Method for customer segmentation with applications to electronic commerce

• US 6,862,574 B1
• Filed: 07/27/2000
• Issued: 03/01/2005
• Est. Priority Date: 07/27/2000
• Status: Expired due to Term

• Current Assignee: Teradata US, Inc. (Teradata Corporation)
• Original Assignee: NCR Corporation (NCR Voyix Corporation)
• Inventors: Srikant, Sreedhar, Boerger, Ellen M., Cunningham, Scott W.
• Primary Examiner(s): Hafiz, Tariq R.
• Assistant Examiner(s): Heck, Michael C.
• Application Number:
  US09/626,887
• Time in Patent Office: 1,678 Days
• Field of Search: 705/10, 706/27
• US Class Current: 705/7.33
• CPC Class Codes: G06Q 30/02 Marketing; Price estimation or determination; Fundraising
• CPC Class Codes: G06Q 30/0204 Market segmentation

Abstract

The customer segmentation software according to the present invention automatically finds or creates profiles of prototypical customers in a large e-commerce database. The software matches all existing customer data in the database to one or more of the prototypical customers. The resulting customer segmentation is an effective summarization of the database and is useful for a range of business applications. Applications of the customer segmentation system include the development of customized web sites, the creation of targeted promotional offers and the prediction of consumer behavior.

Claims

1. A method of describing customer data stored in a database and predicting missing customer data for segmentation of the customer data, comprising:

a) determining customer profiles by a computer;

b) applying a Gaussian mixture model algorithm to the customer data stored an the database to assess a similarity of the customer profiles to existing customer profiles by the computer; and

c) predicting the missing customer data by the computer using the Gaussian mixture model algorithm, including the steps of

c-1) randomly initializing Gaussian mixture model parameters by the computer;

c-2) applying an expectation formula, maximization formulae, and radial basis function formulae, to the customer data to obtain a solution for the messing data by the computer; and

c-3) repeating step (c-2) until the solution for the missing customer data converges on a preferred solution by the computer.

2. The method of claim 1, wherein the expectation formula is:

$H<br><br>\left(i,j\right)=\underset{k}{\sum <br><br>}<br><br>\underset{j}{\sum <br><br>}<br><br>P_i<br><br>\left(k\right)<br><br>f_1<br><br>\left(C,X\right);<br><br>$

where X(j) is define, f₁(C,X) equals exp(−

(C_k−

X_j)²/R_j

else f1(C,X) equals 0,

the maximization formulae are;

$C<br><br>\left(k\right)=\frac{\underset{k}{\sum <br><br>}<br><br>\underset{j}{\sum <br><br>}<br><br>\underset{i}{\sum <br><br>}<br><br>H<br><br>\left(i,j\right)<br><br>X<br><br>\left(i,j\right)}{\underset{j}{\sum <br><br>}<br><br>\underset{i}{\sum <br><br>}<br><br>H<br><br>\left(i,j\right)}$$R<br><br>\left(k\right)=\frac{\underset{k}{\sum <br><br>}<br><br>\underset{j}{\sum <br><br>}<br><br>\underset{i}{\sum <br><br>}<br><br>H<br><br>\left(i,j\right)<br><br>{\left(X<br><br>\left(i,j\right)-C<br><br>\left(j,k\right)\right)}^2}{\underset{j}{\sum <br><br>}<br><br>\underset{i}{\sum <br><br>}<br><br>H<br><br>\left(i,j\right)}$$P_i<br><br>\left(k\right)=\frac{\underset{j}{\sum <br><br>}<br><br>\underset{i}{\sum <br><br>}<br><br>H<br><br>\left(i,j\right)}{\underset{k}{\sum <br><br>}<br><br>\underset{j}{\sum <br><br>}<br><br>\underset{i}{\sum <br><br>}<br><br>H<br><br>\left(i,j\right)}$

the radial basis function formula is;

if X(i,j) is undefined $X<br><br>\left(i,j\right)=\underset{j}{\sum <br><br>}<br><br>\underset{k}{\sum <br><br>}<br><br>C<br><br>\left(k\right)<br><br>H<br><br>\left(i,j\right)<br><br>\text{\hspace{1em}<br><br>}.$

3. A method of describing data and predicting missing data stored in a database, comprising:

a) applying a Gaussian mixture model algorithm to the data stored in the database to assess the similarity of customer profiles to existing customer profiles by a computer; and

b) predicting the missing data using the Gaussian mixture model algorithm by a computer, including the steps of

b-1) applying an expectation formula, maximization formulae and radial basis function formulae to the data stored in the database to derive a solution for the missing data;

b-2) repeating step (b-1) until the model converges on a preferred solution.

4. A method of describing data and predicting missing data stored in a database, comprising:

applying a Gaussian mixture model algorithm to the data stored in the database to assess a similarity of customer profiles to existing customer profiles by a computer;

predicting the missing data using the Gaussian mixture model algorithm by the computer by applying an expectation formula, maximization formulae, and radial basis function formulae; and

outputting matrices generated by the Gaussian mixture model algorithm and a maximum likelihood fit of the Gaussian mixture model algorithm.

5. The method of claim 4, further comprising outputting a textual description of customer segments.