Method and system for calculating value of website visitor

Patent metadata

Field	Value
Publication number	US-9436907-B2
Application number	US-65876710-A
Country	US
Kind code	B2
Filing date	Feb 12, 2010
Priority date	Feb 23, 2009
Publication date	Sep 6, 2016
Grant date	Sep 6, 2016

Abstract

Calculating a value of a website visitor includes initializing a calculation model for calculating the value of the website visitor, the calculation model being a neural network model with visitor information as an input and the visitor's value as an output; training the calculation model by using a data sample and determining the calculation model; and obtaining the visitor information, and calculating the value of the visitor by using the determined calculation model.

First claim

What is claimed is:

1. A method for calculating a value of a website visitor, comprising:
initializing a calculation model for calculating the value of the website visitor, the calculation model being a neural network model with visitor information as an input and the value of the website visitor as an output;
training the calculation model by using a data sample;
determining the calculation model;
obtaining the visitor information, wherein the visitor information includes a plurality of categories; wherein the visitor information is automatically detected from the website visitor or obtained from registration information of the website visitor, and wherein the plurality of categories includes at least two of the following: gender of the visitor, age of the visitor, region of the visitor, number of visits of a website by the visitor, page visited by the visitor, stay time of a visit of the website, whether the visitor performs chatting, whether the visitor performs sending an email, number of times the visitor performs chatting, or number of times the visitor performs sending the email;
converting the obtained information into corresponding numerical values to be input into the calculation model;
calculating the value of the visitor by using the determined calculation model; wherein:
the calculation model includes an input layer, a hidden layer, and an output layer;
the hidden layer is associated with a hidden layer transfer function;
the output layer is associated with an output layer transfer function; and
the hidden layer transfer function is different from the output layer transfer function;

correcting the calculation model that is currently determined, the correcting of the calculation model including:
obtaining a correlation between the visitor's value and a category of the visitor information from the calculation model currently determined; and
deleting the category of the visitor information from the input vector of the calculation model currently determined in the event that the correlation of the category is lower than a preset threshold, wherein the deleting of the category of the visitor information includes eliminating the category of the visitor information from the input vector of the calculation model;

dynamically updating the calculation model with the corrected calculation model, comprising:
calculating the value of the visitor using the corrected calculation model;

determining whether the value of the visitor meets or exceeds a threshold;
in the event that the value of the visitor meets or exceeds the threshold, performing a first service activity; and
in the event that the value of the visitor does not meet and does not exceed the threshold, performing a second service activity.

2. The method according to claim 1 , wherein the initialized calculation model is Y=ƒ 2 (W 2 ƒ 1 (W 1 X+B 1 )+B 2 ); and wherein
X is an input vector; Y is an output vector, W 1 is a hidden layer weight matrix; B 1 is a hidden layer bias vector, ƒ 1 is the hidden layer transfer function, W 2 is an output layer weight matrix, B 2 is an output layer bias vector, and ƒ 2 is the output layer transfer function.

3. The method according to claim 2 , wherein ƒ 1 is a non-linear action function and ƒ 2 is a linear function.

4. The method according to claim 1 , wherein the visitor information as the input to the calculation model is numerical information and the visitor's value as the output from the calculation model is a numerical value.

5. The method according to claim 1 , wherein the training the calculation model comprises training the calculation model by using Back Propagation.

6. The method according to claim 1 , wherein determining the calculation model comprises determining the calculation model when an error between a sample output value of the calculation model and a desired output value meets an accuracy requirement.

7. The method according to claim 1 , wherein correcting the calculation model currently determined comprises:
comparing an actual output value of the calculation model currently determined with a desired output value to obtain an error between the actual output value and the desired output value; and
re-training the calculation model when the error is higher than a preset threshold.

8. The method of claim 1 , wherein the value of the website visitor, at least in part, determines a service activity to be performed for the website visitor.

9. A system for calculating a value of a website visitor, comprising:
one or more processors configured to:
initialize a calculation model for calculating the value of the website visitor, the calculation model being a neural network model with visitor information as an input and the visitor's value as an output;
train the calculation model by using a data sample;
determine the calculation model;
obtain the visitor information, wherein the visitor information includes a plurality of categories; wherein the visitor information is automatically detected from the website visitor or obtained from registration information of the website visitor, and wherein the plurality of categories includes at least two of the following: gender of the visitor, age of the visitor, region of the visitor, number of visits of a website by the visitor, page visited by the visitor, stay time of a visit of the website, whether the visitor performs chatting, whether the visitor performs sending an email, number of times the visitor performs chatting, or number of times the visitor performs sending the email;
convert the obtained information into corresponding numerical values to be input into the calculation model;
calculate the value of the visitor by using the determined calculation model; wherein:
the calculation model includes an input layer, a hidden layer, and an output layer;
the hidden layer is associated with a hidden layer transfer function;
the output layer is associated with an output layer transfer function; and
the hidden layer transfer function is different from the output layer transfer function;

correct the calculation model that is currently determined, the correcting of the calculation model including:
obtain a correlation between the visitor's value and a category of the visitor information from the calculation model currently determined; and
delete the category of the visitor information from the input vector of the calculation model currently determined in the event that the correlation of the category is lower than a preset threshold, wherein the deleting of the category of the visitor information includes eliminate the category of the visitor information from the input vector of the calculation model;

dynamically update the calculation model with the corrected calculation model, wherein model, comprising to:
calculate the value of the visitor using the corrected calculation model;

determine whether the value of the visitor meets or exceeds a threshold;
in the event that the value of the visitor meets or exceeds the threshold, perform a first service activity; and
in the event that the value of the visitor does not meet and does not exceed the threshold, perform a second service activity; and

one or more memories coupled to the one or more processors and configured to provide the processor with instructions.

10. The system according to claim 9 , wherein the initialized calculation model is:
Y=ƒ 2 ( W 2 ƒ 1 ( W 1 X+B 1 )+ B 2 ); and wherein

X is an input vector; Y is an output vector, W 1 is a hidden layer weight matrix; B 1 is a hidden layer bias vector, ƒ 1 is the hidden layer transfer function, W 2 is an output layer weight matrix, B 2 is an output layer bias vector, and ƒ 2 is the ou

Assignees

• Lv Kaili
• Zhang Zheng
• Su Jie
• Alibaba Group Holding Ltd

Inventors

• Lv Kaili
• Zhang Zheng
• Su Jie

Classifications

• G06N3/08

  Learning methods · CPC title

• G06N3/082

  modifying the architecture, e.g. adding, deleting or silencing nodes or connections · CPC title

• G06N3/0499

  Feedforward networks · CPC title

• G06N3/09

  Supervised learning · CPC title

• G06Q30/02

  Marketing; Price estimation or determination; Fundraising · CPC title