METHODS AND SYSTEMS FOR PARTITIONING OF DATASETS FOR RETAIL SALES AND DEMAND CHAIN MANAGEMENT ANALYSIS

• US 20090012979A1
• Filed: 07/02/2007
• Published: 01/08/2009
• Est. Priority Date: 07/02/2007
• Status: Active Grant

• Current Assignee: Teradata US, Inc. (Teradata Corporation)
• Original Assignee: Teradata US, Inc. (Teradata Corporation)
• Inventors: Chan, Andrew, Balendran, Prathayana, Bateni, Arash, Kim, Edward
• Granted Patent: US 7,783,648 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06Q 30/02 Marketing; Price estimation or determination; Fundraising

Abstract

A partitioning system that provides a fast, simple and flexible method for partitioning a dataset. The process, executed within a computer system, retrieves product and sales data from a data store. Data items are selected and sorted by a data attribute of interest to a user and a distribution curve is determined for the selected data and data attribute. The total length of the distribution curve is calculated, and then the curve is divided into k equal pieces, where k is the number of the partitions. The selected data is thereafter partitioned into k groups corresponding to the curve divisions.

Claims

1. A computer-implemented method for partitioning a dataset, the method comprising the steps of:

maintaining an electronic database of data items, said data items containing at least one attribute;

sorting said data items by a first attribute;

determining a distribution curve for said data;

calculating a total length of said distribution curve;

dividing said distribution curve into a plurality of equal curve segments; and

grouping said data items into a plurality of partitions, each partition comprising data items corresponding to one of said curve segments.

2. The computer-implemented method for partitioning a dataset in accordance with claim 1, wherein:

said step of dividing said distribution curve into a plurality of equal curve segments comprises the step of dividing said distribution curve into k equal pieces, where k equals a number of desired partitions; and

said step of grouping said data items into a plurality of partitions comprises grouping said data items into k partitions, each one of said k partitions corresponding to one of said k curve segments.

3. The computer-implemented method for partitioning a dataset in accordance with claim 2, wherein:

said distribution curve plots attribute values for said data items against a sorted sequence of said data items; and

the length of said distribution curve is determined through use of the equation;

Length=f(range,count)≈

√

{square root over ((range)²+(count)²)}{square root over ((range)²+(count)²)},

where count represents the number of data items, and range represents the range in attribute values for said data items.

4. The computer-implemented method for partitioning a dataset in accordance with claim 2, further comprising the step of:

normalizing the values of said attribute prior to determining said distribution curve utilizing the equation;

5. A system for partitioning a dataset, comprising:

an electronic database of data items, said data items containing at least one attribute;

means for sorting said data items by a first attribute;

means for determining a distribution curve for said data;

means for calculating a total length of said distribution curve;

means for dividing said distribution curve into a plurality of equal curve segments; and

means for grouping said data items into a plurality of partitions, each partition comprising data items corresponding to one of said curve segments.

6. The system in accordance with claim 5, wherein:

dividing said distribution curve into a plurality of equal curve segments comprises the dividing said distribution curve into k equal pieces, where k equals a number of desired partitions; and

grouping said data items into a plurality of partitions comprises grouping said data items into k partitions, each one of said k partitions corresponding to one of said k curve segments.

7. The system in accordance with claim 6, wherein:

said distribution curve plots attribute values for said data items against a sorted sequence of said data items; and

the length of said distribution curve is determined through use of the equation;

Length=f(range,count)∞

√

{square root over ((range)²+(count)²)}{square root over ((range)²+(count)²)},

where count represents the number of data items, and range represents the range in attribute values for said data items.

8. The system in accordance with claim 7, wherein:

the values of said attribute are normalized prior to determining said distribution curve utilizing the equation;