CN116955831A

CN116955831A - Information recommendation method and system based on data mining and AI model analysis

Info

Publication number: CN116955831A
Application number: CN202311142002.XA
Authority: CN
Inventors: 杨宇; 高蕾
Original assignee: Mianyang Hengyongsheng Technology Co ltd
Current assignee: Mianyang Hengyongsheng Technology Co ltd
Priority date: 2023-09-06
Filing date: 2023-09-06
Publication date: 2023-10-27

Abstract

The invention provides an information recommendation method and system based on data mining and AI model analysis, and relates to the technical field of artificial intelligence. In the invention, matrix key information description vectors corresponding to candidate user distribution matrixes are mined; obtaining a matching parameter of the analysis matrix area and the user group relation according to the matrix key information description vector; marking a screening matrix area based on the user population relation matching parameters, and analyzing a first to-be-determined matrix area distribution coordinate and a first relation matching evaluation parameter, a second to-be-determined matrix area distribution coordinate and a second relation matching evaluation parameter according to the screening matrix area; and performing association information recommendation operation among the group association users based on the screening matrix area, the first to-be-determined matrix area distribution coordinates, the first relationship matching evaluation parameters, the second to-be-determined matrix area distribution coordinates and the second relationship matching evaluation parameters. Based on the above, the efficiency of information recommendation can be improved to some extent.

Description

Information recommendation method and system based on data mining and AI model analysis

Technical Field

The invention relates to the technical field of artificial intelligence, in particular to an information recommendation method and system based on data mining and AI model analysis.

Background

The application scene of information recommendation is more, such as recommendation of image data, text data, audio data and the like, namely, the information to be recommended is pushed to a required user, however, in the prior art, generally, recommendation operation is carried out for each user independently, and the problem of low efficiency exists.

Disclosure of Invention

In view of the above, the present invention aims to provide an information recommendation method and system based on data mining and AI model analysis, so as to improve the efficiency of information recommendation to a certain extent.

In order to achieve the above purpose, the embodiment of the present invention adopts the following technical scheme:

an information recommendation method based on data mining and AI model analysis, the information recommendation method comprising:

determining a candidate user distribution matrix, wherein each matrix parameter in the candidate user distribution matrix corresponds to one candidate user, and in the candidate user distribution matrix, the distribution relation between matrix parameters is related to the distribution relation between the candidate users, and the matrix parameters are used for reflecting the user characteristic information of the candidate users;

utilizing a key information mining unit included in a target user group analysis network formed by performing network optimization operation to mine a matrix key information description vector corresponding to the candidate user distribution matrix;

According to the matrix key information description vector, an analysis output unit included in the target user group analysis network is utilized to obtain a corresponding analysis matrix area and user group relation matching parameters, wherein the user group relation matching parameters are in one-to-one correspondence with the analysis matrix area;

marking an analysis matrix area under the condition that the user group relation matching parameter is not smaller than a reference relation matching parameter so as to be marked as a screening matrix area, analyzing a first to-be-determined matrix area distribution coordinate and a first relation matching evaluation parameter by utilizing a user group analysis front-end unit included in the target user group analysis network according to the screening matrix area, and analyzing a second to-be-determined matrix area distribution coordinate and a second relation matching evaluation parameter by utilizing a user group analysis rear-end unit included in the target user group analysis network according to the screening matrix area;

analyzing a target matrix area of the candidate user distribution matrix and a user group relationship matching parameter of the candidate user distribution matrix based on the screening matrix area, the first to-be-determined matrix area distribution coordinates, the first relationship matching evaluation parameter, the second to-be-determined matrix area distribution coordinates and the second relationship matching evaluation parameter;

And under the condition that the user group relation matching parameters of the candidate user distribution matrix meet a preset matching rule, taking the candidate users corresponding to each matrix parameter included in the target matrix area as group associated users, and performing associated information recommendation operation among the group associated users.

In some preferred embodiments, in the information recommendation method based on data mining and AI model analysis, the network optimization process of the target user group analysis network includes:

determining an exemplary user distribution matrix, wherein the exemplary user distribution matrix is marked with an exemplary matrix area, each matrix parameter in the exemplary user distribution matrix corresponds to one exemplary user, and in the exemplary user distribution matrix, the distribution relation among matrix parameters is related to the distribution relation among the exemplary users, and the matrix parameters are used for reflecting the user characteristic information of the exemplary users;

analyzing a first number of analysis matrix areas and a first number of user population relation matching parameters in the exemplary user distribution matrix by using an initial user population analysis network, wherein each user population relation matching parameter corresponds to one analysis matrix area;

Constructing an exemplary data cluster based on the exemplary matrix area, the first number of analytical matrix areas and the first number of user population relationship matching parameters, wherein different exemplary data clusters are determined according to different reference coincidence relation characterization parameters, the exemplary data clusters comprise low-coincidence exemplary data clusters and high-coincidence exemplary data clusters, the low-coincidence exemplary data clusters comprise exemplary matching data with coincidence relation characterization parameters not smaller than predetermined low-reference coincidence relation characterization parameters, the high-coincidence exemplary data clusters comprise exemplary matching data with coincidence relation characterization parameters not smaller than predetermined high-reference coincidence relation characterization parameters, and the low-reference coincidence relation characterization parameters are smaller than the high-reference coincidence relation characterization parameters;

according to the low-coincidence exemplary data cluster, a user group analysis front-end unit included in the initial user group analysis network is utilized to obtain a corresponding front-end analysis data cluster, and the front-end analysis data cluster comprises a relationship matching evaluation parameter corresponding to each exemplary data in the low-coincidence exemplary data cluster and a distribution coordinate of a matrix area to be determined;

According to the high-coincidence exemplary data cluster, utilizing a user group analysis back-end unit included in the initial user group analysis network to obtain a corresponding back-end analysis data cluster, wherein the back-end analysis data cluster comprises a relationship matching evaluation parameter corresponding to each exemplary data in the high-coincidence exemplary data cluster and a distribution coordinate of a matrix area to be determined;

and performing network optimization operation on the initial user group analysis network based on the exemplary data cluster, the front end analysis data cluster and the back end analysis data cluster to form a corresponding target user group analysis network, wherein the low-coincidence exemplary data cluster and the front end analysis data cluster are used for performing network optimization operation on the user group analysis front end unit, the high-coincidence exemplary data cluster and the back end analysis data cluster are used for performing network optimization operation on the user group analysis back end unit, the user group analysis front end unit and the user group analysis back end unit are different, and the user group analysis front end unit and the user group analysis back end unit are connected in sequence.

In some preferred embodiments, in the information recommendation method based on data mining and AI model analysis, the step of analyzing, using an initial user population analysis network, a first number of analysis matrix areas and a first number of user population relationship matching parameters in the exemplary user distribution matrix includes:

According to the exemplary user distribution matrix, a key information mining unit included in the initial user group analysis network is utilized to mine out a corresponding matrix key information description vector;

and according to the matrix key information description vector, obtaining a corresponding first number of analysis matrix areas and a corresponding first number of user group relation matching parameters by utilizing an analysis output unit included in the initial user group analysis network.

In some preferred embodiments, in the information recommendation method based on data mining and AI model analysis, the step of mining, according to the exemplary user distribution matrix, a corresponding matrix key information description vector by using a key information mining unit included in the initial user group analysis network includes:

according to the exemplary user distribution matrix, utilizing a key information mining structure in the key information mining unit to mine out a corresponding initial key information description vector;

determining a corresponding front-end gradient optimization description vector by utilizing a front-end gradient optimization structure in the key information mining unit according to the initial key information description vector;

determining a corresponding rear-end gradient optimization description vector by utilizing a rear-end gradient optimization structure in the key information mining unit according to the front-end gradient optimization description vector;

And determining a matrix key information description vector corresponding to the exemplary user distribution matrix by utilizing at least one front-end gradient optimization structure and at least one back-end gradient optimization structure in the key information mining unit according to the back-end gradient optimization description vector.

In some preferred embodiments, in the information recommendation method based on data mining and AI model analysis, the step of constructing an exemplary data cluster based on the exemplary matrix area, the first number of analysis matrix areas, and the first number of user population relationship matching parameters includes:

based on the first number of user group relation matching parameters, a second number of analysis matrix areas are selected from the first number of analysis matrix areas and marked as screening matrix areas, so that a second number of screening matrix areas are formed, the user group relation matching parameters corresponding to the screening matrix areas are not smaller than the reference relation matching parameters, and the second number is not larger than the first number;

analyzing coincidence relation characterization parameters between each screening matrix region and the exemplary matrix region based on the second number of screening matrix regions;

Determining a corresponding low-coincidence exemplary data cluster based on coincidence relation characterization parameters between each screening matrix region and the exemplary matrix region, wherein the low-coincidence exemplary data cluster comprises first exemplary matching data and exemplary non-matching data;

and determining corresponding high-coincidence exemplary data clusters based on coincidence relation characterization parameters between each screening matrix region and each exemplary matrix region, wherein the high-coincidence exemplary data clusters comprise second exemplary matching data and exemplary non-matching data.

In some preferred embodiments, in the information recommendation method based on data mining and AI model analysis, the step of extracting a second number of analysis matrix areas from the first number of analysis matrix areas based on the first number of user population relationship matching parameters and marking the second number of analysis matrix areas as screening matrix areas, thereby forming a second number of screening matrix areas includes:

marking an analysis matrix area corresponding to the user group relation matching parameter under the condition that the user group relation matching parameter is not smaller than the reference relation matching parameter so as to be marked as a screening matrix area, wherein the user group relation matching parameter belongs to the first number of user group relation matching parameters, and the screening matrix area belongs to the second number of screening matrix areas;

And under the condition that the user group relation matching parameters are smaller than the reference relation matching parameters, discarding the analysis matrix areas corresponding to the user group relation matching parameters in the first number of analysis matrix areas to form a second number of screening matrix areas.

In some preferred embodiments, in the information recommendation method based on data mining and AI model analysis, the step of determining the corresponding low-coincidence exemplary data cluster based on coincidence relation characterization parameters between each of the filtering matrix areas and the exemplary matrix areas includes:

marking the corresponding exemplary data of the screening matrix region to be marked as first exemplary matching data in a low-coincidence exemplary data cluster under the condition that the coincidence relation characterization parameter between the screening matrix region and the exemplary matrix region is not less than the low-reference coincidence relation characterization parameter;

marking the corresponding exemplary data of the screening matrix region to be the exemplary non-matching data in the low-coincidence exemplary data cluster under the condition that the coincidence relation representation parameter between the screening matrix region and the exemplary matrix region is smaller than the appointed coincidence relation representation parameter, wherein the appointed coincidence relation representation parameter is smaller than the low-reference coincidence relation representation parameter;

The step of determining the corresponding high-coincidence exemplary data cluster based on the coincidence relation characterization parameter between each screening matrix region and the exemplary matrix region includes:

marking the corresponding exemplary data of the screening matrix region to be marked as second exemplary matching data in a high-coincidence exemplary data cluster under the condition that the coincidence relation representation parameter between the screening matrix region and the exemplary matrix region is not smaller than the high-reference coincidence relation representation parameter;

and marking the exemplary data corresponding to the screening matrix region to be the exemplary non-matching data in the high-coincidence exemplary data cluster under the condition that the coincidence relation representation parameter between the screening matrix region and the exemplary matrix region is smaller than the appointed coincidence relation representation parameter, wherein the appointed coincidence relation representation parameter is smaller than the high-reference coincidence relation representation parameter.

In some preferred embodiments, in the information recommendation method based on data mining and AI model analysis, the step of obtaining, according to the low-coincidence exemplary data cluster, a corresponding front-end analysis data cluster by using a user group analysis front-end unit included in the initial user group analysis network includes:

Determining a corresponding low-coincidence mapping feature representation by utilizing a front-end region mapping unit according to the to-be-analyzed exemplary data in the low-coincidence exemplary data cluster, wherein the to-be-analyzed exemplary data in the low-coincidence exemplary data cluster belongs to the exemplary matching data or the exemplary non-matching data in the low-coincidence exemplary data cluster;

analyzing front-end analysis data corresponding to the to-be-analyzed exemplary data in the low-coincidence exemplary data cluster by utilizing the user population analysis front-end unit according to the low-coincidence mapping characteristic representation, wherein the front-end analysis data comprises a first relation matching evaluation parameter and a first to-be-determined matrix region distribution coordinate corresponding to the to-be-analyzed exemplary data in the low-coincidence exemplary data cluster;

the step of obtaining a corresponding back-end analysis data cluster by using a user group analysis back-end unit included in the initial user group analysis network according to the high-coincidence exemplary data cluster includes:

determining a corresponding high-coincidence mapping feature representation by utilizing a rear-end region mapping unit according to the to-be-analyzed exemplary data in the high-coincidence exemplary data cluster, wherein the to-be-analyzed exemplary data in the high-coincidence exemplary data cluster belongs to the exemplary matching data or the exemplary non-matching data in the high-coincidence exemplary data cluster;

And analyzing back-end analysis data corresponding to the to-be-analyzed exemplary data in the high-coincidence exemplary data cluster by utilizing the user population analysis back-end unit according to the high-coincidence mapping characteristic representation, wherein the back-end analysis data comprises second relation matching evaluation parameters corresponding to the to-be-analyzed exemplary data in the high-coincidence exemplary data cluster and second to-be-determined matrix region distribution coordinates.

In some preferred embodiments, in the information recommendation method based on data mining and AI model analysis, the step of performing network optimization operation on the initial user group analysis network to form a corresponding target user group analysis network based on the exemplary data cluster, the front end analysis data cluster and the back end analysis data cluster includes:

based on the low-coincidence exemplary data cluster and the front-end analysis data cluster, optimizing and adjusting parameters of the user group analysis front-end unit according to a first network optimization rule;

based on the exemplary matrix area and the front-end analysis data cluster, optimizing and adjusting parameters of the user group analysis front-end unit according to a network optimization second rule;

Based on the high-coincidence exemplary data cluster and the back-end analysis data cluster, optimizing and adjusting parameters of the user group analysis back-end unit according to the network optimization first rule;

and based on the exemplary matrix area and the back-end analysis data cluster, optimizing and adjusting parameters of the user group analysis back-end unit according to a second network optimization rule.

The embodiment of the invention also provides an information recommendation system based on data mining and AI model analysis, which comprises a processor and a memory, wherein the memory is used for storing a computer program, and the processor is used for executing the computer program so as to realize the information recommendation method based on the data mining and AI model analysis.

The information recommendation method and system based on data mining and AI model analysis provided by the embodiment of the invention can be used for firstly mining the matrix key information description vector corresponding to the candidate user distribution matrix; obtaining a matching parameter of the analysis matrix area and the user group relation according to the matrix key information description vector; marking a screening matrix area based on the user population relation matching parameters, and analyzing a first to-be-determined matrix area distribution coordinate and a first relation matching evaluation parameter, a second to-be-determined matrix area distribution coordinate and a second relation matching evaluation parameter according to the screening matrix area; and performing association information recommendation operation among the group association users based on the screening matrix area, the first to-be-determined matrix area distribution coordinates, the first relationship matching evaluation parameters, the second to-be-determined matrix area distribution coordinates and the second relationship matching evaluation parameters. Based on the foregoing, before the information recommendation operation is performed, the group associated users are determined, so that the associated information recommendation operation can be performed, and the efficiency of information recommendation can be improved to a certain extent.

In order to make the above objects, features and advantages of the present invention more comprehensible, preferred embodiments accompanied with figures are described in detail below.

Drawings

Fig. 1 is a block diagram of an information recommendation system based on data mining and AI model analysis according to an embodiment of the present invention.

Fig. 2 is a flowchart illustrating steps included in an information recommendation method based on data mining and AI model analysis according to an embodiment of the present invention.

Fig. 3 is a schematic diagram of each module included in the information recommendation device based on data mining and AI model analysis according to the embodiment of the present invention.

Detailed Description

For the purpose of making the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is apparent that the described embodiments are only some embodiments of the present invention, but not all embodiments of the present invention. The components of the embodiments of the present invention generally described and illustrated in the figures herein may be arranged and designed in a wide variety of different configurations.

Thus, the following detailed description of the embodiments of the invention, as presented in the figures, is not intended to limit the scope of the invention, as claimed, but is merely representative of selected embodiments of the invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.

As shown in fig. 1, an embodiment of the present invention provides an information recommendation system based on data mining and AI model analysis. The information recommendation system based on data mining and AI model analysis can include a memory and a processor.

In detail, the memory and the processor are electrically connected directly or indirectly to realize transmission or interaction of data. For example, electrical connection may be made to each other via one or more communication buses or signal lines. The memory may store at least one software functional module (computer program) that may exist in the form of software or firmware. The processor may be configured to execute the executable computer program stored in the memory, so as to implement the information recommendation method based on data mining and AI model analysis provided by the embodiment of the present invention.

It should be appreciated that in some possible embodiments, the Memory may be, but is not limited to, random access Memory (Random Access Memory, RAM), read Only Memory (ROM), programmable Read Only Memory (Programmable Read-Only Memory, PROM), erasable Read Only Memory (Erasable Programmable Read-Only Memory, EPROM), electrically erasable Read Only Memory (Electric Erasable Programmable Read-Only Memory, EEPROM), and the like.

It should be appreciated that in some possible embodiments, the processor may be a general purpose processor, including a central processing unit (Central Processing Unit, CPU), a network processor (Network Processor, NP), a System on Chip (SoC), etc.; but also Digital Signal Processors (DSPs), application Specific Integrated Circuits (ASICs), field Programmable Gate Arrays (FPGAs) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

It should be appreciated that in some possible embodiments, the information recommendation system based on data mining and AI model analysis may be a server with data processing capabilities.

With reference to fig. 2, the embodiment of the invention further provides an information recommendation method based on data mining and AI model analysis, which can be applied to the information recommendation system based on data mining and AI model analysis. The method steps defined by the flow related to the information recommendation method based on the data mining and the AI model analysis can be realized by the information recommendation system based on the data mining and the AI model analysis.

The specific flow shown in fig. 2 will be described in detail.

Step S110, determining a candidate user distribution matrix.

In the embodiment of the invention, the information recommendation system based on data mining and AI model analysis can determine a candidate user distribution matrix. Each matrix parameter in the candidate user distribution matrix corresponds to one candidate user, and in the candidate user distribution matrix, a distribution relation among matrix parameters is related to a distribution relation among the candidate users, and the matrix parameters are used for reflecting user characteristic information of the candidate users, wherein the user characteristic information can be various response actions of the candidate users on different recommended information historically, such as network actions of no-click, click viewing, collection, forwarding and the like. The user characteristic information may be various network behaviors recorded in text form, or may be represented by converting the network behaviors recorded in text into digital characters, and different network behaviors may be converted into different numbers based on a preset conversion relationship.

And step S120, utilizing a key information mining unit included in a target user group analysis network formed by performing network optimization operation to mine out matrix key information description vectors corresponding to the candidate user distribution matrix.

In the embodiment of the invention, the information recommendation system based on data mining and AI model analysis can use a key information mining unit included in a target user group analysis network formed by performing network optimization operation to mine a matrix key information description vector corresponding to the candidate user distribution matrix, that is, can perform key information mining operation, namely feature mining, on the candidate user distribution matrix to represent corresponding key information in a vector form, so that a matrix key information description vector corresponding to the candidate user distribution matrix can be formed.

And step S130, according to the matrix key information description vector, obtaining a corresponding analysis matrix area and a user group relation matching parameter by utilizing an analysis output unit included in the target user group analysis network.

In the embodiment of the invention, the information recommendation system based on data mining and AI model analysis can obtain the corresponding analysis matrix area and user group relation matching parameters by utilizing the analysis output unit included in the target user group analysis network according to the matrix key information description vector. The user group relation matching parameters are in one-to-one correspondence with the analysis matrix areas, that is, for the obtained multiple analysis matrix areas and the obtained multiple user group relation matching parameters, the user group relation matching parameters can be used for reflecting the possibility that the candidate users corresponding to each matrix parameter included in the corresponding analysis matrix areas have a specific user group relation, or the degree of matching between the user group formed by the corresponding candidate users and the specific user group relation. The specific user group relationship may be configured according to actual requirements, for example, in a network optimization process of the target user group analysis network, an exemplary user group relationship corresponding to an exemplary matrix area in an exemplary user distribution matrix may be used as the specific user group relationship. That is, the target user group analysis network may learn, in the exemplary user distribution matrix, exemplary user group relationships that each exemplary user corresponding to an exemplary matrix region has.

And step S140, marking the analysis matrix area to be a screening matrix area under the condition that the user group relation matching parameter is not smaller than the reference relation matching parameter, analyzing a first to-be-determined matrix area distribution coordinate and a first relation matching evaluation parameter by utilizing a user group analysis front-end unit included in the target user group analysis network according to the screening matrix area, and analyzing a second to-be-determined matrix area distribution coordinate and a second relation matching evaluation parameter by utilizing a user group analysis rear-end unit included in the target user group analysis network according to the screening matrix area.

In the embodiment of the present invention, the information recommendation system based on data mining and AI model analysis may mark an analysis matrix area to be a screening matrix area when the user population relationship matching parameter is not less than a reference relationship matching parameter, analyze, according to the screening matrix area, a first to-be-determined matrix area distribution coordinate and a first relationship matching evaluation parameter by using a user population analysis front end unit included in the target user population analysis network, and analyze, according to the screening matrix area, a second to-be-determined matrix area distribution coordinate and a second relationship matching evaluation parameter by using a user population analysis back end unit included in the target user population analysis network. The reference relation matching parameters can be configured according to actual requirements, and specific numerical values of the reference relation matching parameters are not limited. In addition, the distribution coordinates of the area of the undetermined matrix may refer to the distribution coordinates of the analyzed area. That is, the parameters of the screening matrix area are analyzed by the user group analysis front-end unit included in the target user group analysis network to determine corresponding distribution coordinates and evaluation parameters, wherein the evaluation parameters are used for reflecting the matching degree of the relation with the specific user group.

Step S150, analyzing the target matrix area of the candidate user distribution matrix and the user population relationship matching parameters of the candidate user distribution matrix based on the screening matrix area, the first to-be-determined matrix area distribution coordinates, the first relationship matching evaluation parameters, the second to-be-determined matrix area distribution coordinates and the second relationship matching evaluation parameters.

In the embodiment of the present invention, the information recommendation system based on data mining and AI model analysis may analyze the target matrix area of the candidate user distribution matrix and the user population relationship matching parameters of the candidate user distribution matrix based on the screening matrix area, the first predetermined matrix area distribution coordinate, the first relationship matching evaluation parameter, the second predetermined matrix area distribution coordinate and the second relationship matching evaluation parameter. For example, the first to-be-determined matrix area distribution coordinates and the second to-be-determined matrix area distribution coordinates may refer to one distribution coordinate in the screening matrix area, so that average value or weighted summation calculation may be performed on the first to-be-determined matrix area distribution coordinates and the second to-be-determined matrix area distribution coordinates, an average value distribution coordinate corresponding to the one distribution coordinate may be obtained, so that, in combination with the size of the screening matrix area, a target matrix area corresponding to the candidate user distribution matrix may be determined, and average value or weighted summation calculation may be performed on the first relationship matching evaluation parameter and the second relationship matching evaluation parameter, to obtain a user population relationship matching parameter of the candidate user distribution matrix.

Step S160, taking the candidate users corresponding to each matrix parameter included in the target matrix area as group associated users and performing an associated information recommendation operation between the group associated users when the user group relationship matching parameters of the candidate user distribution matrix meet a preset matching rule.

In the embodiment of the invention, the information recommendation system based on data mining and AI model analysis can take candidate users corresponding to each matrix parameter included in the target matrix area as group associated users and perform associated information recommendation operation among the group associated users under the condition that the user group relationship matching parameters of the candidate user distribution matrix meet a preset matching rule. For example, when the user group relationship matching parameter of the candidate user distribution matrix is not smaller than a set value, candidate users corresponding to each matrix parameter included in the target matrix area may be used as group-associated users, where the set value may be configured according to actual requirements, and for example, the set value may be equal to a value of 0.8, 0.7, 0.9, or the like. In addition, the associated information recommending operation may refer to recommending information that has been historically recommended to one group-associated user to other group-associated users, or when it is determined that one piece of information (such as an image, text, etc.) to be processed needs to be recommended to one group-associated user, the information to be processed may be recommended to each group-associated user, for example, to improve the recommending efficiency.

Based on the foregoing, before the information recommendation operation is performed, the group associated users are determined, so that the associated information recommendation operation can be performed, and the efficiency of information recommendation can be improved to a certain extent.

It should be appreciated that, in some possible embodiments, the step S110 may further include the following specific implementation matters:

acquiring a plurality of candidate users, wherein the plurality of candidate users can be network users of the same user information recommendation platform, such as network users with network behaviors in the last period of time;

user characteristic information of each candidate user is obtained respectively;

for each candidate user, determining the interaction correlation degree between the candidate user and each other candidate user, and carrying out mean value calculation on the interaction correlation degree between the candidate user and each other candidate user to form a representative interaction correlation degree corresponding to the candidate user, wherein the interaction correlation degree is determined at least based on data of a target dimension, for example, the interaction correlation degree can be positively correlated with the interaction times between the candidate users, and can be determined by combining data of other dimensions, for example, can be further determined by combining the interaction data quantity between the candidate users or the friendly degree of the interaction data characterization, and the like, and the method is not particularly limited;

Determining a candidate user with the maximum representative interaction correlation degree as a center candidate user;

the user characteristic information corresponding to the center candidate user is used as matrix parameters corresponding to center distribution coordinates of a candidate user distribution matrix, and each other candidate user is ranked based on the interaction correlation degree between the center candidate user and each other candidate user, for example, the ranking is performed according to the sequence of the interaction correlation degree from large to small, the same interaction correlation degree can be randomly ranked, or the ranking can also be performed based on the similarity among the user characteristic information;

and sequentially determining a plurality of rectangular outlines with gradually increased sizes (each rectangular outline passes through each distribution coordinate) by taking the central distribution coordinate as the center, and sequentially distributing user characteristic information corresponding to other candidate users to each distribution coordinate in the candidate user distribution matrix according to the sequence from front to back of the rectangular outlines, so that the candidate user distribution matrix comprising each matrix parameter can be formed.

It should be appreciated that in some possible embodiments, the network optimization process of the target user group analysis network may further include the following specific implementation:

Determining an exemplary user distribution matrix, wherein the exemplary user distribution matrix is marked with an exemplary matrix area, each matrix parameter in the exemplary user distribution matrix corresponds to one exemplary user, and in the exemplary user distribution matrix, the distribution relation among matrix parameters is related to the distribution relation among the exemplary users, and the matrix parameters are used for reflecting user characteristic information of the exemplary users, such as the related description;

analyzing a first number of analysis matrix areas and a first number of user population relation matching parameters in the exemplary user distribution matrix by using an initial user population analysis network, wherein each user population relation matching parameter corresponds to one analysis matrix area, for example, a first number of rectangular area sizes can be predetermined, then, in the exemplary user distribution matrix, the first number of analysis matrix areas are cut out based on the first number of rectangular area sizes, and the possibility that each exemplary user corresponding to matrix parameters included in each analysis matrix area has an exemplary user population relation or the matching degree with the exemplary user population relation is analyzed, so that the first number of user population relation matching parameters can be obtained;

Constructing an exemplary data cluster based on the exemplary matrix area, the first number of analytical matrix areas and the first number of user population relationship matching parameters, wherein different exemplary data clusters are determined according to different reference coincidence relation characterization parameters, for example, the exemplary data cluster comprises a low coincidence exemplary data cluster and a high coincidence exemplary data cluster, the low coincidence exemplary data cluster comprises exemplary matching data of which the coincidence relation characterization parameter is not less than a predetermined low reference coincidence relation characterization parameter, the high coincidence exemplary data cluster comprises exemplary matching data of which the coincidence relation characterization parameter is not less than a predetermined high reference coincidence relation characterization parameter, the low reference coincidence relation characterization parameter is less than the high reference coincidence relation characterization parameter, the specific numerical values of the low reference coincidence relation characterization parameter and the high reference coincidence relation characterization parameter are not limited, and the low coincidence relation characterization parameter and the high reference coincidence relation characterization parameter can be configured according to actual requirements, such as 0.5, 0.6 and the like respectively;

according to the low-coincidence exemplary data cluster, a user group analysis front-end unit included in the initial user group analysis network is utilized to obtain a corresponding front-end analysis data cluster, the front-end analysis data cluster comprises a relationship matching evaluation parameter corresponding to each exemplary data in the low-coincidence exemplary data cluster and a distribution coordinate of a matrix area to be determined, and the relationship matching evaluation parameter can be used for reflecting the possibility that each exemplary user corresponding to the matrix parameter included in the matrix area corresponding to the distribution coordinate of the matrix area to be determined has an exemplary user group relationship;

According to the high-coincidence exemplary data cluster, utilizing a user group analysis back-end unit included in the initial user group analysis network to obtain a corresponding back-end analysis data cluster, wherein the back-end analysis data cluster comprises a relationship matching evaluation parameter corresponding to each exemplary data in the high-coincidence exemplary data cluster and a distribution coordinate of a matrix area to be determined, and the method is as described above;

based on the exemplary data cluster, the front-end analysis data cluster and the back-end analysis data cluster, the initial user group analysis network is subjected to network optimization operation to form a corresponding target user group analysis network, the low-coincidence exemplary data cluster and the front-end analysis data cluster are used for performing network optimization operation on the user group analysis front-end unit, the high-coincidence exemplary data cluster and the back-end analysis data cluster are used for performing network optimization operation on the user group analysis back-end unit, the user group analysis front-end unit and the user group analysis back-end unit are different, and the user group analysis front-end unit and the user group analysis back-end unit can be connected sequentially or can have parallel relations and the like.

It should be appreciated that, in some possible embodiments, the step of analyzing, using the initial user population analysis network, the first number of analysis matrix areas and the first number of user population relationship matching parameters in the exemplary user distribution matrix may further include the following specific implementation matters:

according to the exemplary user distribution matrix, a key information mining unit included in the initial user group analysis network is utilized to mine a corresponding matrix key information description vector, that is, for example, the key information mining unit included in the initial user group analysis network performs key information mining operation on the exemplary user distribution matrix, so that a corresponding matrix key information description vector can be obtained, and in addition, the key information mining unit can be a coding network;

according to the matrix key information description vector, an analysis output unit included in the initial user group analysis network is utilized to obtain a corresponding first number of analysis matrix areas and a first number of user group relation matching parameters, and the analysis output unit may perform full connection processing on the matrix key information description vector first, then perform multiple decimation (perform decimation according to a corresponding size) on the obtained full connection vector to form a first number of local full connection vectors, and then perform activation processing on each of the local full connection vectors to obtain a user group relation matching parameter corresponding to each of the local full connection vectors, and may determine an analysis matrix area in the exemplary user distribution matrix based on a corresponding size of each of the local full connection vectors, where the activation processing on each of the local full connection vectors may be implemented based on a softmax function.

It should be appreciated that, in some possible embodiments, the step of mining the key information description vector of the corresponding matrix by using the key information mining unit included in the initial user group analysis network according to the exemplary user distribution matrix may further include the following specific implementation matters:

mining a corresponding initial key information description vector according to the exemplary user distribution matrix by using a key information mining structure in the key information mining unit, that is, the key information mining structure may be used to perform key information mining operation on the exemplary user distribution matrix to form a corresponding initial key information description vector, and the key information mining structure may include one or more filter matrices (convolution kernels);

It should be understood that, in some possible embodiments, the step of determining, according to the initial critical information description vector, the corresponding front-end gradient optimization description vector by using the front-end gradient optimization structure in the critical information mining unit may further include the following specific implementation matters:

determining a corresponding primary filtering description vector by utilizing a primary filtering substructure included in the front-end gradient optimization structure according to the initial key information description vector, wherein the primary filtering substructure can include one or more filtering matrixes (convolution kernels);

determining a corresponding secondary filtering description vector by utilizing a secondary filtering substructure included in the front-end gradient optimization structure according to the initial key information description vector, wherein the secondary filtering substructure can include one or more filtering matrixes (convolution kernels);

Determining a corresponding three-level filtering description vector by utilizing a three-level filtering substructure included in the front-end gradient optimization structure according to the two-level filtering description vector, wherein the three-level filtering substructure can include one or more filtering matrixes (convolution kernels);

determining a corresponding four-level filtering description vector by utilizing a four-level filtering substructure included in the front-end gradient optimization structure according to the three-level filtering description vector, wherein the four-level filtering substructure can include one or more filtering matrixes (convolution kernels);

and determining a corresponding front-end gradient optimization description vector according to the first-stage filtering description vector and the fourth-stage filtering description vector, wherein the first-stage filtering description vector and the fourth-stage filtering description vector can be subjected to superposition operation to output the front-end gradient optimization description vector.

It should be understood that, in some possible embodiments, the step of determining, according to the front-end gradient optimization description vector, the corresponding back-end gradient optimization description vector by using the back-end gradient optimization structure in the key information mining unit may further include the following specific implementation matters:

Determining a corresponding five-level filtering description vector by utilizing a one-level filtering substructure included in the rear-end gradient optimization structure according to the front-end gradient optimization description vector, wherein the one-level filtering substructure can include one or more filtering matrixes (convolution kernels), the one-level filtering substructure can be different from the one-level filtering substructure included in the front-end gradient optimization structure, for example, the number of the included filtering matrixes can be different, and specific filtering parameters can be determined in the network optimization process;

determining a corresponding six-level filtering description vector according to the five-level filtering description vector through a two-level filtering substructure included in the back-end gradient optimization structure, wherein the two-level filtering substructure can include one or more filtering matrixes (convolution kernels), the two-level filtering substructure included in the front-end gradient optimization structure can be different from the two-level filtering substructure included in the front-end gradient optimization structure, for example, the number of the included filtering matrixes can be different, and specific filtering parameters can be determined in the network optimization process;

according to the six-level filtering description vector, determining a corresponding seven-level filtering description vector through a three-level filtering substructure included in the rear-end gradient optimization structure, wherein the three-level filtering substructure can include one or more filtering matrixes (convolution kernels), the three-level filtering substructure included in the front-end gradient optimization structure can be different from the three-level filtering substructure included in the front-end gradient optimization structure, for example, the number of the included filtering matrixes can be different, and specific filtering parameters can be determined in the network optimization process;

And determining a corresponding rear-end gradient optimization description vector according to the front-end gradient optimization description vector and the seven-stage filtering description vector, wherein the front-end gradient optimization description vector and the seven-stage filtering description vector can be subjected to superposition operation to output the rear-end gradient optimization description vector.

It should be understood that, in some possible embodiments, the step of determining, according to the back-end gradient optimization description vector, the matrix key information description vector corresponding to the exemplary user distribution matrix by using at least one front-end gradient optimization structure and at least one back-end gradient optimization structure in the key information mining unit may further include the following specific implementation matters:

and the step of determining the corresponding front-end gradient optimization description vector by using the front-end gradient optimization structure in the key information mining unit after the step of determining the corresponding front-end gradient optimization description vector according to the initial key information description vector by executing the step of using the front-end gradient optimization description vector as a new initial key information description vector in a revolving manner (namely, performing once for the initial key information description vector and performing at least once for at least one new initial key information description vector), and then sequentially performing at least two times (namely, performing once for the initial key information description vector and performing at least one time for the at least one new initial key information description vector) to obtain the last output back-end gradient optimization description vector serving as the key information description vector corresponding to the exemplary user distribution matrix.

It should be appreciated that, in some possible embodiments, the step of constructing an exemplary data cluster based on the exemplary matrix area, the first number of analysis matrix areas, and the first number of user population relationship matching parameters may further include the following specific implementation matters:

analyzing a coincidence relation characterization parameter between each screening matrix region and the exemplary matrix region based on the second number of screening matrix regions, wherein the coincidence relation characterization parameter can be used for reflecting the coincidence degree between the screening matrix region and the exemplary matrix region;

It should be appreciated that, in some possible embodiments, the step of extracting, based on the first number of user population relationship matching parameters, a second number of analysis matrix areas from the first number of analysis matrix areas and marking the second number of analysis matrix areas as screening matrix areas, thereby forming a second number of screening matrix areas may further include the following specific implementation matters:

marking an analysis matrix area corresponding to the user group relation matching parameter as a screening matrix area under the condition that the user group relation matching parameter is not smaller than the reference relation matching parameter, wherein the user group relation matching parameter belongs to the first number of user group relation matching parameters, and the screening matrix area belongs to the second number of screening matrix areas, namely, the second number of screening matrix areas can be correspondingly marked;

It should be appreciated that, in some possible embodiments, the step of determining the corresponding low-coincidence exemplary data cluster based on the coincidence relation characterization parameter between each of the filtering matrix regions and the exemplary matrix region may further include the following specific implementation matters:

marking the exemplary data corresponding to the screening matrix region under the condition that the coincidence relation characterization parameter between the screening matrix region and the exemplary matrix region is not smaller than the low reference coincidence relation characterization parameter, so as to mark the first exemplary matching data in the low coincidence exemplary data cluster, namely, the first exemplary matching data is considered to have the aforementioned exemplary user group relation or match the exemplary user group relation, so that the first exemplary matching data can be used as forward exemplary data;

marking the exemplary data corresponding to the screening matrix region to be the exemplary non-matching data in the low-coincidence exemplary data cluster under the condition that the coincidence relation characterization parameter between the screening matrix region and the exemplary matrix region is smaller than the appointed coincidence relation characterization parameter, wherein the appointed coincidence relation characterization parameter is smaller than the low-reference coincidence relation characterization parameter, namely, the appointed coincidence relation characterization parameter is considered to have no or not match with the aforementioned exemplary user group relation, so that the data can be used as negative exemplary data; illustratively, the low reference coincidence characterization parameter may be equal to 0.7 and the specified coincidence characterization parameter may be equal to 0.4.

It should be appreciated that, in some possible embodiments, the step of determining the corresponding high-coincidence exemplary data cluster based on the coincidence relation characterization parameter between each of the filtering matrix regions and the exemplary matrix region may further include the following specific implementation matters:

marking the exemplary data corresponding to the screening matrix region under the condition that the coincidence relation characterization parameter between the screening matrix region and the exemplary matrix region is not smaller than the high reference coincidence relation characterization parameter, so as to mark the second exemplary matching data in the high coincidence exemplary data cluster, namely, the second exemplary matching data is considered to have the aforementioned exemplary user group relation or match the exemplary user group relation, so that the second exemplary matching data can be used as forward exemplary data;

marking the exemplary data corresponding to the screening matrix region to be the exemplary non-matching data in the high-coincidence exemplary data cluster under the condition that the coincidence relation characterization parameter between the screening matrix region and the exemplary matrix region is smaller than the specified coincidence relation characterization parameter, wherein the specified coincidence relation characterization parameter is smaller than the high-reference coincidence relation characterization parameter, namely, the specified coincidence relation characterization parameter is considered to have no or not match with the aforementioned exemplary user group relationship, so that the data can be used as negative exemplary data; illustratively, the high reference coincidence characterization parameter may be equal to 0.8 and the specified coincidence characterization parameter may be equal to 0.4.

It should be appreciated that, in some possible embodiments, the step of obtaining the corresponding front-end analysis data cluster by using the user group analysis front-end unit included in the initial user group analysis network according to the low-coincidence exemplary data cluster may further include the following specific implementation matters:

determining, by using a front-end region mapping unit, a corresponding low-coincidence mapping feature representation according to the to-be-analyzed exemplary data in the low-coincidence exemplary data cluster, where the to-be-analyzed exemplary data in the low-coincidence exemplary data cluster belongs to exemplary matching data or exemplary non-matching data in the low-coincidence exemplary data cluster, that is, the front-end region mapping unit may perform a key information mining operation on the to-be-analyzed exemplary data to form the corresponding low-coincidence mapping feature representation, that is, perform a key information mining operation only on information of a corresponding screening matrix region;

according to the low coincidence mapping feature representation, front end analysis data corresponding to the to-be-analyzed exemplary data in the low coincidence exemplary data cluster are analyzed by using the user population analysis front end unit, the front end analysis data comprise first relation matching evaluation parameters and first matrix area distribution coordinates corresponding to the to-be-analyzed exemplary data in the low coincidence exemplary data cluster, and the user population analysis front end unit can comprise two full connection structures for respectively carrying out full connection processing on the low coincidence mapping feature representation, and then prediction output can be carried out on two processed results respectively to form the first relation matching evaluation parameters and the first matrix area distribution coordinates.

It should be appreciated that, in some possible embodiments, the step of obtaining the corresponding back-end analysis data cluster by using the user group analysis back-end unit included in the initial user group analysis network according to the high coincidence exemplary data cluster may further include the following specific implementation matters:

determining, by using a back end region mapping unit, a corresponding high coincidence mapping feature representation according to the to-be-analyzed exemplary data in the high coincidence exemplary data cluster, where the to-be-analyzed exemplary data in the high coincidence exemplary data cluster belongs to exemplary matching data or exemplary non-matching data in the high coincidence exemplary data cluster, as described in the foregoing related description;

and analyzing back-end analysis data corresponding to the to-be-analyzed exemplary data in the high-coincidence exemplary data cluster by utilizing the user population analysis back-end unit according to the high-coincidence mapping characteristic representation, wherein the back-end analysis data comprises second relation matching evaluation parameters corresponding to the to-be-analyzed exemplary data in the high-coincidence exemplary data cluster and second to-be-determined matrix region distribution coordinates, and the second relation matching evaluation parameters are as described in the previous related description.

It should be appreciated that, in some possible embodiments, the step of performing a network optimization operation on the initial user group analysis network to form a corresponding target user group analysis network based on the exemplary data cluster, the front end analysis data cluster, and the back end analysis data cluster may further include the following specific implementation matters:

based on the high-coincidence exemplary data cluster and the back-end analysis data cluster, optimizing and adjusting parameters of the user group analysis back-end unit according to the network optimization first rule, which can be described with reference to the related description;

based on the exemplary matrix area and the backend analysis data clusters, according to the network optimization second rule, the parameters of the user group analysis backend unit are subjected to optimization adjustment operation, which can be described in the related description.

It should be understood, that in some possible embodiments, the step of optimizing and adjusting the parameters of the user group analysis front-end unit according to the first rule of network optimization based on the low-coincidence exemplary data cluster and the front-end analysis data cluster may further include the following specific implementation matters:

For each low-coincidence exemplary data in the low-coincidence exemplary data cluster, assigning a first value to the population relation representation value of the low-coincidence exemplary data if the low-coincidence exemplary data belongs to the exemplary matching data, and assigning a second value to the population relation representation value of the low-coincidence exemplary data if the low-coincidence exemplary data belongs to the exemplary non-matching data, e.g., the first value is 1 and the second value is 0;

calculating absolute difference values of first relation matching evaluation parameters and corresponding group relation characterization values included in front-end analysis data corresponding to each piece of low-coincidence exemplary data respectively so as to output the absolute difference values corresponding to each piece of low-coincidence exemplary data;

calculating a sum value of absolute differences corresponding to each piece of low-coincidence exemplary data, and determining a first network learning cost parameter based on the calculated sum value, wherein the first network learning cost parameter can have a positive correlation with the sum value;

and carrying out optimization adjustment operation on the parameters of the user group analysis front-end unit based on the first network learning cost parameters, for example, optimizing and adjusting the parameters so that the first network learning cost parameters are smaller than or equal to the pre-configured reference first network learning cost parameters.

Wherein, it should be understood that, in some possible embodiments, the step of performing an optimization adjustment operation on the parameters of the user group analysis front-end unit according to the second rule of network optimization based on the exemplary matrix area and the front-end analysis data cluster may further include the following specific implementation matters:

for each low-coincidence exemplary data in the low-coincidence exemplary data cluster, calculating a coordinate distance between a distribution coordinate of a first to-be-determined matrix region included in the low-coincidence exemplary data and a distribution coordinate of the exemplary matrix region to obtain a coordinate distance of the low-coincidence exemplary data;

performing sum value calculation on the coordinate distance of each low-coincidence exemplary data belonging to the exemplary matching data to obtain a first sum value, and performing sum value calculation on the coordinate distance of each low-coincidence exemplary data belonging to the exemplary non-matching data to obtain a second sum value;

and carrying out optimization adjustment operation on the parameters of the user group analysis front-end unit along the direction of reducing the first sum value and along the direction of increasing the second sum value.

With reference to fig. 3, the embodiment of the invention further provides an information recommendation device based on data mining and AI model analysis, which can be applied to the information recommendation system based on data mining and AI model analysis. Wherein, the information recommending device based on data mining and AI model analysis may include:

the distribution matrix determining module is used for determining a candidate user distribution matrix, wherein each matrix parameter in the candidate user distribution matrix corresponds to one candidate user, and in the candidate user distribution matrix, the distribution relation among matrix parameters is related to the distribution relation among the candidate users, and the matrix parameters are used for reflecting the user characteristic information of the candidate users;

the key information mining module is used for mining matrix key information description vectors corresponding to the candidate user distribution matrix by utilizing a key information mining unit included in a target user group analysis network formed by network optimization operation;

the first user group analysis module is used for obtaining a corresponding analysis matrix area and user group relation matching parameters by utilizing an analysis output unit included in the target user group analysis network according to the matrix key information description vector, wherein the user group relation matching parameters are in one-to-one correspondence with the analysis matrix area;

The second user group analysis module is used for marking an analysis matrix area to be a screening matrix area under the condition that the user group relation matching parameter is not smaller than the reference relation matching parameter, analyzing a first to-be-determined matrix area distribution coordinate and a first relation matching evaluation parameter by utilizing a user group analysis front-end unit included in the target user group analysis network according to the screening matrix area, and analyzing a second to-be-determined matrix area distribution coordinate and a second relation matching evaluation parameter by utilizing a user group analysis rear-end unit included in the target user group analysis network according to the screening matrix area;

the third user group analysis module is used for analyzing the target matrix area of the candidate user distribution matrix and the user group relationship matching parameters of the candidate user distribution matrix based on the screening matrix area, the first to-be-determined matrix area distribution coordinates, the first relationship matching evaluation parameters, the second to-be-determined matrix area distribution coordinates and the second relationship matching evaluation parameters;

and the association information recommending module is used for taking the candidate users corresponding to each matrix parameter included in the target matrix area as group association users under the condition that the user group relation matching parameters of the candidate user distribution matrix meet the preset matching rule, and carrying out association information recommending operation among the group association users.

In summary, the information recommendation method and system based on data mining and AI model analysis provided by the invention can firstly mine the matrix key information description vector corresponding to the candidate user distribution matrix; obtaining a matching parameter of the analysis matrix area and the user group relation according to the matrix key information description vector; marking a screening matrix area based on the user population relation matching parameters, and analyzing a first to-be-determined matrix area distribution coordinate and a first relation matching evaluation parameter, a second to-be-determined matrix area distribution coordinate and a second relation matching evaluation parameter according to the screening matrix area; and performing association information recommendation operation among the group association users based on the screening matrix area, the first to-be-determined matrix area distribution coordinates, the first relationship matching evaluation parameters, the second to-be-determined matrix area distribution coordinates and the second relationship matching evaluation parameters. Based on the foregoing, before the information recommendation operation is performed, the group associated users are determined, so that the associated information recommendation operation can be performed, and the efficiency of information recommendation can be improved to a certain extent.

The above description is only of the preferred embodiments of the present invention and is not intended to limit the present invention, but various modifications and variations can be made to the present invention by those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. An information recommendation method based on data mining and AI model analysis is characterized by comprising the following steps:

2. The data mining and AI model analysis-based information recommendation method of claim 1, wherein the network optimization process of the target user population analysis network includes:

and performing network optimization operation on the initial user group analysis network based on the exemplary data cluster, the front-end analysis data cluster and the back-end analysis data cluster to form a corresponding target user group analysis network, wherein the low-coincidence exemplary data cluster and the front-end analysis data cluster are used for performing network optimization operation on the user group analysis front-end unit, the high-coincidence exemplary data cluster and the back-end analysis data cluster are used for performing network optimization operation on the user group analysis back-end unit, and the user group analysis front-end unit and the user group analysis back-end unit are different.

3. The data mining and AI model analysis-based information recommendation method of claim 2, wherein analyzing the first number of analysis matrix regions and the first number of user population relationship matching parameters in the exemplary user distribution matrix using an initial user population analysis network includes:

4. The data mining and AI model analysis-based information recommendation method of claim 3, wherein the step of mining out a corresponding matrix key information description vector by using a key information mining unit included in the initial user group analysis network according to the exemplary user distribution matrix includes:

5. The data mining and AI model analysis-based information recommendation method of claim 2, wherein the step of constructing an exemplary data cluster based on the exemplary matrix area, the first number of analysis matrix areas, and the first number of user population relationship matching parameters includes:

6. The data mining and AI model analysis-based information recommendation method of claim 5, wherein the step of decimating a second number of analysis matrix areas from the first number of analysis matrix areas based on the first number of user population relationship matching parameters and marking the second number of analysis matrix areas as screening matrix areas, thereby forming a second number of screening matrix areas, includes:

7. The data mining and AI model analysis-based information recommendation method of claim 5, wherein determining a corresponding low-coincidence exemplary data cluster based on coincidence relation characterization parameters between each of the screening matrix regions and the exemplary matrix region includes:

8. The method for recommending information based on data mining and AI model analysis of claim 2, wherein the step of obtaining a corresponding front-end analysis data cluster from the low-coincidence exemplary data cluster by using a user group analysis front-end unit included in the initial user group analysis network includes:

9. The data mining and AI model analysis-based information recommendation method of any of claims 2-8, wherein the step of performing network optimization operations on the initial user population analysis network to form a corresponding target user population analysis network based on the exemplary data cluster, the front-end analysis data cluster, and the back-end analysis data cluster includes:

10. An information recommendation system based on data mining and AI model analysis, characterized by comprising a processor and a memory, said memory for storing a computer program, said processor for executing said computer program for implementing the method of any of claims 1-9.