Intelligent pushing method and system based on user data

By constructing user behavior data and encrypting it, the accuracy and security issues in information push are solved, the efficiency and accuracy of information push are improved, and the security of user data is ensured.

CN119089032APending Publication Date: 2024-12-06YUNDONG (SHANGHAI) TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410627394.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-05-21
Publication Date
2024-12-06

AI Technical Summary

Technical Problem

In a network database with messy information, it is difficult for users to quickly find the information of interest, and due to the expansion of data volume, the accuracy and security of information push are affected, resulting in reduced efficiency.

Method used

By obtaining user data, building user behavior data, using TF-IDF algorithm to obtain usage frequency and entity weights, performing binary packet encryption processing, and calling the network database to obtain pre-push content, and sorting and transmission according to the correlation. Ensure the accuracy and security of information push.

Benefits of technology

It improves the efficiency and accuracy of information push, ensures the security of user data, and avoids data leakage.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119089032A_ABST
    Figure CN119089032A_ABST
Patent Text Reader

Abstract

The invention relates to an intelligent pushing method and system based on user data, and belongs to the technical field of data processing. User behavior data are constructed according to the network action information, user entity information is obtained according to the user behavior data, the use frequency is obtained through TF-IDF algorithm processing according to the user behavior set, weight assignment is conducted on the user behavior set according to the use frequency and time information, and an entity weight is obtained; obtaining a user behavior label according to the user behavior entity and the user behavior entity relationship, performing binary grouping according to the user behavior label and the entity weight, performing encryption processing on the binary grouping to obtain ciphertext data, and decrypting the ciphertext data to obtain pre-push content, and obtaining relevancy according to the push feature vector and the user push feature vector, sorting the push content according to the relevancy, and transmitting the push content to a display terminal according to the user IP address. The information pushing efficiency is improved, the information pushing accuracy is realized, and the safety of user data is guaranteed.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of data processing, and in particular relates to an intelligent push method and system based on user data. Background Art

[0002] As the Internet connects information resources around the world, it has formed a rich and all-encompassing ocean of information, allowing people to obtain information, conduct instant communications, and shop online anytime and anywhere. However, this has also led to a huge increase in the amount of information on the Internet, forcing users to spend a lot of time and energy to filter this information in order to find the information they are really interested in and need.

[0003] At present, in the network database with messy information, if users want to retrieve interesting information according to their own needs and conduct in-depth analysis, they will not only be affected by the flood of information in the database, but may also encounter problems such as inconsistency of target information, which affects the accuracy of information push. In addition, since user data contains sensitive information of users, as the amount of data expands, the security of user data is difficult to guarantee, and it also leads to reduced efficiency of information push. Summary of the invention

[0004] In order to solve the above problems existing in the prior art, the present invention provides an intelligent push method and system based on user data, which obtains user data, constructs user behavior data according to network action information, obtains user entity information through information extraction according to user behavior data, obtains usage frequency according to user behavior set through TF-IDF algorithm processing, assigns weights to user behavior set according to usage frequency and time information to obtain entity weight, obtains user behavior label according to user behavior entity and user behavior entity relationship, performs binary grouping according to user behavior label and entity weight, encrypts binary grouping to obtain ciphertext data, cloud server decrypts ciphertext data to obtain user behavior label and entity weight, obtains pre-pushed content by calling network database according to user behavior label, obtains relevance according to push feature vector and user push feature vector, sorts push content according to relevance, and transmits sorted push content to display terminal according to user IP address. The efficiency of information push is improved, the accuracy of information push is achieved, and the security of user data is guaranteed.

[0005] The purpose of the present invention can be achieved through the following technical solutions:

[0006] A smart push method based on user data comprises the following steps:

[0007] S1: Acquire user data, where the user data includes the user's IP address and network action information, where the network action information carries time information;

[0008] S2: construct user behavior data according to the network action information, obtain user entity information through information extraction according to the user behavior data, the user entity information includes user behavior entities and user behavior entity relationships, obtain user behavior sets according to the user behavior entities and the user behavior entity relationships, and obtain usage frequency according to the user behavior sets through TF-IDF algorithm. The calculation formula is expressed as: Among them, P i Indicates the frequency of use, A ij represents the number of times word i appears in j, N represents the user behavior set, n i represents the number of times word i appears in the user behavior set;

[0009] S3: assigning weights to the user behavior set according to the usage frequency and the time information to obtain entity weights, and obtaining user behavior labels according to the user behavior entities and the user behavior entity relationships, wherein the time information includes browsing record time;

[0010] S4: performing binary grouping according to the user behavior label and the entity weight, encrypting according to the binary grouping to obtain corresponding ciphertext blocks, fusing the ciphertext blocks to obtain ciphertext data, and transmitting the ciphertext data to the cloud server through a channel;

[0011] S5: The cloud server decrypts the ciphertext data to obtain the user behavior tag and the entity weight, obtains the pre-pushed content by calling the network database according to the user behavior tag, extracts features according to the pre-pushed content to obtain a push feature vector, extracts features according to the user behavior tag combined with the entity weight to obtain a user feature vector, and calculates the correlation according to the push feature vector and the user push feature vector by using the cosine formula;

[0012] S6: sorting the pushed contents according to the relevance, and transmitting the sorted pushed contents to the display terminal through a channel according to the user IP address.

[0013] Preferably, the step S2 specifically includes the following steps:

[0014] S201: performing word segmentation processing on the user behavior data through the Jieba word segmentation model to obtain a user behavior entity corpus, annotating the user behavior data according to predefined entity types to obtain annotation information, and obtaining user behavior training data according to the annotation information;

[0015] S202: Processing the user behavior training data through a BERT model to obtain a user behavior word vector, processing the user behavior word vector through a bidirectional neural network BiLSTM model to obtain deep semantic information, processing the deep semantic information through a CRF model to obtain a maximum probability annotation sequence, and outputting the user behavior entity according to the maximum probability annotation sequence;

[0016] S203: Extract features based on the user behavior word vector to obtain a feature vector, perform a convolution operation on the feature vector through a convolutional neural network to obtain a convolution result, perform maximum pooling on the convolution result to obtain a maximum eigenvalue, and output the user behavior entity relationship through a convolutional neural network.

[0017] Preferably, the step S3 specifically includes: Calculate the entity weight, Y i represents the entity weight of the i-th user entity information, Max() represents the maximum value, and the formula is used according to the browsing record time. Among them, T i represents the entity weight of the i-th user entity information, T1 represents the user browsing start time, T2 represents the user browsing end time, Ti1 represents the browsing start time of the i-th user entity information, and Ti2 represents the browsing end time of the i-th user entity information.

[0018] Preferably, the step S4 specifically includes the following steps:

[0019] S401: dividing the user behavior label and the entity weight into information blocks of preset length, reading characters in the information blocks respectively, and converting them into binary numbers to obtain the binary groups;

[0020] S402: Process the binary grouping by an encryption algorithm to obtain the ciphertext block. The calculation formula is: K i =m i +R+PQ, where K i represents the ciphertext block of the i-th group, m i represents the binary group of the i-th group, R represents the random value in the encryption process, P represents the encryption key, and Q represents the encryption parameter;

[0021] S403: performing addition processing on the obtained ciphertext blocks to obtain the ciphertext data.

[0022] Preferably, the step S5 specifically includes expressing the cosine formula as: in is the pushed feature vector, is the user behavior feature vector, is the degree of relevance;

[0023] An intelligent push system based on user data, comprising:

[0024] A data collection module is used to obtain user data, including user IP addresses and network action information;

[0025] A data processing module, configured to construct user behavior data according to the network action information, obtain user entity information by information extraction according to the user behavior data, wherein the user entity information includes user behavior entities and user behavior entity relationships, obtain a user behavior set according to the user behavior entities and the user behavior entity relationships, obtain a usage frequency according to the user behavior set by processing through a TF-IDF algorithm, obtain an entity weight by weight matching the user behavior set according to the usage frequency and the time information, and obtain a user behavior label according to the user behavior entity and the user behavior entity relationship;

[0026] A data encryption module, used for performing binary grouping according to the user behavior label and the entity weight, encrypting according to the binary grouping to obtain corresponding ciphertext blocks, fusing the ciphertext blocks to obtain ciphertext data, and transmitting the ciphertext data to the cloud server through a channel;

[0027] Intelligent push module, the cloud server decrypts the ciphertext data to obtain the user behavior label and the entity weight, obtains pre-push content by calling a network database according to the user behavior label, extracts features according to the pre-push content to obtain a push feature vector, extracts features according to the user behavior label combined with the entity weight to obtain a user feature vector, calculates the correlation by using the cosine formula according to the push feature vector and the user push feature vector, sorts the push content according to the correlation, and transmits the sorted push content to the display terminal through a channel according to the user IP address.

[0028] The beneficial effects of the present invention are:

[0029] 1. The purpose of extracting user behavior entities and user behavior entity relationships is to retain the nouns and verbs in the user behavior data and filter the user behavior data. The retained nouns and verbs, on the one hand, eliminate the non-text information and some modal particles in the user behavior data, and on the other hand, simplify the content of the user behavior data, ensure the integrity of the information, and improve the accuracy of information push.

[0030] 2. Binary grouping is performed according to the user behavior label and the entity weight, encryption is performed according to the binary grouping to obtain corresponding ciphertext blocks, the ciphertext blocks are merged to obtain ciphertext data, and the ciphertext data is transmitted to the cloud server through the channel. After the user data processing is completed, the user feature data is encrypted to avoid data interception during the transmission process, which may lead to user data leakage. The advantage is that the cloud server can process the push data without having to contact the plaintext data, thereby ensuring the security of the data transmission process. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] In order to facilitate understanding by those skilled in the art, the present invention is further described below with reference to the accompanying drawings.

[0032] Figure 1 It is a flow chart of the intelligent push method based on user data of the present invention. DETAILED DESCRIPTION

[0033] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the specific implementation methods, structures, features and effects of the present invention are described in detail below in conjunction with the accompanying drawings and preferred embodiments.

[0034] See also Figure 1 , an intelligent push method based on user data, comprising the following steps:

[0035] S1: Acquire user data, where the user data includes the user's IP address and network action information, where the network action information carries time information;

[0036] S2: construct user behavior data according to the network action information, obtain user entity information through information extraction according to the user behavior data, the user entity information includes user behavior entities and user behavior entity relationships, obtain user behavior sets according to the user behavior entities and the user behavior entity relationships, and obtain usage frequency according to the user behavior sets through TF-IDF algorithm. The calculation formula is expressed as: Among them, P i Indicates the frequency of use, A ij represents the number of times word i appears in j, N represents the user behavior set, n i represents the number of times word i appears in the user behavior set;

[0037] S3: assigning weights to the user behavior set according to the usage frequency and the time information to obtain entity weights, and obtaining user behavior labels according to the user behavior entities and the user behavior entity relationships, wherein the time information includes browsing record time;

[0038] S4: performing binary grouping according to the user behavior label and the entity weight, encrypting according to the binary grouping to obtain corresponding ciphertext blocks, fusing the ciphertext blocks to obtain ciphertext data, and transmitting the ciphertext data to the cloud server through a channel;

[0039] S5: The cloud server decrypts the ciphertext data to obtain the user behavior tag and the entity weight, obtains the pre-pushed content by calling the network database according to the user behavior tag, extracts features according to the pre-pushed content to obtain a push feature vector, extracts features according to the user behavior tag combined with the entity weight to obtain a user feature vector, and calculates the correlation according to the push feature vector and the user push feature vector by using the cosine formula;

[0040] S6: sorting the pushed contents according to the relevance, and transmitting the sorted pushed contents to the display terminal through a channel according to the user IP address.

[0041] It should be noted that network action information includes search records, browsing records and social interaction information. Social interaction information includes comments on posts and interest tendency information. At a certain time node, the data that users have recently browsed should be what they are more interested in recently, and they will be given a larger weight. A smaller weight will be given to earlier browsing and search data to improve the accuracy of information push.

[0042] Specifically, the step S2 comprises the following steps:

[0043] S201: performing word segmentation processing on the user behavior data through the Jieba word segmentation model to obtain a user behavior entity corpus, annotating the user behavior data according to predefined entity types to obtain annotation information, and obtaining user behavior training data according to the annotation information;

[0044] S202: Processing the user behavior training data through a BERT model to obtain a user behavior word vector, processing the user behavior word vector through a bidirectional neural network BiLSTM model to obtain deep semantic information, processing the deep semantic information through a CRF model to obtain a maximum probability annotation sequence, and outputting the user behavior entity according to the maximum probability annotation sequence;

[0045] S203: Extract features based on the user behavior word vector to obtain a feature vector, perform a convolution operation on the feature vector through a convolutional neural network to obtain a convolution result, perform maximum pooling on the convolution result to obtain a maximum eigenvalue, and output the user behavior entity relationship through a convolutional neural network.

[0046] The purpose of extracting user behavior entities and user behavior entity relationships is to retain the nouns and verbs in the user behavior data and to filter the user behavior data. The retained nouns and verbs, on the one hand, eliminate the non-text information and some modal particles of the user behavior data, and on the other hand, simplify the content of the user behavior data and ensure the integrity of the information.

[0047] Specifically, the step S3 specifically includes calculating the frequency of use by the formula Calculate the entity weight, Y i represents the entity weight of the i-th user entity information, Max() represents the maximum value, and the formula is used according to the browsing record time. Among them, T i represents the entity weight of the i-th user entity information, T1 represents the user browsing start time, T2 represents the user browsing end time, Ti1 represents the browsing start time of the i-th user entity information, and Ti2 represents the browsing end time of the i-th user entity information.

[0048] User behaviors are weighted according to usage frequency and time information, and the weight assignment can reflect the specific degree of interest bias.

[0049] Specifically, the step S4 comprises the following steps:

[0050] S401: dividing the user behavior label and the entity weight into information blocks of preset length, reading characters in the information blocks respectively, and converting them into binary numbers to obtain the binary groups;

[0051] S402: Process the binary grouping by an encryption algorithm to obtain the ciphertext block. The calculation formula is: K i =m i +R+PQ, where K i represents the ciphertext block of the i-th group, m i represents the binary group of the i-th group, R represents the random value in the encryption process, P represents the encryption key, and Q represents the encryption parameter;

[0052] S403: performing addition processing on the obtained ciphertext blocks to obtain the ciphertext data.

[0053] The process of big data collection, analysis, and transmission is accompanied by the problem of sensitive data leakage. Similarly, in the field of intelligent push technology based on user data, after the user data is processed, the user feature data is encrypted to avoid data interception during transmission and user data leakage. The advantage is that the cloud server can process the push data without having to contact the plaintext data, thus ensuring the security of the data transmission process.

[0054] Preferably, the step S5 specifically includes expressing the cosine formula as: in is the pushed feature vector, is the user behavior feature vector, is the degree of relevance;

[0055] Furthermore, the present application also provides an intelligent push system based on user data, including:

[0056] A data collection module is used to obtain user data, including user IP addresses and network action information;

[0057] A data processing module, configured to construct user behavior data according to the network action information, obtain user entity information by information extraction according to the user behavior data, wherein the user entity information includes user behavior entities and user behavior entity relationships, obtain a user behavior set according to the user behavior entities and the user behavior entity relationships, obtain a usage frequency according to the user behavior set by processing through a TF-IDF algorithm, obtain an entity weight by weight matching the user behavior set according to the usage frequency and the time information, and obtain a user behavior label according to the user behavior entity and the user behavior entity relationship;

[0058] A data encryption module, used for performing binary grouping according to the user behavior label and the entity weight, encrypting according to the binary grouping to obtain corresponding ciphertext blocks, fusing the ciphertext blocks to obtain ciphertext data, and transmitting the ciphertext data to the cloud server through a channel;

[0059] Intelligent push module, the cloud server decrypts the ciphertext data to obtain the user behavior label and the entity weight, obtains pre-push content by calling a network database according to the user behavior label, extracts features according to the pre-push content to obtain a push feature vector, extracts features according to the user behavior label combined with the entity weight to obtain a user feature vector, calculates the correlation by using the cosine formula according to the push feature vector and the user push feature vector, sorts the push content according to the correlation, and transmits the sorted push content to the display terminal through a channel according to the user IP address.

[0060] The working principle and use process of the present invention:

[0061] Obtain user data, construct user behavior data based on network action information, obtain user entity information through information extraction based on user behavior data, obtain usage frequency based on user behavior set through TF-IDF algorithm, assign weights to user behavior set based on usage frequency and time information to obtain entity weights, obtain user behavior tags based on user behavior entities and user behavior entity relationships, perform binary grouping based on user behavior tags and entity weights, encrypt the binary groups to obtain ciphertext data, the cloud server decrypts the ciphertext data to obtain user behavior tags and entity weights, obtain pre-pushed content by calling the network database based on the user behavior tags, obtain relevance based on the push feature vector and the user push feature vector, sort the pushed content based on the relevance, and transmit the sorted pushed content to the display terminal based on the user IP address.

[0062] The program code included in the method in the embodiment of the present invention can be transmitted with any appropriate medium, including but not limited to wireless, electric wire, optical cable, etc., or any suitable combination mentioned above. The computer program code for performing the operation of the present invention can be written with one or more programming languages ​​or their combination, and the programming language includes object-oriented programming languages--such as Java, Smalltalk, C++, and also includes conventional procedural programming languages--such as "C" language or similar programming languages. The program code can be executed completely on the user's computer, partially on the user's computer, executed as an independent software package, partially on the user's computer and partially on the remote computer, or executed completely on the remote computer or server. In the case of a remote computer, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computer (for example, using an Internet service provider to connect through the Internet).

[0063] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Although the present invention has been disclosed as a preferred embodiment as above, it is not used to limit the present invention. Any technical personnel in this field can make some changes or modify the technical contents disclosed above into equivalent embodiments without departing from the scope of the technical solution of the present invention. However, any brief modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention still fall within the scope of the technical solution of the present invention.

Claims

1. An intelligent push method based on user data, characterized in that: The following steps are involved: S1: Acquire user data, where the user data includes the user's IP address and network action information, where the network action information carries time information; S2: construct user behavior data according to the network action information, obtain user entity information through information extraction according to the user behavior data, the user entity information includes user behavior entities and user behavior entity relationships, obtain user behavior sets according to the user behavior entities and the user behavior entity relationships, and obtain usage frequency according to the user behavior sets through TF-IDF algorithm. The calculation formula is expressed as: Among them, P i Indicates the frequency of use, A ij represents the number of times word i appears in j, N represents the user behavior set, n i represents the number of times word i appears in the user behavior set; S3: assigning weights to the user behavior set according to the usage frequency and the time information to obtain entity weights, and obtaining user behavior labels according to the user behavior entities and the user behavior entity relationships, wherein the time information includes browsing record time; S4: performing binary grouping according to the user behavior label and the entity weight, encrypting according to the binary grouping to obtain corresponding ciphertext blocks, fusing the ciphertext blocks to obtain ciphertext data, and transmitting the ciphertext data to the cloud server through a channel; S5: The cloud server decrypts the ciphertext data to obtain the user behavior tag and the entity weight, obtains the pre-pushed content by calling the network database according to the user behavior tag, extracts features according to the pre-pushed content to obtain a push feature vector, extracts features according to the user behavior tag combined with the entity weight to obtain a user feature vector, and calculates the correlation according to the push feature vector and the user push feature vector by using the cosine formula; S6: sorting the pushed contents according to the relevance, and transmitting the sorted pushed contents to the display terminal through a channel according to the user IP address.

2. The intelligent push method based on user data according to claim 1, characterized in that: The step S2 specifically includes the following steps: S201: performing word segmentation processing on the user behavior data through the Jieba word segmentation model to obtain a user behavior entity corpus, annotating the user behavior data according to predefined entity types to obtain annotation information, and obtaining user behavior training data according to the annotation information; S202: Processing the user behavior training data through a BERT model to obtain a user behavior word vector, processing the user behavior word vector through a bidirectional neural network BiLSTM model to obtain deep semantic information, processing the deep semantic information through a CRF model to obtain a maximum probability annotation sequence, and outputting the user behavior entity according to the maximum probability annotation sequence; S203: Extract features based on the user behavior word vector to obtain a feature vector, perform a convolution operation on the feature vector through a convolutional neural network to obtain a convolution result, perform maximum pooling on the convolution result to obtain a maximum eigenvalue, and output the user behavior entity relationship through a convolutional neural network.

3. The intelligent push method based on user data according to claim 1, characterized in that: The step S3 specifically includes calculating the frequency of use by the formula Calculate the entity weight, Y i represents the entity weight of the i-th user entity information, Max() represents the maximum value, and the formula is used according to the browsing record time. Among them, T i represents the entity weight of the i-th user entity information, T1 represents the user browsing start time, T2 represents the user browsing end time, Ti1 represents the browsing start time of the i-th user entity information, and Ti2 represents the browsing end time of the i-th user entity information.

4. The intelligent push method based on user data according to claim 1, characterized in that: The step S4 specifically comprises the following steps: S401: dividing the user behavior label and the entity weight into information blocks of preset length, reading characters in the information blocks respectively, and converting them into binary numbers to obtain the binary groups; S402: Process the binary grouping by an encryption algorithm to obtain the ciphertext block. The calculation formula is: K i =m i +R+PQ, where K i represents the ciphertext block of the i-th group, m i represents the binary group of the i-th group, R represents the random value in the encryption process, P represents the encryption key, and Q represents the encryption parameter; S403: performing addition processing on the obtained ciphertext blocks to obtain the ciphertext data.

5. The intelligent push method based on user data according to claim 1, characterized in that: The step S5 specifically includes the cosine formula being expressed as: in, represents the pushed feature vector, represents the user behavior feature vector, Indicates the degree of correlation.

6. An intelligent push system based on user data, the intelligent push system based on user data adopts the intelligent push method based on user data as claimed in claim 1, characterized in that: include: A data collection module is used to obtain user data, including user IP addresses and network action information; A data processing module, configured to construct user behavior data according to the network action information, obtain user entity information by information extraction according to the user behavior data, wherein the user entity information includes user behavior entities and user behavior entity relationships, obtain a user behavior set according to the user behavior entities and the user behavior entity relationships, obtain a usage frequency according to the user behavior set by processing through a TF-IDF algorithm, obtain an entity weight by weight matching the user behavior set according to the usage frequency and the time information, and obtain a user behavior label according to the user behavior entity and the user behavior entity relationship; A data encryption module, used for performing binary grouping according to the user behavior label and the entity weight, encrypting according to the binary grouping to obtain corresponding ciphertext blocks, fusing the ciphertext blocks to obtain ciphertext data, and transmitting the ciphertext data to the cloud server through a channel; Intelligent push module, the cloud server decrypts the ciphertext data to obtain the user behavior label and the entity weight, obtains pre-push content by calling a network database according to the user behavior label, extracts features according to the pre-push content to obtain a push feature vector, extracts features according to the user behavior label combined with the entity weight to obtain a user feature vector, calculates the correlation by using the cosine formula according to the push feature vector and the user push feature vector, sorts the push content according to the correlation, and transmits the sorted push content to the display terminal through a channel according to the user IP address.