Education empowerment method and system based on all-media traffic aggregation

Abstract

The present invention provides an education empowerment method and system based on full-media traffic aggregation, which relates to the field of data processing. The system includes: a content delivery module, which is used to deliver interactive content on multiple media traffic platforms; a data acquisition module, which is used to obtain user behavior data on interactive content; a user screening module, which is used to determine candidate users based on user behavior data on interactive content; a user determination module, which is used to determine target users based on the target historical behavior data of candidate users through a first artificial intelligence model; a portrait establishment module, which is used to establish user portraits based on the target historical behavior data of target users through a second artificial intelligence model; and an institution matching module, which is used to match target education and training institutions based on user portraits. The system has the advantages of realizing traffic aggregation of multiple media traffic platforms, improving the efficiency of mining user needs, and more accurately and comprehensively analyzing the matching degree between users and education and training institutions, thereby empowering education and training.

Description

Technical Field

[0001] The present invention relates to the field of data processing, and in particular to an education enabling method and system based on omnimedia traffic aggregation. Background Art

[0002] Educational training institutions include educational information for all levels, from preschool education to university, and even doctoral or studying abroad. They also provide skill training for current workers or laid-off workers. They are specialized training institutions whose main content is to provide educational resources and training information.

[0003] Currently, user needs are primarily discovered offline, which is inefficient. Furthermore, once a user need is identified, the system manually matches the user with education and training institutions offering the corresponding services, and then pushes the user's information to one or all matching institutions. The matching methods are simple, either sorted by time or by a certain commission ratio. Operators, based on their own judgment, assign users to any one or all matching education and training institutions, potentially leading to users terminating their training due to inappropriate recommendations.

[0004] Therefore, it is necessary to provide an education empowerment method and system based on all-media traffic aggregation to realize traffic aggregation of multiple media traffic platforms, improve the efficiency of mining user needs, and more accurately and comprehensively analyze the matching degree between users and education and training institutions to empower education and training. Summary of the Invention

[0005] The present invention provides an education empowerment system based on all-media traffic aggregation, including: a content delivery module for delivering interactive content on multiple media traffic platforms; a data acquisition module for acquiring user behavior data on the interactive content; a user screening module for determining candidate users based on the user behavior data on the interactive content; the data acquisition module is also used to acquire the target historical behavior data of the candidate users from the multiple media traffic platforms; a user determination module for determining the target user based on the target historical behavior data of the candidate users acquired from the multiple media traffic platforms through a first artificial intelligence model; a portrait establishment module for establishing a user portrait based on the target historical behavior data of the target user through a second artificial intelligence model; and an institution matching module for matching the target education and training institution based on the user portrait.

[0006] Furthermore, the content delivery module delivers interactive content on multiple media traffic platforms, including: for each of the media traffic platforms, performing user sampling on the media traffic platform to obtain multiple sampled users, and establishing a portrait of the media traffic platform based on the basic data and historical behavior data of the multiple sampled users; for each of the education and training institutions, establishing a portrait of the training objects corresponding to each education and training project provided by the education and training institution; for each of the media traffic platforms, based on the portrait of the training objects corresponding to each education and training project and the portrait of the media traffic platform, determining the education and training projects matched by the media traffic platform, generating interactive content matched by the media traffic platform based on the education and training projects matched by the media traffic platform, and delivering the interactive content on the media traffic platform.

[0007] Furthermore, the user screening module determines candidate users based on the user's behavioral data on the interactive content, including: determining the user's interactive activity parameter based on the user's behavioral data on the interactive content; and determining the candidate users based on the user's interactive activity parameter.

[0008] Furthermore, the user screening module determines the interactive activity parameters of the user based on the user's behavioral data on the interactive content, including: obtaining the user's comment information on the interactive content based on the user's behavioral data on the interactive content; identifying the relevance of the comment information; for relevant comment information, identifying the emotional parameters of the relevant comment information; and determining the user's interactive activity parameters based on the emotional parameters of the relevant comment information.

[0009] Furthermore, the data acquisition module obtains the target historical behavior data of the candidate user from the multiple media traffic platforms, including: obtaining the historical behavior data of the candidate user; extracting key features of the historical behavior data; calculating the association parameters between the historical behavior data and the interactive content based on the key features of the historical behavior data; and judging whether the historical behavior data is the target historical behavior data based on the association parameters between the historical behavior data and the interactive content.

[0010] Furthermore, the user determination module determines the target user through the first artificial intelligence model based on the target historical behavior data of the candidate user obtained from the multiple media traffic platforms, including: determining the matching degree between the candidate user and the training needs corresponding to the interactive content based on the target historical behavior data of the candidate user through the first artificial intelligence model; determining the target user based on the matching degree between the candidate user and the training needs corresponding to the interactive content.

[0011] Furthermore, the portrait establishment module establishes a user portrait based on the basic data and historical behavior data of the target user through a second artificial intelligence model, including: determining multiple known user portrait features of the target user based on the basic data and historical behavior data of the target user; predicting missing user portrait features based on the historical behavior data of the target user through the second artificial intelligence model; and establishing the user portrait based on the multiple known user portrait features and the predicted missing user portrait features.

[0012] Furthermore, the portrait establishment module predicts missing user portrait features based on the historical behavior data of the target user through the second artificial intelligence model, including: determining similar target users based on the historical behavior data of the target user; and predicting missing user portrait features based on the basic data and historical behavior data of the similar target users through the second artificial intelligence model.

[0013] Furthermore, the institution matching module matches the target education and training institution based on the user portrait, including: for each of the education and training institutions, calculating the demand matching value of the education and training institution based on the user portrait and the training object portrait corresponding to each education and training project provided by the education and training institution, calculating the location matching value of the education and training institution based on the user portrait and the address distribution information of the education and training institution, and calculating the matching priority value of the education and training institution based on the historical user evaluation information and achievement transformation information of the education and training institution; determining the target education and training institution based on the demand matching value, location matching value and matching priority value of each of the education and training institutions.

[0014] The present invention provides an education empowerment method based on all-media traffic aggregation, including: delivering interactive content on multiple media traffic platforms; obtaining user behavior data on the interactive content; determining candidate users based on the user behavior data on the interactive content; obtaining target historical behavior data of the candidate users from the multiple media traffic platforms; determining target users based on the target historical behavior data of the candidate users obtained from the multiple media traffic platforms; establishing a user portrait based on the basic data and target historical behavior data of the target user through an artificial intelligence model; and matching target education and training institutions based on the user portrait.

[0015] Compared with the existing technology, the education empowerment method and system based on all-media traffic aggregation provided by the present invention have at least the following beneficial effects:

[0016] 1. By collecting and analyzing user behavioral data on interactive content, we can precisely identify user groups with potential interest or demand for education and training, known as candidate users. Further acquiring target historical behavioral data on candidate users provides a deeper understanding of their interests, learning needs, and consumption habits. Target users identified based on this data are more likely to be interested in purchasing the education and training institution's products or services, thereby improving conversion rates and user satisfaction. User profiles built using AI models comprehensively reflect the multi-dimensional characteristics of target users. This enables education and training institutions to provide personalized course recommendations and learning plans based on their user profiles, enhancing the user experience and increasing user retention. Data analysis and AI models make the decision-making process of education and training institutions more scientific and objective. Data-based decision-making reduces subjective assumptions and blindness, improving the effectiveness and accuracy of decisions. With the continuous accumulation of data and continuous optimization of models, education and training institutions can continuously improve their marketing strategies, products and services, and user experience to adapt to market changes and evolving user needs, maintaining a competitive advantage.

[0017] 2. Evaluating user engagement based solely on simple behavioral data like clicks and browsing may not be comprehensive enough. By incorporating the relevance and sentiment of comment information, we can more comprehensively assess the quality of user engagement. Highly relevant comments indicate that users have deeply considered and engaged with the content, while sentiment reveals positive or negative attitudes. These are important indicators for assessing user engagement. Users with high engagement levels are often more likely to become potential users for educational training institutions, enabling precise screening of candidate users.

[0018] 3. By comprehensively considering the match between the user profile and the training target profile, we ensure that recommended training institutions precisely meet the user's personalized needs and learning goals. This refined matching mechanism helps improve user satisfaction and learning outcomes. The calculation of location matching values ​​allows the selection of training institutions to take into account not only their teaching content and quality, but also their geographical convenience. This helps users choose training institutions that are closer to them or have more convenient transportation, reducing commuting time and costs. The inclusion of historical user evaluation information and results conversion data provides an objective and comprehensive reference for selecting training institutions. These indicators reflect the actual teaching quality, student satisfaction, and results conversion rate of training institutions, helping users select more reliable and efficient training institutions. By comprehensively considering matching values ​​across multiple dimensions to determine target training institutions, we provide users with a more personalized and attentive service experience. Users can not only find learning content that suits them, but also receive high-quality teaching services in a convenient location, thereby improving their overall learning experience and satisfaction. Through an open and transparent matching mechanism, training institutions need to continuously improve their teaching quality, enhance service facilities, and optimize their geographical layout to attract more users and increase their matching priority. This competitive mechanism helps promote the healthy development and improvement of the entire education and training industry. Determining target education and training institutions based on multi-dimensional matching values ​​helps optimize the allocation of educational resources. Users can choose the most appropriate education and training institution based on their needs and conditions, while education and training institutions can attract a more compatible user base based on their own strengths and characteristics, achieving a win-win situation for all parties. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] This specification will be further described in the form of exemplary embodiments, which will be described in detail with reference to the accompanying drawings. These embodiments are not limiting, and in these embodiments, like numbers represent like structures, wherein:

[0020] Figure 1 This is a module diagram of an education empowerment system based on all-media traffic aggregation according to some embodiments of this specification;

[0021] Figure 2 It is a flow chart of an education empowerment method based on all-media traffic aggregation according to some embodiments of this specification. DETAILED DESCRIPTION

[0022] To more clearly illustrate the technical solutions of the embodiments of this specification, the following briefly describes the drawings required for describing the embodiments. Obviously, the drawings described below are merely examples or embodiments of this specification. Those skilled in the art can apply this specification to other similar scenarios based on these drawings without inventive effort. Unless otherwise apparent from the context or otherwise noted, the same reference numerals in the figures represent the same structure or operation.

[0023] Figure 1 This is a module diagram of an education empowerment system based on all-media traffic aggregation according to some embodiments of this specification, such as Figure 1 As shown, the education empowerment system based on all-media traffic aggregation can include a content delivery module, a data acquisition module, a user screening module, a user determination module, a portrait creation module and an institution matching module.

[0024] The content delivery module can be used to deliver interactive content on multiple media traffic platforms.

[0025] The media traffic platform may be a platform with user social functions, such as Tik Tok, Xiaohongshu, Kuaishou, etc.

[0026] Specifically include:

[0027] For each media traffic platform, user sampling is performed on the media traffic platform to obtain multiple sampled users. Based on the basic data and historical behavior data of the multiple sampled users, a media traffic platform profile is established;

[0028] For each education and training institution, establish a training target profile corresponding to each education and training program provided by the education and training institution;

[0029] For each media traffic platform, based on the training target portrait and media traffic platform portrait corresponding to each education and training project, determine the education and training projects that match the media traffic platform. Based on the education and training projects that match the media traffic platform, generate interactive content that matches the media traffic platform, and deliver the interactive content on the media traffic platform.

[0030] Specifically, the basic data of the sampled users may be personal data disclosed on the media traffic platform with the user's consent or authorization, such as age, gender, region, occupation, etc. The historical behavior data of the sampled users may be behavior data disclosed on the media traffic platform with the user's consent or authorization. For example, browsing history, liked content, favorited content, commented content, published content, etc. The content delivery module can randomly select a certain number of users from all users of the media traffic platform as sampled users. This method can reduce sampling errors and make the results more representative. The basic data and historical behavior data of the sampled users are obtained from the media traffic platform through the API (Application Programming Interface) interface, data export function or crawler of the media traffic platform.

[0031] The content delivery module can collect statistics and analyze basic attributes of sampled users such as age, gender, region, occupation, etc., and determine the age distribution, gender ratio, regional distribution, occupation distribution and marital and childbearing status of the media traffic platform. Among them, the age distribution can represent the proportion of users in each age group, the regional distribution can represent the proportion of users in each region, the occupational distribution can represent the proportion of users in each occupation, and the marital and childbearing status can represent the proportion of users who are unmarried and not pregnant, married and not childbearing, and married and have children.

[0032] The content delivery module can analyze the interests and preferences of the sampled users based on their historical behavior data. For example, the content delivery module can analyze the interests and preferences of the sampled users based on their historical behavior data using a data analysis model, where the data analysis model can be a convolutional neural network (CNN) model.

[0033] For each media traffic platform, the content delivery module can perform statistical analysis on the interest preferences of sampled users to determine the distribution of interest points corresponding to the media traffic platform, wherein the distribution of interest points can represent the proportion of users interested in each interest point.

[0034] For each media traffic platform, the content delivery module can determine the content form distribution corresponding to the media traffic platform based on the historical behavior data of sampled users, where the content form distribution can represent the proportion of various content forms (for example, text, image, video, etc.).

[0035] The media traffic platform portrait may include information such as age distribution, gender ratio, regional distribution, occupational distribution, marital status, interest point distribution and content form distribution.

[0036] For each education and training program provided by an education and training institution, the content delivery module can determine the points of interest corresponding to the education and training program (e.g., arts and crafts, music and dance, technology and programming, sports and fitness, language learning, baking and cooking, parenting education, etc.). For example, the points of interest determination model can be used to determine the points of interest corresponding to the education and training program based on the program information of the education and training program (e.g., course information, teaching materials information, etc.). The point of interest determination model can be a convolutional neural network model.

[0037] The training target profile corresponding to the education and training program includes information on points of interest, and may also include information on age group, occupation, etc. The age group information, occupation information, etc. may be determined based on the basic data of users who are currently in or have completed the education and training program.

[0038] For each media traffic platform, the content delivery module can use a project matching model based on the training target profile and media traffic platform profile corresponding to each education and training project to determine the education and training projects that match the media traffic platform. The project matching model can be a convolutional neural network model. Based on the distribution of content forms corresponding to the media traffic platform, the optimal content form is determined. Based on the optimal content form and the project information of the education and training project, interactive content that matches the media traffic platform is generated. The optimal content form can be the content form with the highest proportion.

[0039] The data acquisition module can be used to obtain user behavior data on interactive content.

[0040] Specifically, the data acquisition module can obtain user behavior data on interactive content with the user's consent or authorization through the media traffic platform's API interface, data export function, or crawler, such as the number of views, likes, favorites, comments, and reposts.

[0041] The user screening module can be used to determine candidate users based on the user's behavioral data on interactive content.

[0042] Specifically include:

[0043] Determine the user's interactive activity parameters based on the user's behavioral data on interactive content;

[0044] Determine candidate users based on the user's interactive activity parameters.

[0045] Specifically, the first activity parameter of the user for the interactive content may be calculated based on the number of types of interactive behaviors of the user for the interactive content:

[0046]

[0047] Among them, P (1,i)is the first activity parameter of the i-th user on the interactive content, N i N is the number of types of interactive behaviors of the i-th user on the interactive content, i The total number of preset interaction behavior types.

[0048] For example, the types of interactive behaviors include browsing, liking, collecting, commenting, and forwarding. The user's interactive behaviors on interactive content include browsing and liking. Then the user's first activity parameter for the interactive content is 2 / 5.

[0049] In some embodiments, the user screening module determines the user's interactive activity parameters based on the user's behavioral data on the interactive content, including:

[0050] Based on the user's behavioral data on the interactive content, obtain the user's comment information on the interactive content;

[0051] Identify the relevance of review information;

[0052] For relevant comment information, identifying sentiment parameters of the relevant comment information;

[0053] Based on the sentiment parameters of relevant comment information, the user's interactive activity parameters are determined.

[0054] Specifically, the user screening module can perform text cleaning on the comment information, remove special characters, invalid spaces and other noise data in the comment information, and use the Chinese word segmentation tool to perform word segmentation on the comment information after noise removal, and remove stop words to reduce redundant information, and obtain multiple valid comment words in the comment information.

[0055] For each interactive content, the user screening module may determine multiple comment keywords corresponding to the interactive content through a keyword determination model, wherein the keyword determination model may be a BERT (Bidirectional Encoder Representations from Transformers) model.

[0056] For each valid review word, the semantic distance between the valid review word and each review keyword is calculated to determine the shortest semantic distance corresponding to the valid review word.

[0057] Based on the shortest semantic distance corresponding to each valid comment word, the relevance parameter of the comment information is determined:

[0058]

[0059] Among them, r (i,j) is the relevance parameter of the jth comment information of the i-th user, a 11 、a 12and a 13 All are preset weights, and a 11 +a 12 +a 13 =1,a 11 >0,a 12 >0,a 13 >0, N j is the number of valid comment words in the jth comment information, N0 is the preset number of valid comment words, P (i,j) is the ratio of valid comment words in the jth comment information, P0 is the preset ratio of valid comment words, μ j is the shortest semantic distance mean of the effective comment words of the j-th comment information, μ0 is the preset shortest semantic distance mean, σ0 is the preset shortest semantic distance variance, σ i is the shortest semantic distance variance of the effective comment words of the j-th comment information, L (j,effective) is the sum of the lengths of the valid comment words in the jth comment information, L j is the length of the jth comment in characters, D (j,n) is the shortest semantic distance corresponding to the nth valid comment word of the jth comment information, and N is the total number of valid comment words included in the jth comment information.

[0060] When the relevance parameter of the comment information is greater than a preset relevance parameter threshold, the comment information is relevant.

[0061] The emotion recognition model can be used to identify the emotional characteristics of the relevant review information based on multiple valid comment words in the relevant review information. The emotional characteristics may include the matching probability corresponding to each emotion (e.g., happy, sad, angry, calm, etc.). The emotion recognition model can be a long short-term memory network model.

[0062] The state update of the long short-term memory network model can be described by the following mathematical formula:

[0063] f t =σ(W f ·[h t―1 ,x t ]+b f )

[0064] i t =σ(W i ·[h t―1 ,x t ]+b i )

[0065] C t =tanh(W f ·[h t―1 ,x t ]+bC )

[0066] C t =f t *C t―1 +i t *C t )

[0067] o t =σ(W o ·[h t―1 ,x t ]+b o )

[0068] h t =o t *tanh(C t )

[0069] Among them, f t 、i t and o t Represent the activation values ​​of the forget gate, input gate and output gate respectively, h t―1 is the hidden state of the previous time step, x t is the input of the current time step, C t is the internal state of the current unit, C t is a candidate state, W f 、W i 、W o and b f 、b i 、b c 、b o is the network parameter, σ is the sigmoid activation function, * represents the product between elements, and tanh() is the hyperbolic tangent function.

[0070] During the training of the LSTM network model, the network parameters of the LSTM network model can be adjusted through the back propagation algorithm:

[0071]

[0072] m t =β1m t―1 +(1―β1)g t

[0073] v t =β2v t―1 +(1―β2)g t 2

[0074] Among them, θ t+1 is the parameter of the long short-term memory network model at time step t+1, θ tis the parameter of the long short-term memory network model at time step t, η is the learning rate, m t is the moving average of the gradient at time step t, v t is the moving average of the square of the gradient at time step t, ε is a constant, β1 is the decay rate, m t―1 is the moving average of the gradient at time step t-1, g t is the gradient, β2 is the decay rate, v t―1 is the moving average of the squared gradient at time step t-1.

[0075] The sentiment parameters of relevant review information can be calculated based on the sentiment characteristics of relevant review information according to the following formula:

[0076]

[0077] Among them, E j is the sentiment parameter of the jth relevant comment information, a 21 and a 22 All are preset weights, and a 21 +a 22 =1,a 21 >0,a 22 >0, V m is the emotion value corresponding to the mth emotion, for example, happiness is 3, disappointment is 1, P (j,m) is the matching probability corresponding to the mth emotion in the emotional features of the jth relevant comment information, and M is the total number of emotion types.

[0078] The second activity parameter of the user's interactive content can be calculated based on the relevance parameter and sentiment parameter of the relevant comment information according to the following formula:

[0079] P (2,i) =max(a 31 ×r j +a 32 ×E j ),j=1,2…J

[0080] Among them, P (2,i) is the second activity parameter of the i-th user for interactive content, a 31 and a 32 All are preset weights, and a 31 +a 32 =1,a 31 >0,a 32 >0, J is the total number of relevant comment information of the i-th user on the interactive content, and max() is the maximum value operation.

[0081] In some embodiments, the user screening module determines the user's interactive activity parameter based on the user's behavioral data on the interactive content, and may also include:

[0082] Based on the user's behavioral data on the interactive content, obtain the user's forwarding information on the interactive content;

[0083] Identify the relevance parameters of users' forwarding information on interactive content;

[0084] Identify the sentiment parameters of users’ forwarded information about interactive content;

[0085] The user's third activity parameter is determined based on the relevance parameter and the emotion parameter of the user's forwarding information of the interactive content.

[0086] Specifically, the user screening module can perform text cleaning on the user's forwarded information on interactive content, remove noise data such as special characters and invalid spaces in the forwarded information, and use Chinese word segmentation tools to perform word segmentation on the forwarded information after noise removal, and remove stop words to reduce redundant information, thereby obtaining multiple valid comment words for the forwarded information.

[0087] Based on multiple valid comment words of the forwarded information, the relevance parameter and sentiment parameter of the user's forwarded information on the interactive content are calculated.

[0088] The method of calculating the relevance parameters and sentiment parameters of the user's forwarding information on the interactive content is similar to the method of calculating the relevance parameters and sentiment parameters of the relevant comment information, and will not be repeated here.

[0089] The user's interactive activity parameter is calculated based on the following formula:

[0090] P i =a 41 ×P (1,i) +a 42 ×P (2,i) +a 43 ×P (3,i)

[0091] Among them, P i is the interactive activity parameter of the i-th user, a 41 、a 42 and a 43 All are preset weights, and a 41 +a 42 +a 43 =1,a 41 >0,a 42 >0,a 43 >0, P (3,i) is the third activity parameter of the i-th user for the interactive content.

[0092] The data acquisition module can also be used to obtain target historical behavior data of candidate users from multiple media traffic platforms.

[0093] Specifically include:

[0094] Obtain historical behavior data of candidate users, such as content they liked, collected, posted, and forwarded;

[0095] Extracting key features of the historical behavior data, where the key features of the historical behavior data may include words in the content corresponding to the historical behavior. The content corresponding to the historical behavior may be cleaned to remove noise data such as special characters and invalid spaces in the content corresponding to the historical behavior. A Chinese word segmentation tool is then used to segment the content corresponding to the historical behavior after noise removal, and stop words are removed to reduce redundant information, thereby obtaining multiple valid words in the content corresponding to the historical behavior.

[0096] Based on the key features of historical behavior data, calculate the correlation parameters between historical behavior data and interactive content;

[0097] Based on the correlation parameters between the historical behavior data and the interactive content, it is determined whether the historical behavior data is the target historical behavior data.

[0098] Specifically, for each valid word in the content corresponding to the historical behavior, the semantic distance between the valid word and each valid word in the interactive content is calculated to determine the shortest semantic distance corresponding to the valid word. The association parameter between the historical behavior data and the interactive content is calculated based on the shortest semantic distance of each valid word in the content corresponding to the historical behavior according to the following formula:

[0099]

[0100] Among them, r (i,e) is the correlation parameter between the e-th historical behavior data and interactive content of the i-th candidate user, D ((i,e),f) is the shortest semantic distance corresponding to the fth valid word in the content corresponding to the eth historical behavior of the i-th candidate user, and F is the total number of valid words in the content corresponding to the eth historical behavior of the i-th candidate user.

[0101] When the correlation parameter between the historical behavior data and the interactive content is greater than a preset correlation parameter threshold, the historical behavior data is determined to be target historical behavior data.

[0102] The user determination module can be used to determine the target user based on the target historical behavior data of candidate users obtained from multiple media traffic platforms through a first artificial intelligence model.

[0103] Specifically include:

[0104] Determining, by a first artificial intelligence model based on the candidate user's target historical behavior data, a degree of matching between the candidate user and the training needs corresponding to the interactive content, wherein the first artificial intelligence model may be a Transformer model;

[0105] The target user is determined based on the matching degree between the candidate user and the training demand corresponding to the interactive content. For example, the candidate user whose matching degree is greater than a preset matching degree threshold is used as the target user.

[0106] The portrait creation module can be used to create a user portrait based on the target user's basic data and target historical behavior data through a second artificial intelligence model.

[0107] Specifically include:

[0108] Based on the target user's basic data and historical behavior data, determine multiple known user profile features of the target user, such as age, occupation, living area, etc.

[0109] Predicting missing user profile features (e.g., occupation, etc.) based on the target user's historical behavior data using a second artificial intelligence model, where the second artificial intelligence model may be a convolutional neural network model;

[0110] Build a user profile based on multiple known user profile features and predicted missing user profile features.

[0111] In some embodiments, the profile building module predicts missing user profile features based on the target user's historical behavior data using a second artificial intelligence model, including:

[0112] Identify similar target users based on their historical behavior data;

[0113] The second artificial intelligence model is used to predict missing user portrait features based on the basic data and historical behavior data of similar target users.

[0114] Specifically, a similarity determination model can be used to determine the behavioral similarity between two target users based on their historical behavioral data. When the behavioral similarity between the two target users is greater than a preset behavioral similarity threshold, the two target users can be considered similar target users. The similarity determination model can be a convolutional neural network model.

[0115] The institution matching module can be used to match target education and training institutions based on user portraits.

[0116] Specifically include:

[0117] For each education and training institution, the demand matching value of the education and training institution is calculated based on the user portrait and the training target portrait corresponding to each education and training project provided by the education and training institution. The location matching value of the education and training institution is calculated based on the user portrait and the address distribution information of the education and training institution. The matching priority value of the education and training institution is calculated based on the historical user evaluation information and achievement transformation information of the education and training institution.

[0118] Determine the target education and training institutions based on the demand matching value, location matching value and matching priority value of each education and training institution.

[0119] Specifically, a demand matching prediction model can be used to calculate the demand matching value of an education and training institution based on user profiles and the training target profiles corresponding to each education and training program provided by the education and training institution. The demand matching prediction model can be a convolutional neural network model.

[0120] The information on the transformation of results can represent the growth of users' relevant skills after training by educational and training institutions.

[0121] The institution matching module can perform sentiment analysis on historical user evaluation information, and calculate the matching priority value of education and training institutions based on the sentiment analysis results and achievement transformation information of historical user evaluation information.

[0122] The demand matching value, location matching value and matching priority value of the education and training institution can be weighted and summed to determine the matching value of the education and training institution, and the education and training institution with the largest matching value can be used as the target education and training institution.

[0123] Figure 2 This is a flow chart of an education empowerment method based on all-media traffic aggregation according to some embodiments of this specification, such as Figure 2 As shown, the education empowerment method based on all-media traffic aggregation can include the following steps.

[0124] Step 210: delivering interactive content on multiple media traffic platforms;

[0125] Step 220, obtaining user behavior data on interactive content;

[0126] Step 230, determining candidate users based on the user's behavioral data on the interactive content;

[0127] Step 240 , obtaining target historical behavior data of candidate users from multiple media traffic platforms;

[0128] Step 250 , determining a target user based on target historical behavior data of candidate users obtained from multiple media traffic platforms;

[0129] Step 260: Create a user profile based on the target user's basic data and historical behavior data using an artificial intelligence model.

[0130] Step 270: Match target education and training institutions based on user profiles.

[0131] The education empowerment method based on all-media traffic aggregation can be executed by the education empowerment system based on all-media traffic aggregation. For more descriptions of the education empowerment method based on all-media traffic aggregation, please refer to the relevant descriptions of the education empowerment system based on all-media traffic aggregation, which will not be repeated here.

[0132] Finally, it should be understood that the embodiments described in this specification are intended only to illustrate the principles of the embodiments of this specification. Other variations may also fall within the scope of this specification. Therefore, by way of example and not limitation, alternative configurations of the embodiments of this specification may be considered consistent with the teachings of this specification. Accordingly, the embodiments of this specification are not limited to the embodiments explicitly described and illustrated in this specification.

Claims

1. The education empowerment system based on all-media traffic aggregation is characterized by: include: Content delivery module, used to deliver interactive content on multiple media traffic platforms; A data acquisition module, used to acquire user behavior data on the interactive content; A user screening module, configured to determine candidate users based on user behavior data regarding the interactive content; The data acquisition module is further configured to acquire target historical behavior data of the candidate user from the multiple media traffic platforms; A user determination module, configured to determine a target user based on the target historical behavior data of the candidate users obtained from the multiple media traffic platforms through a first artificial intelligence model; A portrait building module, configured to build a user portrait based on the basic data and historical behavior data of the target user through a second artificial intelligence model; The institution matching module is used to match the target education and training institution based on the user portrait.

2. The education empowerment system based on all-media traffic aggregation according to claim 1 is characterized in that: The content delivery module delivers interactive content on multiple media traffic platforms, including: For each of the media traffic platforms, user sampling is performed on the media traffic platform to obtain a plurality of sampled users, and a profile of the media traffic platform is established based on basic data and historical behavior data of the plurality of sampled users; For each of the education and training institutions, a training target profile corresponding to each education and training program provided by the education and training institution is established; For each of the media traffic platforms, based on the training object portrait corresponding to each education and training project and the media traffic platform portrait, the education and training projects matched by the media traffic platform are determined; based on the education and training projects matched by the media traffic platform, interactive content matched by the media traffic platform is generated, and the interactive content is delivered to the media traffic platform.

3. The education empowerment system based on all-media traffic aggregation according to claim 1 is characterized in that: The user screening module determines candidate users based on the user's behavior data on the interactive content, including: Determining an interactive activity parameter of the user based on the user's behavioral data on the interactive content; The candidate user is determined based on the interactive activity parameter of the user.

4. The education empowerment system based on all-media traffic aggregation according to claim 3 is characterized in that: The user screening module determines the user's interactive activity parameter based on the user's behavioral data on the interactive content, including: Based on the user's behavior data on the interactive content, obtaining the user's comment information on the interactive content; Identifying the relevance of the review information; For relevant comment information, identifying sentiment parameters of the relevant comment information; Based on the sentiment parameters of the relevant comment information, the interactive activity parameter of the user is determined.

5. The education empowerment system based on all-media traffic aggregation according to any one of claims 1 to 4, characterized in that: The data acquisition module acquires the target historical behavior data of the candidate user from the multiple media traffic platforms, including: Obtaining historical behavior data of the candidate user; Extracting key features of the historical behavior data; Calculating correlation parameters between the historical behavior data and the interactive content based on key features of the historical behavior data; Based on the association parameters between the historical behavior data and the interactive content, it is determined whether the historical behavior data is target historical behavior data.

6. The education empowerment system based on all-media traffic aggregation according to any one of claims 1 to 4, characterized in that: The user determination module determines a target user by using a first artificial intelligence model based on target historical behavior data of the candidate users obtained from the multiple media traffic platforms, including: Determining, by the first artificial intelligence model and based on the target historical behavior data of the candidate user, a degree of matching between the candidate user and the training needs corresponding to the interactive content; The target user is determined based on the matching degree between the candidate user and the training needs corresponding to the interactive content.

7. The education empowerment system based on all-media traffic aggregation according to any one of claims 1 to 4, characterized in that: The portrait building module builds a user portrait based on the basic data and historical behavior data of the target user through a second artificial intelligence model, including: Determine multiple known user profile features of the target user based on the basic data and historical behavior data of the target user; Predicting missing user profile features based on the historical behavior data of the target user using the second artificial intelligence model; The user profile is established based on the multiple known user profile features and the predicted missing user profile features.

8. The education empowerment system based on all-media traffic aggregation according to any one of claims 1 to 4, characterized in that: The portrait building module predicts missing user portrait features based on the historical behavior data of the target user using the second artificial intelligence model, including: Determine similar target users based on the historical behavior data of the target user; The second artificial intelligence model predicts missing user portrait features based on the basic data and historical behavior data of the similar target users.

9. The education empowerment system based on all-media traffic aggregation according to any one of claims 1 to 4, characterized in that: The institution matching module matches the target education and training institution based on the user profile, including: For each of the education and training institutions, based on the user portrait and the training target portrait corresponding to each education and training program provided by the education and training institution, the demand matching value of the education and training institution is calculated; based on the user portrait and the address distribution information of the education and training institution, the location matching value of the education and training institution is calculated; based on the historical user evaluation information and achievement conversion information of the education and training institution, the matching priority value of the education and training institution is calculated; The target education and training institution is determined based on the demand matching value, location matching value and matching priority value of each education and training institution.

10. The educational empowerment method based on all-media traffic aggregation is characterized by: include: Deliver interactive content on multiple media traffic platforms; Obtaining user behavior data on the interactive content; Determining candidate users based on user behavior data regarding the interactive content; Acquire target historical behavior data of the candidate user from the multiple media traffic platforms; Determine a target user based on the target historical behavior data of the candidate users obtained from the multiple media traffic platforms; Create a user profile based on the target user's basic data and historical behavior data through an artificial intelligence model; Based on the user portrait, match the target education and training institution.

Images