University network personalized content pushing system and method

Abstract

The invention discloses a university network personalized content pushing system and method, and relates to the field of computer information application. According to the university network personalized content pushing system and method, the interest label pool is established, the label browsing duration of the user is taken as a basic interest growth point, a corresponding data analysis model is designed, content pushing is performed according to the interest label direction of the user along with the increase of the use time of the user, the user is guided to explore and develop according to the interest direction, and the user experience is improved. The user is helped to perform deep learning with interest as a guide, meanwhile, by designing an interest weight calculation model and dynamically analyzing browsing conditions of the user on each label module, whether interest labels of the user have increasing, attenuating or drifting conditions or not is analyzed, and then dynamic maintenance is performed according to interest changes of the user. It is guaranteed that the pushed content can be in the interest increasing direction of the user, the user is helped to further mine the interest, and meanwhile the exploration depth of the user for the interested content is increased.

Description

Technical Field

[0001] The present invention relates to the technical field of computer information applications, and particularly to a personalized content push system and method for college networks. Background Art

[0002] College networks are usually interconnected network systems with a certain degree of privacy built by independent colleges or alliances of multiple colleges, and then maintained by a dedicated team. College students use such local area networks for relevant communication, sharing, and learning, and it is a functional website service with directional function blocking.

[0003] In the prior art, usually, an account is registered and then logged in through the account. However, the functions of such service-based websites are relatively scarce. When pushing messages, usually, relevant unified information is pushed according to the account type, such as teacher or student. This results in a high broad spectrum of the entire service website, but for students, there is a lack of corresponding guiding effect, and it is impossible to provide professional personalized content according to the students' interest directions, thereby guiding students to carry out special development behaviors. Therefore, a personalized content push system and method for college networks are provided. Summary of the Invention

[0004] Aiming at the deficiencies of the prior art, the present invention provides a personalized content push system and method for college networks, which solves the problems that existing college websites have relatively scarce functions, usually push relevant unified information, lack corresponding guiding effects for students, and cannot provide professional personalized content according to the students' interest directions, thereby guiding students to carry out special development.

[0005] To achieve the above objectives, the present invention is realized through the following technical solutions: A personalized content push method for college networks includes the following steps;

[0006] Step 1: Obtain basic user information;

[0007] Step 2: Initially present a unified multi-element push interface;

[0008] Step 3: Establish user interest tags based on user browsing information;

[0009] Step 4: Establish different tag pools for different interest tags;

[0010] Step 5: Secondarily push the content of the interest tags they are concerned about;

[0011] Step 6: Push content in corresponding proportions according to the ratings of each tag, and at the same time push relevant content according to its classification;

[0012] Step 7: Continuously repeat Step 5 and Step 6 to gradually improve the personalized push content for this user;

[0013] In step three, by analyzing the browsing duration of the user on each piece of content, it is determined whether the user has a real interest in the content. By setting a time threshold, behaviors exceeding the threshold are considered valid interests, while behaviors below the threshold are considered misclicks or shallow browsing. The specific implementation logic is as follows;

[0014] Monitor the browsing time H of the web page, and let T short be the invalid duration, and T valid be the valid duration. At the same time, set T short = 5s, and T valid = 30s;

[0015] If H < T short , it is judged as a misclick; if H > T valid , it is judged as a valid interest. If T short ≤ H < T valid , it is judged that there is a potential interest;

[0016] Thus, based on the above-designed interest weight calculation model;

[0017]

[0018] where H ι represents the duration of the user's ι-th browsing; where valid ι represents whether this browsing is a valid browsing, 1 for valid and 0 for invalid;

[0019] Then, according to the distribution of the browsing duration, the weights of the interest tags are dynamically adjusted. When the daily browsing duration of a certain type of tag increases, the weight of that tag is increased.

[0020] Preferably, in step one, when obtaining the user's basic information, let the user fill in relevant information through the registration interface, and at the same time, comprehensive analysis can be carried out according to the relevant information of the school, major, and campus clubs the user participates in;

[0021] In step two, the initial push is in the cold start state. At this time, by providing a short interest questionnaire, the user's interests are initially understood.

[0022] Preferably, in step three, a dynamic adjustment mechanism is established for the user's interest tags. As time goes by, the types of the user's interest tags drift, and there are problems of growth and decay in the interest degrees of various tags. At this time, the weights are adjusted according to the browsing volumes of various tags of the user;

[0023] By establishing a browsing threshold function, let D = browsing duration; L ι = the duration of the user's ι-th browsing, so as to obtain a new weight dynamic adjustment function as;

[0024] W new W(t)=α·W old (t)+(1 - α)·W now (t)

[0025] where W old (t) is the historical weight; W now (t) is the current weight; α is the weight coefficient.

[0026] Preferably, in the fourth and fifth steps, first, the tag content needs to be classified in a general direction according to the type, then, according to the large field of the tag, it is further classified systematically according to the detailed information in this field, then accurate tags are given according to the browsing situation of the user, and personalized content is further pushed to the user based on the detailed tags, and the browsing situation of the user is focused on at this stage to avoid incorrect tags.

[0027] Preferably, the seventh step executes a closed-loop cycle process. By continuously adjusting the weights of the user's various interest tag contents, then adjusting the proportion of the pushed content, and then analyzing the user feedback according to the browsing situation, whether the browsing duration of the corresponding increased pushed content increases proportionally. If so, it is judged that the weight of this tag content increases, and at this time, the push of this tag content continues to increase. If not, it is necessary to judge whether the user has interest drift and interest decline. By understanding the user's search record, whether there is a new tag development direction for the user. If so, content recommendation is made according to the new tag. If not, by reducing the weight coefficient of this tag, recalculating, and then pushing the tag content with relatively increased weight according to the weight change.

[0028] Preferably, according to the change of the user's interest tags, a model of the weight ratio of the user's interest content is designed. This model comprehensively considers the user's behavior data, time decay factor, and interest relevance, so as to effectively adjust the weights of the user's interest tag changes according to different situations;

[0029] In the model of the weight ratio of the user's interest content, the weight W(Q) of the user for an interest tag Q can be expressed as;

[0030]

[0031] where A(Q) is the interest weight based on the user's behavior data; B(Q) is the interest weight based on time decay; C(Q) is the interest weight based on interest relevance; meanwhile β and γ are weight coefficients, satisfying

[0032] Preferably, in the interest weight based on user behavior data, A(Q) reflects the user's direct behavior preference for the interest tag Q, including browsing duration, click times, and interaction behaviors, and its functional expression can be represented as;

[0033]

[0034] D(Q) is the total browsing duration of the user under the interest tag Q; E(Q) is the total number of clicks of the user under the interest tag Q; F(Q) is the total number of interactions of the user under the interest tag Q.

[0035] Preferably, in the interest weight based on time decay, B(Q) reflects the situation of the user's interest in the interest tag Q decaying over time, and its functional expression can be represented as;

[0036]

[0037] where Q now is the current time; Q ι is the time of the user's ι-th behavior; b ι (Q) is the contribution of the user's ι-th behavior to the interest Q; λ is the time decay coefficient, which controls the speed of interest decay.

[0038] Preferably, in the interest weight based on interest correlation, C(Q) reflects the correlation between the interest tag Q and other interest tags. If the user is also interested in other tags associated with Q, the weight of Q should increase, and its functional expression can be represented as;

[0039]

[0040] where S(Q) represents the set of other interest tags associated with the interest tag Q; sim(Q,S) represents the similarity between the interest tag Q and S; W(S) is the weight of the user for the interest tag S.

[0041] Preferably, the university network personalized content push system includes;

[0042] A data collection module, which is used to collect the user's basic information when the user registers, and at the same time, through real-time web monitoring, obtain the user's dynamic data;

[0043] A data processing module, which is used to clean and classify the obtained user data information, and then convert it into a computer language;

[0044] A user tag model, which is established based on the tag pool, and by continuously feeding in user tag information, automatically generates a corresponding user personalized tag model, and adjusts the model in real time according to the dynamic weight changes;

[0045] The data storage module stores the user's historical tag information by establishing a data storage server, providing a database for the generation of the user tag model.

[0046] The data push module automatically pushes corresponding tag content to the user account based on the user tag model.

[0047] The present invention discloses a personalized content push system and method for college networks, and the beneficial effects thereof are as follows:

[0048] 1. In the personalized content push method for college networks, by establishing an interest tag pool, with the user's tag browsing duration as the basis for the interest growth point, through designing a corresponding data analysis model, as the user's usage time increases, data snowballing is continuously carried out, and at the same time, various data modules are utilized to achieve data circulation. According to the user's interest tag direction, corresponding content is pushed to guide the user to explore and develop in the interest direction, and help the user conduct in-depth learning with interest as the guide.

[0049] 2. In the personalized content push method for college networks, by designing an interest weight calculation model, by dynamically analyzing the user's browsing situation of each tag module to analyze whether the user's interest tags are growing, decaying or drifting, and then dynamically maintaining according to the user's interest changes, ensuring that the pushed content can be in the user's interest growth direction, helping the user further explore interests, and at the same time increasing the exploration depth of the content the user is interested in. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0051] Figure 1 It is the overall flowchart of the personalized content push method of the present invention;

[0052] Figure 2 It is the flowchart of the method for automatically adjusting the proportion of interest tags according to browsing data during the personalized content push process of the present invention;

[0053] Figure 3 It is the functional framework diagram of the personalized content push system of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0054] To make the objectives, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be described clearly and completely. Apparently, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0055] By providing a personalized content push system and method for college networks in the embodiments of the present application, the problems of existing college websites are solved. The functions of existing college websites are relatively scarce. Usually, relevant unified information is pushed. For students, there is a lack of corresponding guiding effects and it is impossible to provide professional personalized content according to the students' interest directions, thus guiding students to develop specifically.

[0056] To better understand the above technical solutions, the above technical solutions will be described in detail below in conjunction with the accompanying drawings of the specification and specific implementation manners.

[0057] Embodiment 1

[0058] An embodiment of the present invention discloses a method for pushing personalized content on a college network. As shown in the appendix Figures 1 - 3 , it includes the following steps;

[0059] Step 1: Obtain basic user information;

[0060] Step 2: Initially present a unified multi-element push interface;

[0061] Step 3: Establish user interest tags according to user browsing information;

[0062] Step 4: Establish different tag pools for different interest tags;

[0063] Step 5: Secondarily push the content of the interest tags they are concerned about;

[0064] Step 6: Push content in corresponding proportions according to the ratings of each tag, and at the same time push relevant content according to its classification;

[0065] Step 7: Continuously repeat Step 5 and Step 6 to gradually improve the personalized push content for this user;

[0066] In Step 3, by analyzing the browsing duration of the user on each piece of content, it is judged whether the user has a real interest in the content. By setting a time threshold, behaviors exceeding the threshold are considered valid interests, and behaviors below the threshold are considered misclicks or shallow browsing. The specific implementation logic is as follows;

[0067] By monitoring the browsing time H of the web page, let T short be the invalid duration, Tvalid is the effective duration, and at the same time, set T short = 5s, T valid = 30s;

[0068] If H < T short , it is judged as a misclick; if H > T valid , it is judged as effective interest. If T short ≤ H < T valid , it is judged that there is potential interest;

[0069] Thus, based on the above design of the interest weight calculation model;

[0070]

[0071] where H ι represents the duration of the user's ι-th browsing; where valid ι represents whether this browsing is an effective browsing. Effective is 1, and ineffective is 0;

[0072] Then, according to the distribution of the browsing duration, dynamically adjust the weights of the interest tags. When the daily browsing duration of a certain type of tag increases, increase the weight of this tag.

[0073] In step three, establish a dynamic adjustment mechanism for the user's interest tags. As time goes by, the types of the user's interest tags drift, and there are problems of growth and decay in the interest degrees of various tags. At this time, adjust the weights according to the browsing volumes of various tags of the user;

[0074] By establishing a browsing threshold function, let D = browsing duration; L ι = the duration of the user's ι-th browsing, so as to obtain a new weight dynamic adjustment function as;

[0075] W new (t) = α·W old (t) + (1 - α)·W now (t)

[0076] where W old (t) is the historical weight; W now (t) is the current weight; α is the weight coefficient.

[0077] In step one, when obtaining the user's basic information, let the user fill in relevant information through the registration interface, and at the same time, comprehensive analysis can be carried out according to the relevant information of the school, major, and campus clubs participated by the user;

[0078] In step two, the initial push is in the cold start state. At this time, provide a short interest questionnaire to initially understand the user's interests.

[0079] In Steps 4 and 5, first, the tag content needs to be classified in a general direction according to the type, and then, according to the major field of the tag, it is further classified systematically according to the detailed information in this field. Then, accurate tags are given according to the user's browsing situation, and further personalized content is pushed to the user based on the detailed tags. At this stage, key attention is paid to the user's browsing situation to avoid incorrect tags.

[0080] Step 7 executes a closed-loop cycle process. By continuously adjusting the weights of the user's various interest tag contents, then adjusting the proportion of the pushed content, and then analyzing the user feedback based on the browsing situation to see if the browsing duration of the corresponding increased pushed content shows a proportional increase. If so, it is judged that the weight of this tag content increases, and at this time, continue to increase the push of this tag content. If not, it is necessary to judge whether the user has interest drift and interest decline. By looking at the user's search records to see if there is a new tag development direction. If there is, content is recommended according to the new tag. If not, by reducing the weight coefficient of this tag, recalculating, and then pushing the tag content with a relatively increased weight according to the weight change.

[0081] Design a model for the weight ratio of the user's interest content according to the changes in the user's interest tags. This model comprehensively considers the user's behavior data, time decay factors, and interest relevance, so as to effectively adjust the weights of the changes in the user's interest tags according to different situations;

[0082] In the model of the weight ratio of the user's interest content, the weight W(Q) of the user for a certain interest tag Q can be expressed as;

[0083]

[0084] Among them, A(Q) is the interest weight based on the user's behavior data; B(Q) is the interest weight based on time decay; C(Q) is the interest weight based on interest relevance; meanwhile β and γ are weight coefficients, satisfying

[0085] In the interest weight based on the user's behavior data, A(Q) reflects the user's direct behavior preference for the interest tag Q, including browsing duration, click times, interaction behaviors, etc., and its functional expression can be;

[0086]

[0087] D(Q) is the total browsing duration of the user under the interest tag Q; E(Q) is the total click times of the user under the interest tag Q; F(Q) is the total interaction times of the user under the interest tag Q.

[0088] In the time decay-based interest weight, B(Q) reflects the decay of the user's interest in the interest tag Q over time, and its functional form can be expressed as;

[0089]

[0090] where Q now is the current time; Q ι is the time of the user's ι-th behavior; b ι (Q) is the contribution of the user's ι-th behavior to the interest Q; λ is the time decay coefficient, which controls the speed of interest decay.

[0091] In the interest weight based on interest relevance, C(Q) reflects the relevance of the interest tag Q to other interest tags. If the user is also interested in other tags associated with Q, the weight of Q increases, and its functional form can be expressed as;

[0092]

[0093] where S(Q) represents the set of other interest tags associated with the interest tag Q; sim(Q, S) represents the similarity between the interest tag Q and S; W(S) is the weight of the user's interest in the interest tag S.

[0094] Example Two

[0095] Refer to the appendix Figures 1 - 3 , the university network personalized content push system includes;

[0096] A data collection module, which is used to collect the user's basic information when the user registers, and at the same time, through real-time web monitoring, obtain the user's dynamic data;

[0097] A data processing module, which is used to clean and classify the obtained user data information, and then convert it into computer language;

[0098] A user tag model, which is established based on a tag pool. By continuously feeding in user tag information, it automatically generates a corresponding user personalized tag model and adjusts the model in real time according to the dynamic weight changes;

[0099] A data storage module, which stores the user's historical tag information by establishing a data storage server and provides a database for the generation of the user tag model;

[0100] A data push module, which automatically pushes corresponding tag content to the user account based on the user tag model.

[0101] Through this personalized content push system for college networks, through the coordinated cooperation of each module, it is possible to establish a corresponding dynamic data tag model based on the user's account starting from the account registration. In the initial cold start stage, due to the lack of corresponding data samples as the underlying support, a dense push scheme can be adopted to push multi-type tag content, or the data can be started by designing a corresponding questionnaire survey on the registration page for simple tag content. Then, as the user's usage time increases, data snowballing is continuously carried out, and at the same time, each data module is utilized to achieve data circulation, and corresponding content is pushed according to the user's interest tag direction, guiding the user to explore and develop in the interest direction, and helping the user to conduct in-depth learning with interest as the guide.

[0102] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art of this industry should understand that the present invention is not limited by the above embodiments. What is described in the above embodiments and the specification only illustrates the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the present invention claimed. The scope of protection claimed by the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for personalized content push in college networks, characterized in that including the following steps; Step 1: Obtain the user's basic information; Step 2: Initially present a unified multi-source push interface; Step 3: Establish user interest tags based on the user's browsing information; Step 4: Create different tag pools for different interest tags; Step 5: Push the content of the interest tags the user is concerned about for the second time; Step 6: Push content in corresponding proportions according to the ratings of each tag, and at the same time push relevant content according to its classification; Step 7: Continuously repeat Steps 5 and 6 to gradually improve the personalized push content for this user; In Step 3, by analyzing the browsing duration of the user on each piece of content, it is judged whether the user has a real interest in this content. By setting a time threshold, behaviors exceeding this threshold are considered valid interests, and behaviors below this threshold are considered misclicks or shallow browsing. The specific implementation logic is as follows; Monitor the browsing time H through the web page, and let T short be the invalid duration, and T valid be the valid duration. At the same time, set T short = 5s, and T valid = 30s; If H < T short , it is judged as a misclick; if H > T valid , it is judged as effective interest. If T short ≤ H < T valid , it is judged that there is potential interest; Thus, based on the above-designed interest weight calculation model; where H ι represents the duration of the ι-th browsing by the user; where valid ι represents whether this view is a valid view. 1 indicates valid and 0 indicates invalid; Then, according to the distribution of the browsing duration, dynamically adjust the weights of the interest tags. When the daily browsing duration of a certain type of tag increases, increase the weight of this tag.

2. The personalized content push method for college networks according to claim 1, wherein: In Step 1, when obtaining the user's basic information, let the user fill in relevant information through the registration interface, and at the same time, comprehensive analysis can be carried out according to the information of the school, major, and campus clubs the user participates in; In Step 2, the initial push is in a cold start state. At this time, provide a short interest questionnaire to preliminarily understand the user's interests.

3. The personalized content push method for college networks according to claim 1, wherein: In Step 3, establish a dynamic adjustment mechanism for the user's interest tags. As time goes by, there are drifts in the types of the user's interest tags, and there are growth and decay problems in the interest degrees of various tags. At this time, adjust the weights according to the browsing volumes of various tags of the user; By establishing a browsing threshold function, let D = browsing duration; L ι = the duration of the user's ι-th browsing, so as to obtain a new weight dynamic adjustment function as; W new W(t) = α·W old (t) + (1 - α)·W now (t) Among them, W old (t) is the historical weight; W now (t) is the current weight; α is the weight coefficient.

4. The method for personalized content push in a university network according to claim 1, wherein: In Steps 4 and 5, first, the tag content needs to be classified in a general direction according to the type, and then, according to the major field of the tag, it is further classified systematically according to the detailed information in this field. Then, give accurate tags according to the user's browsing situation, and push further personalized content to the user according to the detailed tags. And in this stage, focus on the user's browsing situation to avoid incorrect tags.

5. The method for personalized content push in a university network according to claim 1, wherein: Step 7 executes a closed-loop cycle process. By continuously adjusting the weights of the content of each user interest tag, then adjusting the proportion of the pushed content, and then analyzing the user feedback according to the browsing situation, whether the browsing duration of the corresponding increased pushed content increases proportionally. If so, it is judged that the weight of this tag content increases. At this time, continue to increase the push of this tag content. If not, it is necessary to judge that the user has interest drift and interest decay. Understand whether the user has a new tag development direction through the user's search records. If so, recommend content according to the new tag. If not, reduce the weight coefficient of this tag, recalculate, and then push the content of the tag with a relatively increased weight according to the weight change.

6. The method for personalized content push in a university network according to claim 5, characterized in that: Design a model for the weight ratio of the user's interest content according to the changes in the user's interest tags. This model comprehensively considers the user's behavior data, time decay factors, and interest relevance, so as to effectively adjust the weights of the changes in the user's interest tags according to different situations; In the model of the weight ratio of user interest content, the weight W(Q) of a user for a certain interest tag Q can be expressed as; Among them, A(Q) is the interest weight based on user behavior data; B(Q) is the interest weight based on time decay; C(Q) is the interest weight based on interest relevance; meanwhile β and γ are weight coefficients, satisfying 7. The method for personalized content push in a university network according to claim 6, characterized in that: In the interest weight based on user behavior data, A(Q) reflects the user's direct behavior preference for the interest tag Q, including browsing duration, click times, and interaction behavior, and its functional expression can be written as; D(Q) is the total browsing duration of the user under the interest tag Q; E(Q) is the total number of clicks of the user under the interest tag Q; F(Q) is the total number of interactions of the user under the interest tag Q.

8. The method for personalized content push in a college network according to claim 6, wherein: In the interest weight based on time decay, B(Q) reflects the decay of the user's interest in the interest tag Q over time, and its functional expression can be written as; Where Qnow is the current time; Q ι is the time of the ι-th behavior of the user; b ι (Q) is the contribution of the ι-th behavior of the user to the interest Q; λ is the time decay coefficient, which controls the speed of interest decay.

9. The method for personalized content push in a university network according to claim 8, characterized in that: In the interest weight based on interest correlation, C(Q) reflects the correlation between the interest tag Q and other interest tags. If the user is also interested in other tags associated with Q, the weight of Q should increase, and its functional expression can be written as; Among them, S(Q) represents the set of other interest tags associated with the interest tag Q; sim(Q, S) represents the similarity between the interest tag Q and S; W(S) is the weight of the user for the interest tag S.

10. A personalized content push system for college networks, which is used to execute the personalized content push method for college networks described in any one of claims 1-9, and is characterized in that: Including; A data collection module, which is used to collect the basic information of the user when the user registers, and at the same time obtain the user's dynamic data through real-time web monitoring; A data processing module, which is used to clean and classify the obtained user data information, and then convert it into a computer language; A user tag model, which is established based on a tag pool. By continuously feeding in user tag information, it automatically generates a corresponding user personalized tag model and adjusts the model in real time according to the dynamic weight changes; A data storage module, which stores the user's historical tag information by establishing a data storage server and provides a database for the generation of the user tag model; A data push module, which automatically pushes the corresponding tag content to the user account based on the user tag model.

Images