CN-122022889-A - Method, device, medium and equipment for predicting purchase intention of customer

CN122022889ACN 122022889 ACN122022889 ACN 122022889ACN-122022889-A

Abstract

The application relates to the technical field of behavior prediction, and particularly provides a method, a device, a medium and equipment for predicting purchasing intention of a customer, wherein the method can comprise the steps of acquiring characteristic information corresponding to data information of a product of interest of an electric marketing customer; the feature information comprises basic features corresponding to basic client information, client behavior features corresponding to client behavior information and text embedding vectors corresponding to client historical call information, the text embedding vectors are determined based on label content of flow nodes of dialogue texts in the client historical call information, the basic features, the client behavior features and the text embedding vectors are spliced to obtain feature sets, and a pre-trained willingness prediction model is utilized to predict the feature sets to obtain purchase willingness of electric marketing clients to interested products. The embodiment of the application can accurately predict the purchase intention of the electric marketing customer.

Inventors

FU ZHUORAN
ZENG WENJIA
SONG CHENGYE
LIU XIAOPING

Assignees

零犀(北京)科技有限公司

Dates

Publication Date: 20260512
Application Date: 20260129

Claims (12)

1. A method of customer purchase intent prediction, comprising: The method comprises the steps of acquiring characteristic information corresponding to data information of a product of interest of an electric marketing customer, wherein the data information comprises customer basic information, customer behavior information and customer history call information, the customer behavior information comprises purchase records, application use time or consumption data, the characteristic information comprises basic characteristics corresponding to the customer basic information, the customer behavior characteristics corresponding to the customer behavior information and text embedded vectors corresponding to the customer history call information, and the text embedded vectors are determined based on label content of flow nodes of dialogue texts in the customer history call information; Splicing the basic features, the client behavior features and the text embedded vectors to obtain feature sets; and predicting the feature set by using a pre-trained willingness prediction model to obtain the purchase willingness degree of the electric marketing customer to the interested product.
2. The method of claim 1, wherein the obtaining the characteristic information corresponding to the data information of the product of interest to the electric marketing customer comprises: Dividing the client behavior information according to the time length to obtain a first behavior feature and a second behavior feature, wherein the first behavior feature and the second behavior feature form the client behavior feature.
3. The method according to claim 1 or 2, wherein the obtaining the feature information corresponding to the data information of the product of interest to the electric sales client includes: extracting a plurality of key label contents corresponding to the flow node in the client history call information; Splicing the plurality of key label contents according to the time nodes to obtain spliced sentences; and inputting the spliced statement into a language embedding model, and outputting the text embedding vector.
4. The method of claim 3, wherein extracting the plurality of key tag contents corresponding to the flow node in the client history call information comprises: Identifying a plurality of key flow nodes and a plurality of client query nodes in the client history call information; And taking the text summaries corresponding to the key flow nodes and the client query nodes as the key tag contents.
5. The method of claim 1 or 2, wherein predicting the feature set using a pre-trained willingness prediction model to obtain the purchase willingness of the electric marketing customer for the product of interest comprises: acquiring willingness prediction results of the basic characteristics and the client behavior characteristics; And correcting the willingness prediction result by using the text embedding vector to obtain the purchase willingness degree.
6. An apparatus for customer purchase intent prediction, comprising: The system comprises a feature processing module, a feature information processing module and a text embedding vector, wherein the feature information is used for acquiring feature information corresponding to data information of a product of interest of an electric marketing customer, the data information comprises customer basic information, customer behavior information and customer history call information, the customer behavior information comprises purchase records, application use time length or consumption data, the feature information comprises basic features corresponding to the customer basic information, the customer behavior features corresponding to the customer behavior information and the text embedding vector corresponding to the customer history call information, and the text embedding vector is determined based on label content of flow nodes of dialogue texts in the customer history call information; The feature splicing module is used for splicing the basic features, the client behavior features and the text embedded vectors to obtain feature sets; and the prediction module is used for predicting the feature set by utilizing a pre-trained willingness prediction model and acquiring the purchase willingness degree of the electric pin client to the interested product.
7. The apparatus of claim 6, wherein the feature processing module is to: Dividing the client behavior information according to the time length to obtain a first behavior feature and a second behavior feature, wherein the first behavior feature and the second behavior feature form the client behavior feature.
8. The apparatus of claim 6 or 7, wherein the feature processing module is to: extracting a plurality of key label contents corresponding to the flow node in the client history call information; Splicing the plurality of key label contents according to the time nodes to obtain spliced sentences; and inputting the spliced statement into a language embedding model, and outputting the text embedding vector.
9. The apparatus of claim 8, wherein the feature processing module is to: Identifying a plurality of key flow nodes and a plurality of client query nodes in the client history call information; And taking the text summaries corresponding to the key flow nodes and the client query nodes as the key tag contents.
10. The apparatus of claim 6 or 7, wherein the prediction module is to: acquiring willingness prediction results of the basic characteristics and the client behavior characteristics; And correcting the willingness prediction result by using the text embedding vector to obtain the purchase willingness degree.
11. A computer readable storage medium, characterized in that the computer readable storage medium has stored thereon a computer program, wherein the computer program when run by a processor performs the method according to any of claims 1-5.
12. An electronic device comprising a memory, a processor, and a computer program stored on the memory and running on the processor, wherein the computer program when run by the processor performs the method of any one of claims 1-5.

Description

Method, device, medium and equipment for predicting purchase intention of customer Technical Field The application relates to the technical field of behavior prediction, in particular to a method, a device, a medium and equipment for predicting purchasing intention of a customer. Background With the popularity of communication technology and big data, telemarketing (simply called electronic marketing) has become a key channel for business marketing and customer relationship management. In order to promote sales efficiency, reduce operating costs and achieve accurate marketing in a strong market competition, it becomes important to effectively predict customer purchase will. Currently, when predicting a purchase intention of a customer, a field expert sets a series of judgment rules by collecting static attribute information (such as age, sex, region, occupation, etc.) and limited historical interaction data (such as past purchase records, product browsing duration, etc.) of the customer, and predicts the purchase intention of the customer based on the judgment rules. However, the multidimensional static and historical structured data relied on in the prior art has single data dimension, and cannot realize accurate prediction of the purchase intention of the customer. Therefore, how to provide a precise method for predicting the purchase intention of a customer is a technical problem to be solved. Disclosure of Invention It is an object of some embodiments of the present application to provide a method, apparatus, medium and device for customer purchase intent prediction, according to the technical scheme provided by the embodiment of the application, the accurate prediction of the purchase intention of the customer can be improved, and the accuracy and efficiency of the electric pin are improved. According to the method, feature information corresponding to data information of a product of interest of an electric sales customer is obtained, the data information comprises customer basic information, customer behavior information and customer historical call information, the customer behavior information comprises purchase records, application using time or consumption data, the feature information comprises basic features corresponding to the customer basic information, the customer behavior features corresponding to the customer behavior information and text embedded vectors corresponding to the customer historical call information, the text embedded vectors are determined based on label content of flow nodes of dialogue texts in the customer historical call information, feature sets are obtained by splicing the basic features, the customer behavior features and the text embedded vectors, the feature sets are predicted by means of a pre-trained willingness prediction model, and the purchase willingness of the electric sales customer for the product of interest is obtained. According to the method, the device and the system, the characteristic set is obtained by splicing the characteristic information after the characteristic information of the data information of the product of interest by the electric marketing client is obtained, and finally the characteristic set is input into a pre-trained willingness prediction model to predict the purchase willingness of the product of interest by the electric marketing client. The embodiment of the application can realize accurate prediction of purchase intention of the electric marketing clients, and further can improve the electric marketing accuracy and the client experience. In some embodiments, the obtaining feature information corresponding to the data information of the product of interest of the electric marketing customer includes dividing the customer behavior information according to a time length to obtain a first behavior feature and a second behavior feature, wherein the first behavior feature and the second behavior feature form the customer behavior feature. According to the method, the device and the system, the first behavior characteristic and the second behavior characteristic are obtained by processing the client behavior information, so that the characteristic analysis of the client behavior information is realized, and the data support is improved for subsequent prediction. In some embodiments, the obtaining feature information corresponding to the data information of the product of interest of the electric marketing customer includes extracting a plurality of key tag contents corresponding to the flow node in the customer history call information, splicing the plurality of key tag contents according to a time node to obtain a spliced sentence, inputting the spliced sentence into a language embedding model, and outputting the text embedding vector. According to some embodiments of the application, the plurality of key label contents in the client history call information are extracted, spliced to obtain spliced sentences, and a language embeddi