING PATTERNS OF SIGNIFICANCE IN NUMERIC ARRAYS OF DATA

• US 20100287153A1
• Filed: 05/06/2009
• Published: 11/11/2010
• Est. Priority Date: 05/06/2009
• Status: Active Grant

• Current Assignee: Business Objects Incorporated (SAP SE)
• Original Assignee: Business Objects Incorporated (SAP SE)
• Inventors: MACGREGOR, JOHN
• Granted Patent: US 8,166,064 B2
• Time in Patent Office: Days
• US Class Current: 707/707
• CPC Class Codes: G06F 16/283 Multi-dimensional databases or data warehouses, e.g. MOLAP or ROLAP
• CPC Class Codes: G06F 18/2433 Single-class perspective, e.g. one-against-all classification; Novelty detection; Outlier detection
• CPC Class Codes: G06F 2218/12 Classification; Matching

Abstract

Disclosed is a computer method and system for identifying significance of patterns across a plurality of data patterns, which involves identifying pattern types of the plurality of data patterns, determining a relative pattern significance factor to compare the pattern types. Determining the relative pattern significance factor further involves calculating a percentage change of an identified outlier from a median for a outlier pattern, calculating a value of a step change as a percentage of a last value of a step preceding the step change for a step change pattern and calculating a percentage change from a start value on the fitted curve to an end value on the fitted curve for a trend pattern. A ranked list of the pattern types are returned based on their corresponding relative pattern significant factors.

Claims

1. An article of manufacture, comprising:

a machine readable storage medium comprising instructions which when executed by a machine cause the machine to perform operations comprising;

receiving a data set;

based on a plurality of data patterns, determining relative pattern significance factors for the plurality of data patterns, the determining comprising;

if the pattern type is an outlier pattern, calculating a percentage change of an identified outlier from a median;

if the pattern type is a step change pattern, calculating a percentage of a step change between a last value of a step and a first value preceding the step, wherein the step corresponds to an outlier in a first difference on the data set; and

if the pattern type is a trend pattern, calculating a degree of freedom factor;

accepting a fitted curve approximation based on the degree of freedom factor;

calculating a percentage change from a start value on the fitted curve to a end value on the fitted curve; and

returning a ranked list of the pattern types based on their corresponding relative pattern significance factors.

2. The article of manufacture in claim 1, wherein determining the trend pattern comprises determining a linear regression for the data set for a linear fitted curve.

3. The article of manufacture in claim 1, wherein determining the trend pattern comprises determining a non-linear regression for the data set for a non-linear fitted curve.

4. The article of manufacture in claim 3, wherein determining the non-linear regression for the data set comprises transforming the non-linear regression to a linear regression.

5. The article of manufacture in claim 3, wherein the non-linear regression pattern is selected from a geometric regression pattern, a natural logarithm regression pattern or an exponential regression pattern.

6. The article of manufacture in claim 1, wherein calculating the percentage of the step change further comprises a difference between a first average of values after a step and a second average for values after the step.

7. The article of manufacture in claim 1, wherein the data set is included in a multi-dimensional data source.

8. The method of claim 1, wherein calculating the percentage change from the start value on the fitted curve to the end value on the fitted curve comprises accepting the fitted curve if the value of freedom factor is more than 95% of the freedom factor.

9. A method for identifying significance of patterns across a plurality of data patterns, the method comprising:

receiving a data set from a multi-dimensional data source;

based on a plurality of data patterns, determining relative pattern significance factors for the plurality of data patterns;

the determining comprising;

if the pattern type is an outlier pattern, calculating a percentage change of an identified outlier from a median;

if the pattern type is an step change pattern, calculating a value of a step change as a percentage of a last value of a step preceding the step change; and

if the pattern type is a trend pattern, calculating a degree of freedom factor;

accepting a fitted curve approximation based on the degree of freedom factor;

calculating a percentage change from a start value on the fitted curve to a end value on the fitted curve; and

displaying a ranked list of the pattern types based on their corresponding relative pattern significance factors.

10. The method of claim 9, wherein determining the trend pattern comprises a linear regression for the data set.

11. The method of claim 9, wherein determining the trend pattern comprises a non-linear regression for the data set.

12. The method of claim 11, wherein determining the non-linear regression for the data set comprises transforming the non-linear regression to a linear regression.

13. The method of claim 11, wherein the non-linear regression comprises a geometric regression.

14. The method of claim 11, wherein the non-linear regression comprises a natural logarithm regression.

15. The method of claim 11, wherein the non-linear regression comprises an exponential regression.

16. The method of claim 9, wherein displaying a ranked list based on their corresponding relative pattern significance factors comprises displaying the ranked list in a table format.

17. A computer system including a processor and a memory, the memory comprising instructions that are executable by the processor, the instructions comprising:

a pattern type identifier module for identifying a pattern type;

a relative pattern significance calculator in communication with the pattern type identifier for determining the relative pattern significance factor for pattern type; and

a user interface in communication with the relative significance determiner for displaying a list of the relative pattern significance factors.

18. The computer system of claim 17, wherein determining the relative pattern significance factor for an outlier comprises calculating a percentage deviation of distance of an identified outlier from a median.

19. The computer system of claim 17, wherein determining the relative pattern significance factor for a step change pattern comprises calculating a value of a step change as a percentage deviation of a last value of a step preceding the step change.

20. The computer system of claim 17, wherein determining the relative pattern significance factor for a step change pattern comprises:

determining a degree of freedom factor; and

calculating a percentage change from a first value on a fitted curve to a last value on the fitted curve.