METHODS AND SYSTEMS FOR THE DETERMINATION OF THRESHOLDS VIA WEIGHTED QUANTILE ANALYSIS

• US 20100082638A1
• Filed: 09/30/2008
• Published: 04/01/2010
• Est. Priority Date: 09/30/2008
• Status: Active Grant

• Current Assignee: Vmware LLC (Broadcom, Inc.)
• Original Assignee: VMware, Inc. (Broadcom, Inc.)
• Inventors: Marvasti, Mazda A.
• Granted Patent: US 8,171,033 B2
• Time in Patent Office: Days
• US Class Current: 707/748
• CPC Class Codes: G06F 17/18 for evaluating statistical data , e.g. average values, frequency distributions, probability functions, regression analysis forecasting specially adapted for a specific administrative, business or logistic context G06Q10/04

Abstract

Methods and systems for determination of thresholds for time-series data. Data is transformed by reducing outliers, dividing the time series data into discrete time intervals, and taking parts of the data corresponding to the range that the thresholds will bound. If data cycles are known, they may be applied to the data and the resulting sets are weighted. Thresholds are then derived from the weighted means and variances of the sets of weighted data.

Claims

1. A method comprising:

dividing a set of time-series data into at least one time period;

creating at least one set of ranged time-series data by selecting data points from each time period corresponding to a percentage of values useful in defining a threshold;

applying cycles to the set of ranged time-series data;

applying weights to the data for each cycle of ranged time-series data;

determining at least one threshold from the weighted time-series data.

2. The method of claim 1, wherein an upper and a lower threshold is determined from the weighted time-series data;

wherein a percentage of the highest values from each time period and a percentage of the lowest values from each time period comprise an upper set of ranged time-series data and a lower set of ranged time-series data.

3. The method of claim 1, wherein more recent data to has a greater weight than less recent data.

4. The method of claim 1, further comprising reducing outliers prior to dividing the time-series data into at least one time period.

5. The method of claim 1, wherein the time periods are selected from the group consisting of days, weeks, months, or years.

6. The method of claim 1, wherein each threshold is a multiple of weighted standard deviations from the weighted mean.

7. A method comprising:

providing a system for determination of dynamic thresholds, wherein the system determines dynamic thresholds by;

dividing a set of time-series data into at least one time period;

creating at least one set of ranged time-series data by selecting data points from each time period corresponding to a percentage of values useful in defining a threshold;

applying cycles to the set of ranged time-series data;

applying weights to the data for each cycle of ranged time-series data;

determining at least one threshold from the weighted time-series data.

8. The method of claim 7, wherein an upper and a lower threshold is determined from the weighted time-series data;

wherein a percentage of the highest values from each time period and a percentage of the lowest values from each time period comprise an upper set of ranged time-series data and a lower set of ranged time-series data.

9. The method of claim 7, wherein more recent data to has a greater weight than less recent data.

10. The method of claim 7, further comprising reducing outliers prior to dividing the time-series data into at least one time period.

11. The method of claim 7, wherein the time periods are selected from the group consisting of days, weeks, months, or years.

12. The method of claim 7, wherein each threshold is a multiple of weighted standard deviations from the weighted mean.

13. A machine-readable medium having program instructions stored thereon executable by a processing unit for performing the steps of:

dividing a set of time-series data into at least one time period;

creating at least one set of ranged time-series data by selecting data points from each time period corresponding to a percentage of values useful in defining a threshold;

applying cycles to the set of ranged time-series data;

applying weights to the data for each cycle of ranged time-series data;

determining at least one threshold from the weighted time-series data.

14. The machine-readable medium of claim 13, wherein an upper and a lower threshold is determined from the weighted time-series data;

wherein a percentage of the highest values from each time period and a percentage of the lowest values from each time period comprise an upper set of ranged time-series data and a lower set of ranged time-series data.

15. The machine-readable medium of claim 13, wherein more recent data to has a greater weight than less recent data.

16. The machine-readable medium of claim 13, further comprising reducing outliers prior to dividing the time-series data into at least one time period.

17. The machine-readable medium of claim 13, wherein the time periods are selected from the group consisting of days, weeks, months, or years.

18. The machine-readable medium of claim 13, wherein each threshold is a multiple of weighted standard deviations from the weighted mean.