Methods for estimating the seasonality of groups of similar items of commerce data sets based on historical sales data values and associated error information

• US 6,834,266 B2
• Filed: 10/11/2001
• Issued: 12/21/2004
• Est. Priority Date: 10/11/2001
• Status: Active Grant

• Current Assignee: Oracle International Corporation (Oracle Corporation)
• Original Assignee: ProfitLogic, Inc. (Oracle Corporation)
• Inventors: Kumar, Mahesh, Woo, Jonathan W., Gaidarev, Peter
• Primary Examiner(s): Hapiz, Tariq R.
• Assistant Examiner(s): BOSWELL, BETH V
• Application Number:
  US09/975,769
• Publication Number:
  US 20030074251A1
• Time in Patent Office: 1,167 Days
• Field of Search: 705/10, 705/1, 705/7, 705/16, 705/21, 702/182, 702/183, 702/187, 702/197, 702/179-181, 707/10, 707/1
• US Class Current: 705/7.29
• CPC Class Codes: G06Q 10/087 Inventory or stock management, e.g. order filling, procurement or balancing against orders
• CPC Class Codes: G06Q 30/0201 Market modelling; Market analysis; Collecting market data
• CPC Class Codes: G06Q 30/0204 Market segmentation

Abstract

A set of data is received containing values associated with respective data points, the values associated with each of the data points being characterized by a distribution. The values for each of the data points are expressed in a form that includes information about a distribution of the values for each of the data points. The distribution information is used in clustering the set of data with at least one other set of data containing values associated with data points.

Claims

1. A computer-based method of estimating a seasonality of clusters of similar items of commerce based on historical sales data values and associated error information comprising:

executing on a computer the steps (A)-(E) as follows;

A. receiving a first set of data containing a plurality of data points, each of which data points represent historical sales data for one or more of the items of commerce and which is expressed as a value and an associated error,

B. determining a distance between the first set of data and each of one or more other sets of data, where each of the other data sets contains a plurality of data points, each of which data points represents historical sales data for one or more of the items of commerce and which is expressed as a value and an associated error,

C. the determining step including measuring each distance based on the relation;

$d_{i<br><br>\text{\hspace{1em}<br><br>}<br><br>j}=\mathrm{ChiSqr\_PDF}<br><br>\left(\underset{l=1}{\overset{k}{\sum <br><br>}}<br><br>{\left(\frac{{\mathrm{\mu <br><br>}}_{i<br><br>\text{\hspace{1em}<br><br>}<br><br>l}-{\mathrm{\mu <br><br>}}_{j<br><br>\text{\hspace{1em}<br><br>}<br><br>l}}{s_l}\right)}^2,k-1\right)$

where

 

d_ij is a measure of the distance between set of data i and of data j

 

ChiSqr_PDF is a Chi-square distribution function

 

k is a number of data points in each of sets of data i and j

 

l is an index

 

μ

_il is a l^th value of set of data i

 

μ

jl is a l^th value of set of data j

 

s={square root over (s_il²+s_jl²)}, where s_il is error associated with the l^th value of set of data i and s_ij is the error associated with the l^th value of set of data j.

D. clustering the first set of data with at least one of the other sets of data, the clustering step including comparing with a threshold one or more distances determined in the determining step,

E. generating a composite data set that estimates the seasonality of one or more clusters of the items of commerce, where the composite data set is based on clustering performed in step (D) and is generated as a function of data points contained in the first set of data and the one or more other sets of data, if any, whose distances compared favorably with the threshold.

2. The method of claim 1 in which at least one value and the at least one value'"'"'s associated error comprise a Gaussian distribution.

3. The method of claim 1 in which the clustering step includes merging using a weighted average data sets whose distances compared favorably with the threshold.

4. The method of claim 1, wherein the clustering step includes merging seasonalities contained in data sets belonging to a cluster.

5. A computer-based method of estimating a seasonality of clusters of similar items of commerce based on historical sales data values and associated error information comprising

executing on a computer steps (A)-(C) as follows

A. determining distances between respective pairs of data sets in a plurality of data sets, where each data set contains a plurality of data points forming a time-series of retail data, where each data point is expressed as a value and an associated error,

B. the determining step including measuring each distance as a function of the relation $d_{i}$

 



j=ChiSqr_PDF

(∑

l=1k

(μ

i

 



l-μ

j

 



lsl)2,k-1)

where

 

d_ij is a measure of the distance between set of data i and set of data j

 

Chisqr_PDF is a Chi-square distribution function

 

k is a number of data points in/each of sets of data i and j

 

l is an index

 

μ

_u is a l^th value of set of data i

 

μ

_jl is a l^th value of set of data j

 

s=√

{square root over (s_il²+s_jl²)}, where s_il is error associated with the l^th value of set of data i and s_jl is the error associated with the l^th value of set of data j, and

C. estimating the seasonality of one or more clusters of the items of commerce based on clustering the plurality of data sets as a function of the distances determined in the determining step, the clustering step including forming one or more composite data sets, each containing a plurality of data points forming a time-series of retail data based on data points contained in data sets whose distances compared favorably with a threshold.

6. The method of claim 5, wherein the retail data is sales demand data.

7. The method of claim 6, wherein the retail data is seasonality data for items of commerce.

8. The method of claim 5, wherein the sets of data contain a plurality of data points forming a time-series of sales demand data, the method further comprising the step of processing one or more of the sets of data to filter factors other than seasonality, where seasonality is defined as the underlying demand for a group of merchandise items as a function of time of the year.

9. The method of claim 8, wherein the processing step includes filtering the one or more set of data to filter product life cycle effects.

10. The method of claim 9, wherein the processing filtering step includes at least one of averaging and scaling data points in one or more of the sets.