CN-121998777-A - Recommendation method and device for insurance products, electronic equipment and storage medium

CN121998777ACN 121998777 ACN121998777 ACN 121998777ACN-121998777-A

Abstract

The application provides a recommendation method, a recommendation device, electronic equipment and a storage medium of an insurance product, wherein the method comprises the steps of extracting explicit feature vectors of users to be recommended according to user information of the users to be recommended, extracting implicit feature vectors of the users to be recommended based on social relations of the users to be recommended in a community network, splicing the explicit feature vectors and the implicit feature vectors of the users to be recommended to generate fusion feature vectors of the users to be recommended, determining a plurality of target reference users based on similarity between the fusion feature vectors of the users to be recommended and fusion feature vectors of reference users, and determining dangerous recommendation lists of the users to be recommended based on dangerous purchase lists of the target reference users.

Inventors

WEN JIAMEI
YANG YINGHUI
WANG TENG
LIU YUE
CHANG YUHANG

Assignees

人保信息科技有限公司
中国人民人寿保险股份有限公司
中国人民保险集团股份有限公司

Dates

Publication Date: 20260508
Application Date: 20260122

Claims (10)

1. A method of recommending insurance products, the method comprising: according to user information of the user to be recommended, extracting an explicit feature vector of the user to be recommended; Extracting implicit feature vectors of users to be recommended based on social relations of the users to be recommended in a community network; splicing the explicit feature vector and the implicit feature vector of the user to be recommended to generate a fusion feature vector of the user to be recommended; determining a plurality of target reference users based on the similarity between the fusion feature vectors of the users to be recommended and the fusion feature vectors of the reference users; and determining a dangerous seed recommendation list of the user to be recommended based on the dangerous seed purchase list of the target reference user.
2. The method of claim 1, wherein the implicit feature vector is extracted by: Constructing a community network diagram, wherein the community network diagram comprises a second-order path formed by a plurality of first user nodes, community nodes and second user nodes; Generating a plurality of node path sequences with preset lengths by random walk based on a community network diagram; Inputting the node path sequence into a Skip-Gram model, and enabling the Skip-Gram model to output a corresponding implicit feature vector so as to train and generate an implicit embedded model.
3. The method of claim 1, wherein the implicit feature vector is extracted by: Constructing a community network diagram, wherein the community network diagram comprises a second-order path formed by a plurality of first user nodes, community nodes and second user nodes; Generating a plurality of node path sequences with preset lengths by random walk based on a community network diagram; generating positive and negative samples based on the node path sequence; and inputting the positive and negative samples into a Skip-Gram model, and enabling the Skip-Gram model to output corresponding implicit feature vectors so as to train and generate an implicit embedded model.
4. The method of claim 1, wherein the user information includes at least basic personal information, policy behavior information, browsing behavior information, and medical claim information of the user, and wherein the explicit feature vector is extracted by: And encoding basic personal information, policy behavior information, browsing behavior information and medical claim information of the user and carrying out feature fusion to generate an explicit feature vector with specified dimension.
5. The method of claim 4, further comprising, prior to the step of encoding the user's underlying personal information, policy behavior information, browsing behavior information, and medical claims information: Filling or removing missing values of basic personal information, policy behavior information, browsing behavior information and medical claim information of a user; And standardizing the basic personal information, the policy behavior information, the browsing behavior information and the medical claim information of the user.
6. The method of claim 1, wherein the reference users are ranked based on the similarity, and the first K reference users are targeted reference users.
7. The method of claim 1, wherein the user to be recommended is a user who browses an insurance product purchased by a reference user for a preset period of time but does not purchase the insurance product.
8. A recommendation device for insurance products, the device comprising: The first vector extraction module is used for extracting the explicit feature vector of the user to be recommended according to the user information of the user to be recommended; The second vector extraction module is used for extracting an implicit characteristic vector of the user to be recommended based on the social relation of the user to be recommended in the community network; the fusion module is used for splicing the explicit feature vector and the implicit feature vector of the user to be recommended so as to generate a fusion feature vector of the user to be recommended; The searching module is used for determining a plurality of target reference users based on the similarity between the fusion feature vectors of the users to be recommended and the fusion feature vectors of the reference users; And the recommending module is used for determining the dangerous seed recommending list of the user to be recommended based on the dangerous seed purchasing list of the target reference user.
9. An electronic device comprising a processor, a memory and a bus, the memory storing machine readable instructions executable by the processor, the processor and the memory in communication via the bus when the electronic device is in operation, the processor executing the machine readable instructions to perform the steps of the proposed method of insurance products according to any one of claims 1 to 7.
10. A computer readable storage medium, characterized in that the computer readable storage medium has stored thereon a computer program which, when executed by a processor, performs the steps of the recommendation method of an insurance product according to any of claims 1 to 7.

Description

Recommendation method and device for insurance products, electronic equipment and storage medium Technical Field The application relates to the technical field of artificial intelligence, in particular to a recommendation method and device of insurance products, electronic equipment and a storage medium. Background With the continuous development of insurance markets, insurance products are increasingly abundant in variety, and users face a plurality of difficulties in selecting the insurance products. The current insurance recommendation system refers to the Internet intelligent recommendation field method, and performs user personalized recommendation depending on basic information, historical browsing and purchasing behavior tracks of users. However, for insurance companies that are still mainly in an offline marketing mode, the behavior data of the user is relatively sparse, and even lacks records of online behavior data, and especially for new users, the related data is basically empty, so that the recommendation accuracy is low. Disclosure of Invention The embodiment of the application aims to provide a recommendation method and device for insurance products, electronic equipment and storage media, which are used for solving the technical problem that the existing insurance product recommendation is inaccurate. The invention provides a recommendation method of an insurance product, which comprises the steps of extracting explicit feature vectors of users to be recommended according to user information of the users to be recommended, extracting implicit feature vectors of the users to be recommended based on social relations of the users to be recommended in a community network, splicing the explicit feature vectors and the implicit feature vectors of the users to be recommended to generate fusion feature vectors of the users to be recommended, determining a plurality of target reference users based on similarity between the fusion feature vectors of the users to be recommended and fusion feature vectors of reference users, and determining dangerous seed recommendation lists of the users to be recommended based on dangerous seed purchase lists of the target reference users. In an alternative embodiment, the implicit feature vector is extracted by: Constructing a community network diagram, wherein the community network diagram comprises a second-order path formed by a plurality of first user nodes, community nodes and second user nodes; Generating a plurality of node path sequences with preset lengths by random walk based on a community network diagram; Inputting the node path sequence into a Skip-Gram model, and enabling the Skip-Gram model to output a corresponding implicit feature vector so as to train and generate an implicit embedded model. In an alternative embodiment, the implicit feature vector is extracted by: Constructing a community network diagram, wherein the community network diagram comprises a second-order path formed by a plurality of first user nodes, community nodes and second user nodes; Generating a plurality of node path sequences with preset lengths by random walk based on a community network diagram; generating positive and negative samples based on the node path sequence; and inputting the positive and negative samples into a Skip-Gram model, and enabling the Skip-Gram model to output corresponding implicit feature vectors so as to train and generate an implicit embedded model. In an alternative embodiment, the user information includes at least basic personal information, policy behavior information, browsing behavior information, and medical claim information of the user, and the explicit feature vector is extracted by: And encoding basic personal information, policy behavior information, browsing behavior information and medical claim information of the user and carrying out feature fusion to generate an explicit feature vector with specified dimension. In an alternative embodiment, before the step of encoding the user's basic personal information, policy action information, browsing action information and medical claim information, the method further comprises: Filling or removing missing values of basic personal information, policy behavior information, browsing behavior information and medical claim information of a user; And standardizing the basic personal information, the policy behavior information, the browsing behavior information and the medical claim information of the user. In an alternative embodiment, the reference users are ranked based on similarity, with the top K reference users being the target reference users. In an alternative embodiment, the user to be recommended is a user who browses the insurance product purchased by the reference user for a preset period of time, but does not purchase the insurance product. In a second aspect, the present invention provides an insurance product recommendation device, the device comprising: The first vector extraction