Page user big data processing method and system for intelligent push

Abstract

The present invention relates to the field of big data push technology, and specifically provides a page user big data processing method and system for intelligent push. The present invention achieves in-depth analysis of user behavior patterns by obtaining page user interaction data for preliminary screening of potential push targets, especially interactive behavior events to be mined and multiple operation response image information contained therein. By performing correlation analysis on these operation response image information, a set of associated image information reflecting user interest associations can be constructed, and then the push demand relationship topology of each interactive behavior event to be mined is generated, which not only reveals the user's potential needs and preferences in different interactive behaviors, but also extracts the basic push decision features of each interactive behavior event through push decision feature mining. Finally, based on the determined target push decision features, an intelligent push strategy for the preliminary screening of potential push targets can be generated, which significantly improves the accuracy of the push and optimizes the information push service.

Description

Technical Field

[0001] The present invention relates to the field of big data push technology, and in particular to a method and system for processing page user big data applied to intelligent push. Background Art

[0002] In the era of information explosion, how to accurately push content of interest to users has become an important challenge. Traditional push strategies often rely on users' historical behavior data, ignoring the immediate interests and preference changes shown by users in specific interactive behaviors, resulting in limited accuracy of pushed content and user satisfaction. In particular, existing technologies fail to effectively integrate and analyze the rich image information generated by users during page interaction, which contains key clues to user interests. Therefore, there is an urgent need for a technical solution that can deeply mine the information in user interactive behaviors and build push strategies based on it. Summary of the invention

[0003] The present invention at least provides a method and system for processing page user big data applied to intelligent push.

[0004] The present invention provides a page user big data processing method for intelligent push, which is applied to an intelligent push system. The method comprises: obtaining page user interaction data for preliminary screening of potential push targets, the page user interaction data comprising at least one interactive behavior event to be mined of the preliminary screening potential push targets, the interactive behavior event to be mined comprising a plurality of operation response image information; correlating the plurality of operation response image information included in each of the interactive behavior events to be mined according to the page user interaction data, and obtaining at least one associated image information set corresponding to each of the interactive behavior events to be mined, the associated image information set being used to characterize the user interest association result of at least two operation response image information of the interactive behavior events to be mined; generating a push demand relationship topology of each of the interactive behavior events to be mined according to the associated image information set; mining push decision features of the push demand relationship topology of each of the interactive behavior events to be mined, and obtaining basic push decision features corresponding to each of the interactive behavior events to be mined; determining the target push decision features of each of the interactive behavior events to be mined according to the basic push decision features, and generating an intelligent push strategy for the preliminary screening potential push targets according to the target push decision features of each of the interactive behavior events to be mined.

[0005] The present invention also provides an intelligent push system, comprising a processor and a memory; the processor and the memory are communicatively connected, and the processor is used to read a computer program from the memory and execute it to implement the above method.

[0006] The present invention also provides a computer-readable storage medium on which a computer program is stored, and the computer program implements the above method when running.

[0007] The technical solution provided by the embodiments of the present invention may include the following beneficial effects: the present invention achieves in-depth analysis of user behavior patterns by obtaining page user interaction data for preliminary screening of potential push targets, especially the interactive behavior events to be mined and their multiple operation response image information contained therein. By performing correlation analysis on these operation response image information, the present invention can construct a set of associated image information that reflects user interest associations, and then generate a push demand relationship topology for each interactive behavior event to be mined. This process not only reveals the user's potential needs and preferences in different interactive behaviors, but also extracts the basic push decision features of each interactive behavior event through push decision feature mining. Finally, based on the target push decision features determined by these basic features, the present invention can generate an intelligent push strategy for preliminary screening of potential push targets, significantly improving the accuracy of push and user satisfaction, and optimizing the service experience of information push.

[0008] For a description of the effects of the above-mentioned intelligent push system and computer-readable storage medium, please refer to the description of the above-mentioned method.

[0009] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] In order to more clearly illustrate the technical solution of the present invention, the following is a brief introduction to the drawings required for use in the embodiments. The drawings herein are incorporated into the specification and constitute a part of the specification. These drawings illustrate embodiments consistent with the present invention and are used together with the specification to illustrate the technical solution of the present invention. It should be understood that the following drawings only illustrate certain embodiments of the present invention and should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can be obtained based on these drawings without creative work.

[0011] Figure 1 It is a block diagram of an intelligent push system shown in the present invention.

[0012] Figure 2 It is a flow chart of a method for processing page user big data applied to intelligent push shown in the present invention. DETAILED DESCRIPTION

[0013] Here, exemplary embodiments will be described in detail, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Instead, they are only examples of devices and methods consistent with some aspects of the present invention.

[0014] Figure 1 The structural diagram of the intelligent push system 10 provided by the present invention includes a processor 102, a memory 104, and a bus 106. Among them, the memory 104 is used to store execution instructions, including a memory and an external memory. The memory can also be understood as an internal memory, which is used to temporarily store the operation data in the processor 102 and the data exchanged with the external memory such as the hard disk. The processor 102 exchanges data with the external memory through the memory. When the intelligent push system 10 is running, the processor 102 communicates with the memory 104 through the bus 106, so that the processor 102 executes the page user big data processing method applied to intelligent push of the present invention.

[0015] Please combine Figure 2 , Figure 2 It is a flow chart of a method for processing page user big data for intelligent push provided by the present invention, which is applied to an intelligent push system. The method may exemplarily include the following steps 101 to 105.

[0016] Step 101: The intelligent push system obtains page user interaction data of initially screened potential push targets, wherein the page user interaction data includes at least one to-be-mined interactive behavior event of the initially screened potential push targets, and the to-be-mined interactive behavior event includes multiple operation response image information.

[0017] The first step of the intelligent push system is to obtain the page user interaction data for the initial screening of potential push targets. The page user interaction data contains at least one interactive behavior event to be mined for the initial screening of potential push targets, and each interactive behavior event to be mined includes multiple operation response image information. In the actual scenario of enterprise operation management, the above content may come from various user behaviors on the enterprise website, mobile application or social media platform, such as clicks, browsing, collections, comments, etc. The intelligent push system uses relevant data collection technology to capture and record these interactive behavior events to be mined and their related operation response image information in real time.

[0018] Step 102: The intelligent push system associates the multiple operation response image information included in each of the interactive behavior events to be mined based on the page user interaction data, and obtains at least one associated image information set corresponding to each of the interactive behavior events to be mined, and the associated image information set is used to characterize the user interest association results of at least two operation response image information of the interactive behavior event to be mined.

[0019] After obtaining the page user interaction data, the next step of the intelligent push system is to associate the multiple operation response image information included in each interactive behavior event to be mined. The purpose of this step is to reveal the user's interest association between different operation response images, so as to more comprehensively understand the user's preferences and behavior patterns. The intelligent push system can conduct in-depth analysis of these image information through AI algorithms and models to obtain at least one associated image information set corresponding to each interactive behavior event to be mined. These associated image information sets are an important basis for the subsequent generation of push demand relationship topology.

[0020] Step 103: The intelligent push system generates a push demand relationship topology of each interactive behavior event to be mined based on the associated image information set.

[0021] After obtaining the associated image information set, the intelligent push system generates the push demand relationship topology of each interactive behavior event to be mined based on this information set. The push demand relationship topology is a graphical representation method that can intuitively display the user's demand association and transfer path between different interactive behavior events. Through this step, the intelligent push system can more accurately grasp the user's demand dynamics and change trends, and provide strong support for the subsequent push decision feature mining.

[0022] Step 104: The intelligent push system mines the push decision features of the push demand relationship topology of each of the interactive behavior events to be mined, and obtains the basic push decision features corresponding to each of the interactive behavior events to be mined.

[0023] With the push demand relationship topology as the basis, the intelligent push system then mines the push decision features of the push demand relationship topology of each interactive behavior event to be mined. The purpose of this step is to extract feature information with push decision value from the complex topological structure. The intelligent push system can deeply mine and analyze these feature information through machine learning algorithms and data mining technology to obtain the basic push decision features corresponding to each interactive behavior event to be mined. These feature information will be directly used for subsequent push strategy formulation.

[0024] Step 105: The intelligent push system determines the target push decision feature of each of the interactive behavior events to be mined based on the basic push decision feature, and generates the intelligent push strategy for pre-screening potential push targets based on the target push decision feature of each of the interactive behavior events to be mined.

[0025] The last step is for the intelligent push system to determine the target push decision features of each interactive behavior event to be mined based on the basic push decision features, and generate an intelligent push strategy for initial screening of potential push targets based on these target push decision features. This step is the core and destination of the entire technical solution. The intelligent push system formulates personalized push strategies by comprehensively considering multiple factors such as user interest preferences, behavior patterns, and demand dynamics. These strategies can accurately meet user needs, improve user satisfaction and loyalty, and also improve the company's operational efficiency and market competitiveness.

[0026] In order to better understand the application of the present invention in enterprise operation management, an example is given below. For example, an e-commerce company wants to improve the user activity and purchase conversion rate on its mobile application. They introduced this intelligent push system and customized its development and deployment. First, the intelligent push system obtains various interactive behavior events and operation response image information of users in the application in real time through the data interface of the mobile application. This information includes the user's click stream, browsing history, shopping cart addition, order submission, etc. Then, the intelligent push system conducts in-depth association analysis on this information and obtains the user's interest association and transfer path between different interactive behavior events. For example, the intelligent push system finds that after browsing a certain type of goods, users often search for related discount information or check other users' comments in a short time. Based on these association analysis results, the intelligent push system generates the push demand relationship topology of each interactive behavior event to be mined. These topological diagrams intuitively show the user's demand dynamics and changing trends in the application. Then, the intelligent push system mines the push decision features of these topological diagrams and obtains the basic push decision features of users under different interactive behavior events. For example, the intelligent push system finds that when users browse goods, they have significant differences in price sensitivity and brand preference. Finally, the smart push system developed a personalized push strategy based on these basic push decision characteristics. For users who are highly price sensitive, the smart push system will emphasize discounts and offers when pushing product information; and for users with strong brand preferences, the smart push system will prioritize pushing their favorite brand products. Through the application of this smart push system, e-commerce companies have successfully improved user activity and purchase conversion rates on their mobile applications. When using the application, users can experience more personalized and caring services; and companies can better meet user needs through precise push strategies.

[0027] It is worth mentioning that in the scenario where the intelligent push system is applied to enterprise operation and management, the push demand relationship topology, as the core invention point, plays a pivotal role. It not only reveals the deep-seated needs of users in interactive behaviors, but also provides a solid basis for the subsequent push strategy formulation.

[0028] First, the construction of the push demand relationship topology is based on the comprehensive capture and recording of user interaction behavior events. In the actual scenario of enterprise operation and management, every user behavior may contain potential needs and preferences. The intelligent push system uses advanced data collection technology to capture and record various user behavior events on websites, mobile applications or social media platforms in real time, such as clicks, browsing, collections, comments, etc. These behavior events constitute the basic nodes of the push demand relationship topology, and each node represents the specific needs or preferences of users at a certain moment.

[0029] Secondly, the core of push demand relationship topology is to reveal the demand association and transfer path between different interactive behavior events of users. In traditional push strategies, people often only focus on the current single behavior or attribute of users, while ignoring the internal connection and dynamic changes between user behaviors. Push demand relationship topology uses complex algorithms and models to conduct in-depth analysis of the association between different behavior events of users, thereby revealing the internal logic and transfer rules of user needs. For example, in an e-commerce scenario, the system may find that after browsing a certain category of goods, users often further search for related discount information or check other users' reviews. The revelation of this demand association and transfer path provides strong support for the subsequent formulation of push strategies.

[0030] Then, the push demand relationship topology is also dynamic and real-time. User needs and preferences are constantly changing, and the push demand relationship topology can capture such changes in real time and update and optimize them. Through continuous learning and adjustment, the system enables the push demand relationship topology to more accurately reflect the user's current needs and preferences, thereby improving the accuracy and effectiveness of the push strategy. This dynamic and real-time feature makes the push demand relationship topology more adaptable and flexible in dealing with complex and changing user needs.

[0031] In the specific application process of pushing demand relationship topology, the intelligent push system will also combine the user's personal attributes and historical behavior data to further refine and optimize it. For example, in an e-commerce scenario, the system may generate a personalized push demand relationship topology for the user based on multi-dimensional data such as the user's purchase history, browsing history, and search behavior. In this way, each user will have a unique topology that accurately reflects the user's demand dynamics and changing trends in the e-commerce application.

[0032] In addition, push demand relationship topology can also be combined with other advanced technical means to further improve the effectiveness of push strategies and user experience. For example, the system can use machine learning algorithms to deeply mine and analyze topology maps to discover hidden rules and patterns; or use natural language processing technology to perform sentiment analysis on user behaviors and comments, so as to more accurately grasp user needs and preferences. The application of these technical means makes the application of push demand relationship topology in enterprise operation management more extensive and in-depth.

[0033] In actual enterprise operation and management scenarios, the application of push demand relationship topology has brought significant results and improvements. Taking e-commerce companies as an example, by introducing intelligent push systems and push demand relationship topology technology, companies can more accurately grasp user needs and preferences and develop more personalized push strategies.

[0034] In addition to e-commerce companies, push demand relationship topology can also be widely used in other types of enterprise operation and management scenarios. For example, in social media platforms, the system can use push demand relationship topology to discover the association and transfer path between different social behaviors of users, so as to formulate more accurate social recommendation strategies; in online education platforms, the system can use topology maps to reveal the changes in students' needs and preferences during the learning process, and provide teachers with more personalized teaching suggestions; in the financial industry, the system can use topology maps to analyze users' behavior patterns and risk preferences in financial transactions, and provide financial institutions with more accurate risk control and investment suggestions.

[0035] It can be seen that the push demand relationship topology, as the core invention of the intelligent push system, has wide application and practicality in enterprise operation management. It can not only deeply mine the user's interactive behavior events and reveal the user's demand association and transfer path between different behaviors; it can also combine the user's personal attributes and historical behavior data to generate personalized push strategies for them.

[0036] In addition, as far as the smart push strategy is concerned, first of all, the core of the smart push strategy lies in its in-depth mining and analysis of the user's interactive behavior events and related operation response image information. The smart push strategy not only focuses on the user's current single behavior or attribute, but also fully and deeply understands the user's needs and preferences in multiple interactive behaviors. Through advanced algorithms and models, the smart push strategy can reveal the inherent connections and transfer paths between different behaviors of users, so as to more accurately grasp the overall needs and preference changes of users. This comprehensive understanding and accurate grasp of user behavior provides a solid foundation for the subsequent push strategy formulation.

[0037] Secondly, the smart push strategy focuses on personalized and differentiated push. In traditional push strategies, a one-size-fits-all approach is often adopted to push the same content to all users. However, this approach ignores the individual differences and diverse needs among users. The smart push strategy, on the other hand, tailors push content to each user based on their unique needs and preferences. For example, in e-commerce scenarios, the smart push strategy recommends products that best meet their needs and preferences based on multi-dimensional data such as the user's purchase history, browsing history, and search behavior. This personalized and differentiated push method has greatly improved user satisfaction and loyalty, and also brought higher conversion rates and sales to companies.

[0038] Then, the smart push strategy is also dynamic and real-time. User needs and preferences are constantly changing, and the smart push strategy can capture such changes in real time and update and optimize them. Through continuous learning and adjustment, the smart push strategy can more accurately reflect the current needs and preferences of users, thereby improving the accuracy and effectiveness of the push strategy. This dynamic and real-time feature makes the smart push strategy more adaptable and flexible in dealing with complex and changing user needs.

[0039] In actual business operation and management scenarios, the application of intelligent push strategies has brought significant results and improvements. Taking e-commerce companies as an example, by introducing intelligent push strategies, companies can more accurately grasp user needs and preferences and develop more personalized push strategies.

[0040] In addition to e-commerce companies, smart push strategies can also be widely used in other types of enterprise operation and management scenarios. For example, in social media platforms, smart push strategies can recommend social content and users that are more in line with their interests based on their historical behaviors and preferences; in online education platforms, smart push strategies can recommend more personalized learning resources and courses based on students' learning progress and preferences; in the financial industry, smart push strategies can provide users with more accurate investment advice and risk management solutions based on their trading behaviors and risk preferences.

[0041] The implementation of intelligent push strategies requires advanced technical means and data support. First, companies need to establish a complete data collection and storage system to capture and record user interactive behavior events and operation response image information in real time. Secondly, companies need to use advanced algorithms and models to conduct in-depth mining and analysis of these data to reveal the associations and transfer paths between different user behaviors. Finally, companies need to develop personalized push strategies based on the analysis results and accurately reach users through advanced push technologies.

[0042] In the process of implementing smart push strategies, enterprises also need to pay attention to protecting user privacy and data security. User personal information and behavioral data are sensitive. When collecting, storing and using this data, enterprises need to strictly abide by relevant laws, regulations and ethical standards to ensure that user privacy and data security are not violated. At the same time, enterprises also need to establish a complete user feedback mechanism, collect and process user feedback on push content in a timely manner, continuously optimize and improve push strategies, and improve user experience and satisfaction.

[0043] It can be seen that the intelligent push strategy can not only deeply mine users' interactive behavior events and operation response image information, revealing the associations and transfer paths between different behaviors of users; it can also tailor push content for each user based on their unique needs and preferences.

[0044] In summary, the present invention achieves an in-depth analysis of user behavior patterns by obtaining page user interaction data for preliminary screening of potential push targets, especially the interactive behavior events to be mined and their multiple operation response image information contained therein. By performing correlation analysis on these operation response image information, the present invention can construct a set of associated image information that reflects user interest associations, and then generate a push demand relationship topology for each interactive behavior event to be mined. This process not only reveals the user's potential needs and preferences in different interactive behaviors, but also extracts the basic push decision features of each interactive behavior event through push decision feature mining. Finally, based on the target push decision features determined by these basic features, the present invention can generate an intelligent push strategy for preliminary screening of potential push targets, significantly improving the accuracy of push and user satisfaction, and optimizing the service experience of information push.

[0045] In some optional embodiments, based on the page user interaction data, multiple operation response image information included in each interactive behavior event to be mined are associated to obtain at least one associated image information set corresponding to each interactive behavior event to be mined, including: for each interactive behavior event to be mined in the page user interaction data, the following steps are implemented respectively: when the number of the operation response image information included in the interactive behavior event to be mined is two, the two operation response image information of the interactive behavior event to be mined are associated to obtain the associated image information set; when the number of the operation response image information included in the interactive behavior event to be mined is at least three, for the at least three operation response image information included in the interactive behavior event to be mined, user interest association of at least two operation response image information is implemented multiple times to obtain multiple initial associated image information sets, and based on the multiple initial associated image information sets, the associated image information set of the interactive behavior event to be mined is determined.

[0046] It can be understood that the process of associating multiple operation response image information included in each interactive behavior event to be mined based on the page user interaction data to obtain at least one associated image information set corresponding to each interactive behavior event to be mined is intended to deeply mine the image information generated by the user during the page interaction process, and reveal the user's potential interests and preferences in different interactive behaviors through association analysis, thereby providing strong support for the subsequent push strategy formulation.

[0047] First of all, the associated image information set is a core concept, which refers to the image information set that can represent the user interest association results obtained through association analysis in the interactive behavior event to be mined. The image information in this set is not isolated, but has some connection or commonality between each other, which reflects the user's interest or preference in a specific interactive behavior.

[0048] Next, for each interactive behavior event to be mined in the page user interaction data, correlation analysis is performed respectively. This process can be divided into two cases: one is that the number of operation response image information included in the interactive behavior event to be mined is two, and the other is that the number of operation response image information included in the interactive behavior event to be mined is at least three.

[0049] For the first case, when the number of operation response image information included in the interactive behavior event to be mined is two, the association analysis process is relatively simple. By directly associating the two operation response image information, a set of associated image information can be obtained. This association may be based on the similarity of image content, the consistency of user behavior, or other indicators that can reflect user interests. Through association analysis, the user's possible interests or preferences between the two operation response image information can be revealed, thereby providing a basis for the subsequent push strategy formulation.

[0050] For the second case, that is, when the number of operation response image information included in the interactive behavior event to be mined is at least three, the association analysis process is relatively complicated. At this time, it is necessary to implement the user interest association of at least two operation response image information multiple times to obtain multiple initial associated image information sets. This process may involve a variety of association analysis algorithms or models, such as content-based image association, user behavior-based image association, or deep learning-based image association. Through multiple association analyses, multiple initial associated image information sets can be obtained, each of which reflects a part of the user's interest or preference in a specific interactive behavior.

[0051] However, obtaining multiple initial associated image information sets is only the first step. It is necessary to further determine the final associated image information set of the interactive behavior event to be mined. This process may involve the fusion, screening or optimization of multiple initial associated image information sets. For example, by calculating the similarity or consistency between the various initial associated image information sets, the most representative set or the set that best reflects the user's interests can be selected as the final associated image information set. In addition, other information such as the user's historical behavior data and user portraits can also be considered to further improve the accuracy and effectiveness of the associated image information set.

[0052] In this way, the process of associating multiple operation response image information included in each interactive behavior event to be mined based on the page user interaction data is a complex and sophisticated data processing process. Through association analysis, the potential interests and preferences of users in different interactive behaviors can be revealed, which can provide strong support for the subsequent push strategy formulation. At the same time, it is also necessary to consider the selection and optimization of multiple factors and methods in the processing process to ensure the accuracy and effectiveness of the associated image information set. The final associated image information set not only reflects the user's interests or preferences in specific interactive behaviors, but also provides important basis and support for the formulation of intelligent push strategies.

[0053] In the next step, the associated image information set of the interactive behavior event to be mined is determined based on the multiple initial associated image information sets, including: obtaining the number of operation response image information in each of the initial associated image information sets; and determining the initial associated image information sets whose number is greater than the set number as the associated image information sets of the interactive behavior event to be mined.

[0054] It can be understood that this embodiment involves selecting the image information set that best represents the user's interests and preferences from multiple initial related image information sets as the related image information set for the interactive behavior event to be mined. This process is of great significance for the subsequent formulation of accurate push strategies and the improvement of user experience and operational effects.

[0055] First, multiple initial associated image information sets are obtained in the previous step by performing an associated analysis on multiple operation response image information included in each interactive behavior event to be mined. Each initial associated image information set reflects a part of the user's interest or preference in a specific interactive behavior. However, due to the complexity and diversity of user behavior, these initial associated image information sets may contain different amounts of operation response image information, and not all sets can accurately reflect the user's real interests and preferences.

[0056] Therefore, in the next step, it is necessary to further screen and determine these initial associated image information sets. Specifically, first obtain the number of operation response image information in each of the initial associated image information sets. The purpose of this step is to quantify the amount of information about user interests or preferences contained in each set. In theory, a set containing more operation response image information is more likely to accurately reflect the user's interests and preferences in a specific interactive behavior.

[0057] Secondly, the initial associated image information set with a number greater than the set number is determined as the associated image information set of the interactive behavior event to be mined. The set number here is a preset threshold, which is used to measure whether an initial associated image information set contains enough information to represent the user's interests and preferences. By setting this threshold, those sets that contain less operation response image information and may not accurately reflect the user's interests can be filtered out.

[0058] In actual applications, the specific value of the number needs to be adjusted according to the specific scenario and data. For example, in the application scenario of enterprise operation management, a reasonable threshold can be set based on historical data and experience. This threshold should be able to filter out those sets with insufficient information while retaining those sets that contain enough information to represent user interests and preferences.

[0059] Through the above steps, the intelligent push system can filter out the image information set that best represents the user's interests and preferences from multiple initial associated image information sets as the associated image information set for the interactive behavior event to be mined. This set not only reflects the user's interests and preferences in specific interactive behaviors, but also provides strong support for the subsequent push strategy formulation. For example, based on the image content, user behavior and other features in the associated image information set, a more accurate push strategy can be formulated to improve user experience and operational results.

[0060] In addition, it should be noted that in practical applications, the determination of the associated image information set may also need to consider other factors. For example, the user's historical behavior data, user portrait, business scenario, etc. may have an impact on the determination of the associated image information set. Therefore, when determining the associated image information set, it is necessary to comprehensively consider the selection and optimization of multiple factors and methods to ensure the accuracy and effectiveness of the associated image information set.

[0061] In summary, determining the associated image information set of the interactive behavior event to be mined based on multiple initial associated image information sets is a complex and important data processing link. By quantifying the amount of information about user interests or preferences contained in each set and setting reasonable thresholds for screening and determination, the intelligent push system can obtain the image information set that best represents user interests and preferences. This set not only provides strong support for the subsequent push strategy formulation, but also can bring greater value and benefits to enterprise operation management.

[0062] In some possible embodiments, generating the push demand relationship topology of each interactive behavior event to be mined based on the associated image information set includes: implementing the following steps for each interactive behavior event to be mined: for each associated image information set corresponding to the interactive behavior event to be mined, generating the push demand relationship topology corresponding to each associated image information set; generating the push demand relationship topology of the interactive behavior event to be mined based on the push demand relationship topology corresponding to each associated image information set.

[0063] The process involved in this embodiment, which generates the push demand relationship topology of each interactive behavior event to be mined based on the associated image information set, is a data processing and relationship building link. This link aims to deeply explore and construct the push demand relationship topology of each interactive behavior event to be mined based on the determined associated image information set, so as to provide strong support for the subsequent formulation of precise push strategies. Among them, the push demand relationship topology is a core concept, which refers to the topological structure that can characterize the relationship between push demands obtained through association analysis in the interactive behavior events to be mined. This structure not only reflects the intrinsic connection between push demands, but also provides an important reference basis for the subsequent formulation of push strategies.

[0064] For each interactive behavior event to be mined, the following steps are performed to generate its corresponding push demand relationship topology:

[0065] First, for each of the associated image information sets corresponding to the interactive behavior events to be mined, the push demand relationship is mined and analyzed. This process may involve a variety of data mining and relationship analysis algorithms, such as content-based image association analysis, user behavior-based push demand mining, etc. Through the application of these algorithms, key features that can characterize push demand can be extracted from the associated image information set, and the internal connections between these features can be revealed;

[0066] Secondly, based on the results of mining and analysis, the push demand relationship topology corresponding to each of the associated image information sets is generated. This topology structure graphically displays the relationship between push demands, where nodes represent push demands and edges represent the relationship between push demands. Through this topology structure, the correlation and dependency between different push demands can be clearly seen, providing an important reference for subsequent strategy formulation;

[0067] Finally, further fusion and optimization are performed based on the push demand relationship topology corresponding to each of the associated image information sets to generate the push demand relationship topology of the interactive behavior event to be mined. This process may involve operations such as merging, deduplication, and optimization of multiple topological structures to ensure that the push demand relationship topology finally generated can accurately reflect the push demand relationship in the interactive behavior event to be mined.

[0068] In practical applications, the process of generating push demand relationship topology requires comprehensive consideration of multiple factors and the selection and optimization of methods. For example, factors such as user historical behavior data, user portraits, and business scenarios can be considered to more accurately mine and analyze push demand relationships. At the same time, advanced technologies such as machine learning and deep learning can also be used to improve the accuracy and efficiency of push demand relationship mining.

[0069] Through the above steps, the intelligent push system can obtain the push demand relationship topology of each interactive behavior event to be mined. This topological structure not only reflects the internal connection and dependency between push demands, but also provides an important reference for the subsequent push strategy formulation. For example, a more accurate push strategy can be formulated based on the push demand relationship topology to improve user experience and operational results. At the same time, this topological structure can also be used to simulate and predict business scenarios, bringing greater value and benefits to enterprise operation management.

[0070] In the present invention, on the one hand, for each of the associated image information sets corresponding to the interactive behavior event to be mined, the push demand relationship topology corresponding to each of the associated image information sets is generated, including: for each of the associated image information sets corresponding to the interactive behavior event to be mined, the push demand relationship topology corresponding to each of the operation response image information used to create the associated image information set is obtained; the push demand relationship topology corresponding to each of the operation response image information is merged to obtain the push demand relationship topology of the associated image information set. On the other hand, based on the push demand relationship topology corresponding to each of the associated image information sets, the push demand relationship topology of the interactive behavior event to be mined is generated, including: obtaining the number of relationship topologies of the push demand relationship topology corresponding to the associated image information set; when the number of relationship topologies is one, the push demand relationship topology corresponding to the associated image information set is determined as the push demand relationship topology of the interactive behavior event to be mined; when the number of relationship topologies is multiple, the push demand relationship topologies corresponding to each of the associated image information sets are merged to obtain the push demand relationship topology of the interactive behavior event to be mined.

[0071] In the present invention, for the enterprise operation management application scenario, accurate push demand relationship topology can be generated for each interactive behavior event to be mined through sophisticated data processing and relationship construction, thereby providing solid data support for the subsequent formulation of precise push strategies.

[0072] On the one hand, for each associated image information set corresponding to the interactive behavior event to be mined, the present invention proposes a specific method for generating a push demand relationship topology corresponding to each associated image information set. First, for each associated image information set corresponding to the interactive behavior event to be mined, the system will obtain the push demand relationship topology corresponding to each operation response image information that creates these associated image information sets. These push demand relationship topologies are obtained through complex data mining and relationship analysis algorithms based on multiple factors such as the content of the operation response image information, user behavior, and business scenarios. They can accurately reflect the intrinsic connections and dependencies between push demands.

[0073] Secondly, after obtaining the push demand relationship topology corresponding to each operation response image information, the system fuses these topologies to obtain the push demand relationship topology of the associated image information set. This fusion process may involve operations such as merging, deduplication, and optimization of multiple topological structures to ensure that the final generated push demand relationship topology can fully and accurately reflect the push demand relationship in the associated image information set. Through this fused push demand relationship topology, enterprises can more clearly understand the correlation and dependency between different push demands, providing an important reference for subsequent strategy formulation.

[0074] On the other hand, the present invention also proposes a specific method for generating a push demand relationship topology of an interactive behavior event to be mined based on the push demand relationship topology corresponding to each associated image information set. First, the system obtains the number of relationship topologies of the push demand relationship topology corresponding to each associated image information set. This number reflects the complexity and diversity of the push demand relationship in the interactive behavior event to be mined.

[0075] Secondly, depending on the number of relationship topologies, the system will adopt different strategies to generate the push demand relationship topology of the interactive behavior event to be mined. When the number of relationship topologies is one, it means that the push demand relationship in the interactive behavior event to be mined is relatively simple. At this time, the system will directly determine this unique push demand relationship topology as the push demand relationship topology of the interactive behavior event to be mined. When the number of relationship topologies is multiple, it means that the push demand relationship in the interactive behavior event to be mined is relatively complex. At this time, the system needs to fuse the push demand relationship topologies corresponding to each associated image information set to obtain the push demand relationship topology of the interactive behavior event to be mined. This fusion process may also involve operations such as merging, deduplication, and optimization of multiple topological structures to ensure that the final generated push demand relationship topology can fully and accurately reflect the push demand relationship in the interactive behavior event to be mined.

[0076] In practical applications, the process of generating push demand relationship topology requires comprehensive consideration of multiple factors and selection and optimization of methods. For example, factors such as user historical behavior data, user portraits, and business scenarios can be considered to more accurately mine and analyze push demand relationships. At the same time, advanced technologies such as machine learning and deep learning can also be used to improve the accuracy and efficiency of push demand relationship mining. In addition, in order to ensure the accuracy and effectiveness of the push demand relationship topology, the generated topology needs to be continuously verified and optimized.

[0077] In summary, the present invention provides a powerful data support and decision-making basis for enterprise operation management by respectively generating push demand relationship topologies for each associated image information set corresponding to the interactive behavior event to be mined, and generating the push demand relationship topologies of the interactive behavior event to be mined based on these topologies. Through the application of this technical solution, enterprises can have a deeper understanding of users' interests and preferences, formulate more accurate push strategies, and improve user experience and operational results. At the same time, this technical solution can also help enterprises simulate and predict business scenarios, bringing greater value and benefits to enterprise operation management.

[0078] In some preferred embodiments, the target push decision feature of each of the interactive behavior events to be mined is determined based on the basic push decision feature, including: obtaining the scene push involvement weight of each of the basic push decision features, the scene push involvement weight being used to characterize the correlation between the push matching viewpoint corresponding to the basic push decision feature and the ideal matching viewpoint of the corresponding interactive behavior event to be mined; when the scene push involvement weight of the basic push decision feature is not less than the preset push involvement weight, the basic push decision feature is determined as the target push decision feature of the corresponding interactive behavior event to be mined; when the scene push involvement weight of the basic push decision feature is less than the preset push involvement weight, the basic push decision feature is determined as the target push decision feature of the corresponding interactive behavior event to be mined; When the preset push involvement weight is less than the preset push involvement weight, and the number of the operation response image information included in the corresponding interactive behavior event to be mined is two, the basic push decision feature is determined as the target push decision feature of the corresponding interactive behavior event to be mined; when the scene push involvement weight of the basic push decision feature is less than the preset push involvement weight, and the number of the operation response image information included in the corresponding interactive behavior event to be mined is at least three, the associated image information set of the interactive behavior event to be mined is adjusted to obtain an associated image adjustment set, and based on the associated image adjustment set, the target push decision feature of the corresponding interactive behavior event to be mined is determined.

[0079] It can be understood that for the application scenario of enterprise operation management, the present invention proposes a technical solution to determine the target push decision features of each interactive behavior event to be mined based on the basic push decision features. In detail, through fine feature matching and weight judgment, the target push decision features are accurately determined for each interactive behavior event to be mined, thereby providing a solid data foundation for the subsequent push strategy formulation.

[0080] First, the system needs to obtain the scenario push involvement weight of each basic push decision feature. This weight is used to characterize the correlation between the push matching viewpoint corresponding to the basic push decision feature and the ideal matching viewpoint of the corresponding interactive behavior event to be mined. In other words, it reflects the push matching degree of the basic push decision feature for the interactive behavior event to be mined in a specific scenario. The calculation of this weight may involve multiple factors, such as user behavior, business scenarios, historical push effects, etc., to ensure its accuracy and comprehensiveness.

[0081] Secondly, based on the comparison results between the scene push involvement weight and the preset push involvement weight, the system will adopt different strategies to determine the target push decision feature. When the scene push involvement weight of the basic push decision feature is not less than the preset push involvement weight, it means that the basic push decision feature has a high correlation with the ideal matching view of the interactive behavior event to be mined, so the system determines it as the target push decision feature of the corresponding interactive behavior event to be mined. This judgment ensures the accuracy and effectiveness of the target push decision feature.

[0082] However, in some cases, the scene push involvement weight of the basic push decision feature may be less than the preset push involvement weight. At this time, the system needs to further consider the number of operation response image information contained in the interactive behavior event to be mined. When the number of operation response image information is two, although the scene push involvement weight is low, the system still determines the basic push decision feature as the target push decision feature of the corresponding interactive behavior event to be mined. This is because in this case, the number of operation response image information is relatively small and may not be sufficient to provide sufficient feature information for a more accurate match, so the system chooses to retain the basic push decision feature as the target push decision feature.

[0083] Finally, when the scene push involvement weight of the basic push decision feature is less than the preset push involvement weight, and the number of operation response image information contained in the corresponding interactive behavior event to be mined is at least three, the system will adopt a more complex strategy to determine the target push decision feature. In this case, the system will adjust the associated image information set of the interactive behavior event to be mined to obtain an associated image adjustment set. This adjustment process may involve operations such as screening, optimization, and reorganization of the associated image information set to ensure that the adjusted associated image information set can more accurately reflect the characteristics of the interactive behavior event to be mined. Then, the system will determine the target push decision features of the corresponding interactive behavior event to be mined based on the associated image adjustment set. This strategy ensures that in complex situations, the system can still accurately determine the target push decision features, providing strong support for the subsequent push strategy formulation.

[0084] In practical applications, the process of determining the target push decision features requires comprehensive consideration of multiple factors and the selection and optimization of methods. For example, factors such as user historical behavior data, user portraits, and business scenarios can be considered to more accurately determine the target push decision features. At the same time, advanced technologies such as machine learning and deep learning can also be used to improve the accuracy and efficiency of feature matching and weight judgment. In addition, in order to ensure the accuracy and effectiveness of the target push decision features, the determined features need to be continuously verified and optimized.

[0085] In summary, the present invention obtains the scene push involvement weights of the basic push decision features, and determines the technical solution of the target push decision features of each interactive behavior event to be mined based on the comparison results of the weights and preset values ​​and the number of operation response image information, thereby providing strong data support and decision-making basis for enterprise operation management.

[0086] In some alternative embodiments, the associated image information set of the interactive behavior event to be mined is adjusted to obtain an associated image adjustment set, including: obtaining multiple initial associated image information sets corresponding to the interactive behavior event to be mined, and the number of operation response image information in each of the initial associated image information sets; and determining each of the initial associated image information sets whose number is greater than the set adjustment number as the associated image adjustment set of the interactive behavior event to be mined.

[0087] For example, the system needs to obtain multiple initial associated image information sets corresponding to the interactive behavior events to be mined. These initial associated image information sets are screened from a large amount of image information during the early processing of the system based on the relevance and feature matching degree of the interactive behavior events to be mined. They contain various image information closely related to the interactive behavior events to be mined, and are an important basis for the subsequent push strategy formulation.

[0088] Secondly, the system needs to obtain the number of operation response image information in each initial associated image information set. Operation response image information refers to the operation response generated by the user to specific image information in the interactive behavior event, such as click, browse, favorite, etc. The number of these operation response image information reflects the user's attention to the image information and interest preference, and is an important basis for subsequent adjustment of the associated image information set.

[0089] Then, the system will filter and adjust the initial associated image information set according to the number of operation response image information. Specifically, the system will set an adjustment number, which is determined based on the actual business scenario and needs, to determine whether the initial associated image information set needs to be adjusted. Then, the system will traverse all initial associated image information sets, and for the initial associated image information sets in which the number of operation response image information is greater than the set adjustment number, the system will determine them as the associated image adjustment sets for the interactive behavior events to be mined.

[0090] The significance of this screening and adjustment process is that by retaining image information with high user attention and strong interest preference, and removing image information with low user attention and weak interest preference, the system can more accurately grasp the user's interests and needs, and provide a more accurate and effective image information set for the subsequent push strategy formulation. At the same time, this can also reduce the amount of data to be processed later, and improve the system's processing efficiency and response speed.

[0091] In practical applications, the process of determining the associated image adjustment set needs to comprehensively consider the selection and optimization of multiple factors and methods. For example, factors such as user historical behavior data, user portraits, and business scenarios can be considered to more accurately determine the associated image adjustment set. At the same time, advanced technologies such as machine learning and deep learning can also be used to improve the accuracy and efficiency of image information screening and processing. In addition, in order to ensure the accuracy and effectiveness of the associated image adjustment set, the determined adjustment set needs to be continuously verified and optimized.

[0092] Furthermore, the determination of the associated image adjustment set can also be carried out in combination with other features and information. For example, factors such as the content, quality, and source of the image information can be considered to more comprehensively evaluate the value and importance of the image information. At the same time, data such as user feedback and push effects can also be used to continuously optimize and adjust the strategy and method for determining the associated image adjustment set.

[0093] In summary, the present invention obtains multiple initial associated image information sets corresponding to the interactive behavior events to be mined, as well as the number of operation response image information in each initial associated image information set, and determines the technical solution of the associated image adjustment set based on the comparison result of the number and the set adjustment number, thereby providing strong data support and decision-making basis for enterprise operation management.

[0094] Under some optional design ideas, the scene push involvement weights of each of the basic push decision features are obtained, including: according to each of the basic push decision features, push analysis is performed on the corresponding interactive behavior events to be mined, and push matching viewpoints corresponding to each of the basic push decision features are obtained; the ideal matching viewpoints of each of the interactive behavior events to be mined are obtained, and each of the ideal matching viewpoints is compared and analyzed with the corresponding push matching viewpoints, and the commonality evaluation coefficients corresponding to each of the basic push decision features are obtained; each of the commonality evaluation coefficients is determined as the scene push involvement weights of the corresponding basic push decision features.

[0095] It can be understood that through refined push analysis and comparison, accurate scenario push involvement weights can be assigned to each basic push decision feature, thereby providing solid data support for subsequent push strategy formulation.

[0096] First, the system will perform push analysis on the corresponding interactive behavior events to be mined based on each basic push decision feature. This analysis process involves an in-depth exploration of the matching degree, correlation, and potential impact between the basic push decision features and the interactive behavior events to be mined. Through push analysis, the system can obtain push matching opinions corresponding to each basic push decision feature. These push matching opinions reflect the push adaptability and effect expectations of the basic push decision features for the interactive behavior events to be mined in specific scenarios.

[0097] Secondly, the system needs to obtain the ideal matching viewpoint for each interactive behavior event to be mined. The ideal matching viewpoint refers to the best push effect or state that the interactive behavior event to be mined is expected to achieve under the enterprise operation and management goals. The acquisition of this viewpoint may involve comprehensive consideration and analysis of multiple factors such as market trends, user behavior, and business goals.

[0098] Then, the system compares and analyzes each ideal match view with the corresponding push match view. This comparison and analysis process aims to evaluate the gap or consistency between the basic push decision features in actual push and the ideal state. Through comparison and analysis, the system can obtain the commonality evaluation coefficient corresponding to each basic push decision feature. The commonality evaluation coefficient is a quantitative indicator used to characterize the degree of commonality and similarity between the push match view of the basic push decision feature and the ideal match view.

[0099] Finally, the system determines each common evaluation coefficient as the scene push involvement weight of the corresponding basic push decision feature. The scene push involvement weight is a comprehensive evaluation indicator that reflects the push involvement and importance of the basic push decision feature in a specific scene for the interactive behavior event to be mined. By converting the common evaluation coefficient into the scene push involvement weight, the system can provide accurate and quantitative decision-making basis for the subsequent push strategy formulation.

[0100] In actual applications, the calculation of the weight involved in scene push may involve complex numerical formulas and algorithm models. For example, weighted average method, fuzzy comprehensive evaluation method, neural network model and other methods can be used to calculate the commonality evaluation coefficient and scene push weight involved. The selection and optimization of these methods need to be determined according to the actual business scenarios and needs to ensure the accuracy and effectiveness of the calculation results.

[0101] Furthermore, in order to improve the accuracy and reliability of the scene push weights, it is also possible to consider introducing more data and information for auxiliary calculations. For example, historical push data, user feedback data, business operation data, etc. can be used to correct and optimize the commonality evaluation coefficient and scene push weights. At the same time, methods such as cross-validation and model evaluation can also be used to verify and optimize the effectiveness and accuracy of the calculation model.

[0102] Specifically, according to each of the basic push decision features, push analysis is performed on the corresponding interactive behavior events to be mined, and push matching opinions corresponding to each of the basic push decision features are obtained. This process can be achieved by calculating the similarity or correlation between the basic push decision features and the interactive behavior events to be mined:

[0103] Push matching view = f (basic push decision features, interactive behavior events to be mined)

[0104] Among them, f represents the calculation function of similarity or correlation, and an appropriate algorithm can be selected according to the specific application scenario, such as cosine similarity, Pearson correlation coefficient, etc.

[0105] Secondly, the ideal matching viewpoints of each interactive behavior event to be mined are obtained, and each ideal matching viewpoint is compared and analyzed with the corresponding push matching viewpoint to obtain the commonality evaluation coefficient corresponding to each basic push decision feature. This can be achieved by calculating the distance or difference between the push matching viewpoint and the ideal matching viewpoint:

[0106] Commonality evaluation coefficient = g (push matching viewpoint, ideal matching viewpoint)

[0107] Among them, g represents the calculation function of distance or difference, and an appropriate algorithm can also be selected according to the specific application scenario, such as Euclidean distance, Manhattan distance, etc. The smaller the commonality evaluation coefficient, the closer the pushed matching view is to the ideal matching view, and the better the push effect of the basic push decision feature.

[0108] Finally, each commonality evaluation coefficient is determined as the scene push involvement weight of the corresponding basic push decision feature. The specific value of the scene push involvement weight can be obtained by normalizing the commonality evaluation coefficient:

[0109] Scenario push involvement weight = h (commonality evaluation coefficient)

[0110] Among them, h represents a normalized function, which can convert the commonality evaluation coefficient into a value between 0 and 1, indicating the degree of involvement of the basic push decision feature in the scene push. The larger the scene push involvement weight, the greater the influence of the basic push decision feature in the scene push.

[0111] In some alternative embodiments, determining the target push decision feature of the corresponding interactive behavior event to be mined based on the associated image adjustment set includes: generating the push demand relationship adjustment topology of the corresponding interactive behavior event to be mined based on the associated image adjustment set; performing push decision feature mining on the push demand relationship adjustment topology to obtain the push decision optimization feature corresponding to the interactive behavior event to be mined; when the scene push involvement weight of the push decision optimization feature is not less than the preset push involvement weight, determining the push decision optimization feature as the target push decision feature of the corresponding interactive behavior event to be mined; when the scene push involvement weight of the push decision optimization feature is less than the preset push involvement weight, performing at least one round of adjustment on the associated image adjustment set again to obtain the target associated image information set corresponding to each round of adjustment, and generating the target push decision feature based on the target associated image information set, until the scene push involvement weight of the determined target push decision feature is greater than the preset push involvement weight, determining the target push decision feature with the scene push involvement weight greater than the preset push involvement weight as the target push decision feature of the corresponding interactive behavior event to be mined.

[0112] In a further refined technical embodiment, for the enterprise operation management application scenario, this embodiment deeply analyzes the image information in the associated image adjustment set, and mines out the push decision features that can reflect the user's interactive behavior patterns, thereby providing data support and decision-making basis for the enterprise to formulate accurate push strategies.

[0113] First, the system will build an adjustment topology that reflects the push demand relationship of the interactive behavior event to be mined based on the associated image adjustment set. This process involves in-depth analysis of each image in the associated image adjustment set to extract numerical features that can represent multiple dimensions such as image content, quality, and user attention. For example, for a product image, the intelligent push system can extract a feature vector containing the following dimensions: [image brightness (value range 0-255), image contrast (value range 0-1), color richness (value range 0-100), number of user clicks (value range 0-N), user viewing time (value range 0-T seconds), number of user shares (value range 0-M)]. These feature vectors not only contain the attribute information of the image itself, but also imply the interactive behavior pattern between the user and the image.

[0114] Next, the system will further mine the push decision optimization features. This process aims to extract the push strategy, push timing, push content and other decision features that are closely related to the interactive behavior events to be mined from the complex push demand relationship adjustment topology. To achieve this goal, the system will conduct in-depth analysis of the nodes, edges, and subgraphs in the topology structure, and calculate the numerical indicators such as similarity and correlation strength between them. For example, for the push strategy, the intelligent push system can calculate a feature vector containing the following dimensions based on historical data and business rules: [strategy type (such as personalized recommendation, popular recommendation, etc., represented by one-hot encoding), strategy execution frequency (numerical range 0-F times / day), strategy execution time (such as morning, afternoon, evening, etc., represented by one-hot encoding), strategy execution effect (such as click-through rate, conversion rate, etc., numerical range 0-1)]. These push decision optimization features are represented in the form of numerical vectors, each of which contains numerical information of multiple dimensions related to the push decision.

[0115] Then, the system will evaluate the scenario push involvement weights of these push decision optimization features. The scenario push involvement weight is a quantitative indicator used to evaluate the push involvement and importance of push decision features in mining interactive behavior events in specific scenarios. Specifically, the intelligent push system can set a weight for the numerical features of each dimension based on historical data and business rules, and then multiply the values ​​of each dimension in the push decision optimization feature by the corresponding weight, and add the results to obtain a comprehensive scenario push involvement weight. For example, if the weight of the strategy execution frequency is 0.3, the weight of the strategy execution time is 0.2, and the weight of the strategy execution effect is 0.5, then for a certain push decision optimization feature, its scenario push involvement weight can be calculated as: 0.3 strategy execution frequency + 0.2 strategy execution time (value after one-hot encoding) + 0.5*strategy execution effect.

[0116] Next, the system will determine whether the scene push involvement weight of the push decision optimization feature meets the preset requirements. If the scene push involvement weight of the push decision optimization feature is not less than the preset push involvement weight threshold (such as 0.6), then the system will determine these features as target push decision features for the interactive behavior events to be mined, and formulate corresponding push strategies accordingly. For example, if the scene push involvement weight of a push decision optimization feature is 0.7, which is higher than the preset threshold of 0.6, it means that the feature plays an important role in the push process of the interactive behavior events to be mined. The system can use this feature as a target push decision feature and formulate corresponding push strategies, such as increasing the push frequency, optimizing the push time, and adjusting the push content.

[0117] However, in actual applications, due to the complexity and variability of enterprise operation and management scenarios, the scene push involvement weights of push decision optimization features may not always meet the preset requirements. In this case, the system will make at least one round of adjustments to the associated image adjustment set. This adjustment process may involve re-screening, processing or optimization of image information, aiming to obtain an image information set that better meets push requirements. For example, the system can screen and sort images in the associated image adjustment set based on user feedback and business goals, and select images that are more attractive to users and more closely related to user interaction behaviors. By adjusting the associated image adjustment set, the system hopes to generate more effective push decision features and improve the scene push involvement weights.

[0118] After each round of adjustment, the system will regenerate the push decision feature based on the adjusted target-related image information set, and evaluate its scene push involvement weight. This process will be iterated continuously until the scene push involvement weight of the generated target push decision feature is greater than the preset push involvement weight. At this time, the system will determine this target push decision feature as the final target push decision feature of the interactive behavior event to be mined, and formulate corresponding push strategies accordingly. For example, after multiple rounds of adjustment and optimization, the system may obtain a target push decision feature with a scene push involvement weight of 0.8, which indicates that in specific scenarios (such as during the user's active time period), the use of a certain push strategy (such as a personalized recommendation strategy) can significantly improve the user's interactive behavior (such as click-through rate, conversion rate, etc.). Therefore, the system can use this feature as the final target push decision feature and formulate corresponding push strategies accordingly.

[0119] In summary, the present invention achieves accurate determination of the target push decision features for the interactive behavior events to be mined by building a push demand relationship adjustment topology, mining push decision optimization features, evaluating the push weights involved in the scenarios, and performing multiple rounds of adjustment and optimization. This technical solution not only provides powerful decision support for enterprise operation management, but also helps to formulate more accurate push strategies and improve user experience and operation results.

[0120] In the specific implementation process, enterprises can flexibly adjust the construction method of the associated image adjustment set, the mining algorithm of the push decision features, and the calculation model of the scene push weight according to their own business needs and operation scenarios to achieve the best push effect and business value. At the same time, through the introduction of multiple rounds of adjustment and optimization processes, the present invention can also continuously adapt to and respond to the complex and changing situations in the operation and management of enterprises, ensuring the continued effectiveness and competitiveness of the push strategy.

[0121] In some examples, the page user interaction data also includes at least one historical interactive behavior event of the initially screened potential push target, and the historical interactive behavior event includes an operation response image information. After obtaining the page user interaction data of the initially screened potential push target, the method further includes: obtaining the push demand relationship topology of each of the historical interactive behavior events; performing push decision feature mining on the push demand relationship topology of each of the historical interactive behavior events to obtain the target push decision feature corresponding to each of the historical interactive behavior events. Furthermore, the generation of the intelligent push strategy for the initially screened potential push target based on the target push decision feature of each of the interactive behavior events to be mined includes: utilizing the target push decision feature of each of the interactive behavior events to be mined, and the target push decision feature corresponding to each of the historical interactive behavior events, to generate the intelligent push strategy for the initially screened potential push target.

[0122] In the actual application scenario of enterprise operation management, in order to formulate push strategies more accurately, the present invention proposes a specific implementation method for determining the target push decision features of the interactive behavior events to be mined based on the associated image adjustment set. This method not only considers the immediate interactive behavior in the page user interaction data, but also deeply mines the user's historical interactive behavior events, so as to more comprehensively understand the user's interaction mode and preference.

[0123] First, after obtaining the page user interaction data for the initial screening of potential push targets, the present invention further obtains the push demand relationship topology of each historical interactive behavior event. This step is critical because it allows the system to understand how users have interacted with different elements on the page in the past and what the push demand behind these interactions is. For example, if a user has frequently clicked on images related to a certain product in the past and made a purchase, the system can construct a push demand relationship topology that includes "click on the image-view product details-purchase the product."

[0124] Next, the present invention mines the push decision features of the push demand relationship topology of each historical interactive behavior event to obtain the target push decision features corresponding to each historical interactive behavior event. This process involves an in-depth analysis of the topological structure to extract features that can reflect the user's interactive behavior patterns and preferences. For example, for the above-mentioned "click image-view product details-purchase product" topology, the system may mine the following target push decision features: [User preference type: product details viewing and purchase; interactive behavior pattern: click the image first and then view details; push timing: user active time period; push content: product discount information and user evaluation]. These features are represented in the form of numerical vectors, which provide strong data support for the subsequent push strategy formulation.

[0125] Furthermore, when generating an intelligent push strategy for initially screening potential push targets, the present invention not only utilizes the target push decision features of the interactive behavior events to be mined, but also fully considers the target push decision features corresponding to each historical interactive behavior event. This step ensures the comprehensiveness and accuracy of the push strategy. For example, for a potential push target that has been initially screened, the system may comprehensively formulate an intelligent push strategy containing the following elements based on its immediate page browsing behavior and past purchase history: [Push type: personalized product recommendations; Push timing: user's active time period at night; Push content: user's preferred product categories and latest discount information; Push method: push through the user's commonly used interactive channels].

[0126] In specific implementation, the present invention achieves a comprehensive understanding and accurate characterization of user interactive behaviors by constructing a complex push demand relationship topology and mining rich push decision features. These features not only include the user's immediate interactive behavior patterns, but also cover the user's past historical interactive behaviors and preferences. For example, for a user of an e-commerce platform, the system may mine the following target push decision features: [User ID: XYZ123; Historical purchase frequency: 2 times a month; Average purchase amount: 200-300 yuan; Preferred product categories: electronic products and books; Active time period: 8pm to 10pm; Common interaction channels: APP push and email]. These features provide a detailed data basis for formulating personalized push strategies.

[0127] In addition, the present invention continuously improves the accuracy and effectiveness of push decision features through multiple rounds of adjustment and optimization processes. For example, the system may update and optimize the associated image adjustment set based on user feedback and behavior changes to better reflect the user's current needs and preferences. At the same time, the system will also regularly re-explore and evaluate the push decision features to ensure that they can continue to effectively support the formulation of push strategies.

[0128] In summary, the present invention constructs a comprehensive push demand relationship topology by deeply mining the user's historical interactive behavior events and real-time page interaction data, and mines rich target push decision-making features. These features provide strong data support and decision-making basis for formulating accurate and personalized intelligent push strategies. In the actual application of enterprise operation management, this technical solution helps to improve user experience, enhance user stickiness, and ultimately achieve business growth and market competitiveness. At the same time, through continuous adjustment and optimization processes, the present invention can also adapt to and respond to complex and changeable situations in enterprise operation management, ensuring the continued effectiveness and competitiveness of push strategies.

[0129] In some independent embodiments, the use of the target push decision features of each of the interactive behavior events to be mined and the target push decision features corresponding to each of the historical interactive behavior events to generate the intelligent push strategy for initially screening potential push targets includes: obtaining an interactive behavior knowledge relationship spectrum corresponding to the page user interaction data, the interactive behavior knowledge relationship spectrum including multiple knowledge entities, the knowledge entities in the interactive behavior knowledge relationship spectrum corresponding one-to-one to the interactive behavior events contained in the page user interaction data; performing knowledge relationship chain walking on the interactive behavior knowledge relationship spectrum to obtain at least one knowledge relationship chain of the interactive behavior knowledge relationship spectrum, the head knowledge entity of the interactive behavior knowledge relationship spectrum being the same as the head knowledge entity of each of the knowledge relationship chains, and the tail knowledge entity of the interactive behavior knowledge relationship spectrum being the same as the tail knowledge entity of each of the knowledge relationship chains; using the target push decision features of each of the interactive behavior events to be mined and the target push decision features corresponding to each of the historical interactive behavior events, feature embedding is performed on the knowledge entities in each of the knowledge relationship chains to obtain the chain push decision features corresponding to each of the knowledge relationship chains; and generating the intelligent push strategy for initially screening potential push targets according to each of the chain push decision features.

[0130] In the actual application scenario of enterprise operation management, in order to more deeply understand user behavior and accurately formulate push strategies, the present invention proposes an embodiment that combines the interactive behavior knowledge relationship spectrum and the chain push decision feature. This embodiment aims to generate an intelligent push strategy for initially screening potential push targets by analyzing the user's page interaction data and historical interactive behavior events.

[0131] First, the present invention obtains the interactive behavior knowledge relationship spectrum corresponding to the page user interaction data. This relationship spectrum is a structured representation containing multiple knowledge entities, where each knowledge entity corresponds one-to-one to the interactive behavior event contained in the page user interaction data. For example, in the scenario of an e-commerce platform, the interactive behavior knowledge relationship spectrum may include knowledge entities such as "browse products", "click on product images", "view product details", "add to shopping cart" and "purchase products". These entities reflect the typical interactive behaviors of users on the page.

[0132] Secondly, the present invention walks the knowledge relationship chain of the interactive behavior knowledge relationship spectrum to discover the correlation and pattern therein. Through this step, at least one knowledge relationship chain can be obtained, and the head and tail knowledge entities of these chains are respectively the same as the head and tail entities of the interactive behavior knowledge relationship spectrum. Take "browse products-click on product images-view product details-add to shopping cart-buy products" as an example. This is a typical knowledge relationship chain, which describes the complete process from browsing to purchasing of users on the e-commerce platform.

[0133] Then, the present invention uses the target push decision features of the interactive behavior events to be mined, as well as the target push decision features corresponding to the historical interactive behavior events, to embed features into the knowledge entities in each knowledge relationship chain. The purpose of this step is to combine the push decision features with the knowledge entities to form richer and more specific chain push decision features. For example, for the knowledge entity "click on the product image", the following features may be embedded: [User ID: XYZ123, click frequency: 5 times a day, click to view details conversion rate: 30%, preferred product category: sports shoes]. These features provide detailed data support for the subsequent push strategy formulation.

[0134] In specific implementation, the generation process of chain push decision features involves complex feature embedding technology. For example, for each knowledge entity in the knowledge relationship chain, the system may generate a feature vector containing multiple dimensions based on the user's historical behavior and instant interaction data. These dimensions may include the user's click frequency, browsing time, purchase history, preference category, etc. By associating these feature vectors with knowledge entities, the system can more comprehensively understand the user's needs and preferences under specific interactive behaviors.

[0135] Furthermore, the present invention generates an intelligent push strategy for initially screening potential push targets based on the characteristics of each chain push decision. This step is to convert all the previous analysis and mining results into a specific push strategy. For example, for a potential push target that has been initially screened (such as user XYZ123), the system may formulate an intelligent push strategy containing the following elements based on its chain push decision characteristics: [Push type: personalized product recommendation; Push timing: user active time period (2pm to 4pm); Push content: user preferred sports shoe categories and latest discount information; Push method: through user's commonly used interaction channels (such as APP push and email); Push expected effect: expected click-through rate to increase by 25%, add-to-cart rate to increase by 20%, purchase conversion rate to increase by 15%].

[0136] By introducing the interactive behavior knowledge relationship spectrum and chain push decision-making features, the present invention can more deeply understand the user's interactive behavior patterns and preferences, and formulate more accurate and personalized push strategies accordingly. This method not only takes into account the user's immediate interactive behavior, but also fully mines the user's historical interactive behavior data, thereby achieving a more comprehensive grasp of user needs.

[0137] In addition, in actual applications, enterprises can customize the specific content and form of interactive behavior knowledge relationship spectrum and chain push decision features according to their own business needs and data conditions. For example, for social media platforms, the interactive behavior knowledge relationship spectrum may include knowledge entities such as "browsing posts", "liking", "commenting" and "sharing"; while for online education platforms, it may include entities such as "watching course videos", "downloading course materials", "participating in discussions" and "completing homework".

[0138] At the same time, the present invention also provides a wealth of data analysis and mining tools to help enterprises extract valuable information and patterns from massive amounts of user interaction data. These tools not only support the construction and update of interactive behavior knowledge relationship spectra, but also enable real-time monitoring and optimization of chain push decision features to ensure the continued effectiveness and competitiveness of push strategies.

[0139] In summary, the present invention provides enterprises with a new, more accurate and personalized intelligent push strategy formulation method by combining the interactive behavior knowledge relationship spectrum and chain push decision-making characteristics. This method not only improves user experience and stickiness, but also brings higher business growth and market competitiveness to enterprises.

[0140] It is worth mentioning that the knowledge relationship chain walk is essentially a depth-first or breadth-first traversal process of the interactive behavior knowledge relationship spectrum. In this process, the system starts from the head knowledge entity of the interactive behavior knowledge relationship spectrum and gradually walks to the tail knowledge entity along the relationship chain between knowledge entities. During the walk, the system records and analyzes the knowledge entities passed through at each step and the relationship between them, thereby revealing the potential patterns and correlations between user interactive behaviors.

[0141] Taking the user purchase process on an e-commerce platform as an example, the interactive behavior knowledge relationship spectrum may include knowledge entities such as "browse products", "click on product images", "view product details", "add to shopping cart" and "purchase products". When walking the knowledge relationship chain, the system will start from the head knowledge entity "browse products" and gradually walk to the tail knowledge entity "purchase products". In this process, the system will record the knowledge entities passed through at each step, such as "click on product images" and "view product details", and analyze the relationship between them, such as "what percentage of users will click on product images and view product details after browsing products".

[0142] In order to more specifically illustrate the process of knowledge relationship chain wandering, the intelligent push system can introduce some specific numerical examples. For example, in the user behavior data of the e-commerce platform, 1,000 users clicked on the product image after browsing the product, 800 of them further viewed the product details, and finally 500 users added the product to the shopping cart, and 200 users completed the purchase. By wandering the knowledge relationship chain, the intelligent push system can obtain the following relationship chain and numerical features:

[0143] “Browse products” -> “Click on product image”: conversion rate 80% (800 / 1000);

[0144] “Click on product image” -> “View product details”: conversion rate is 100% (800 / 800), because all users who click on the product image view the product details;

[0145] “View product details” -> “Add to cart”: conversion rate 62.5% (500 / 800);

[0146] “Add to cart” -> “Purchase product”: conversion rate 40% (200 / 500).

[0147] Through these numerical features, the smart push system can clearly see the conversion status of users at each step in the purchase process, thereby revealing the potential patterns and correlations between user behaviors. For example, the smart push system can find that although most users will click on the product image and view the details after browsing the product, only about 62.5% of users will choose to add the product to the shopping cart, and only about 40% of users will eventually complete the purchase.

[0148] This information is crucial for the subsequent push strategy formulation. For example, the smart push system can push more attractive product detail pages to users who clicked on the product image but did not view the product details, in order to increase their interest in learning more about the product; for users who added the product to the shopping cart but did not complete the purchase, the smart push system can push some promotional information or coupons to stimulate their desire to buy.

[0149] It can be seen that the knowledge relationship chain walking, as the core invention of the present invention, provides valuable information and insights for the subsequent formulation of precise push strategies by deeply mining and analyzing the correlation between user interactive behaviors.

[0150] In some other independent embodiments, the target push decision feature of each interactive behavior event to be mined and the target push decision feature corresponding to each historical interactive behavior event are used to embed features into the knowledge entities in each knowledge relationship chain to obtain the chain push decision feature corresponding to each knowledge relationship chain, including: implementing the following steps for the knowledge entities in each knowledge relationship chain: when the interactive behavior event corresponding to the knowledge entity is the interactive behavior event to be mined, labeling the knowledge entity with the corresponding target push decision feature of the interactive behavior event to be mined; when the interactive behavior event corresponding to the knowledge entity is the historical interactive behavior event, labeling the knowledge entity with the corresponding target push decision feature of the historical interactive behavior event; when all knowledge entities in the knowledge relationship chain have completed feature labeling, obtaining the chain push decision feature corresponding to the knowledge relationship chain.

[0151] It can be understood that in order to formulate a push strategy more effectively, the present invention proposes an embodiment that combines knowledge relationship chains and chain push decision features. The core of this embodiment is to embed features of knowledge entities in each knowledge relationship chain, thereby obtaining chain push decision features corresponding to each knowledge relationship chain, providing strong support for accurate push.

[0152] Among them, knowledge entities are the basic units in the interactive behavior knowledge relationship spectrum, and they correspond one-to-one to the user's interactive behavior events, such as "browsing products", "clicking product images", etc. The chain push decision feature is obtained by embedding the features of knowledge entities. It reflects the user's needs and preferences under a specific interactive behavior chain and is an important basis for formulating push strategies.

[0153] In specific implementation, for knowledge entities in each knowledge relationship chain, the intelligent push system will adopt different labeling strategies according to their actual situation. When the interactive behavior event corresponding to the knowledge entity is an interactive behavior event to be mined, the intelligent push system will use the target push decision features of the corresponding interactive behavior event to be mined to label the knowledge entity. These features may include the user's click frequency, browsing time, purchase intention, etc., which can provide the intelligent push system with preliminary demand and preference information about the user under specific interactive behavior.

[0154] Taking the e-commerce platform as an example, for example, "clicking on the product image" is an interactive behavior event to be mined, and the intelligent push system may use the following target push decision features to label it: [User ID: XYZ123, click frequency: 5 times a day, average browsing time: 30 seconds, purchase intention: high]. These features can reveal the needs and preferences of user XYZ123 in the behavior of clicking on the product image to the intelligent push system, providing strong support for the subsequent push strategy formulation of the intelligent push system.

[0155] When the interactive behavior event corresponding to a knowledge entity is a historical interactive behavior event, the intelligent push system will use the target push decision features of the corresponding historical interactive behavior event to mark the knowledge entity. These features also include the user's click frequency, browsing time, purchase history, etc., but they are based on the user's past behavior data, so they can provide the intelligent push system with more comprehensive and in-depth user behavior pattern information.

[0156] Continuing with the example of e-commerce platforms, for example, if "viewing product details" is a historical interactive behavior event, the smart push system may use the following target push decision features to label it: [User ID: XYZ123, click frequency in the past 30 days: 100 times, average browsing time: 60 seconds, purchase history: 2 pairs of sports shoes, 1 pair of casual shoes]. These features can reveal the long-term needs and preferences of user XYZ123 in the behavior of viewing product details to the smart push system, providing strong support for the smart push system to formulate more accurate push strategies.

[0157] When all knowledge entities in the knowledge relationship chain have completed feature annotation, the intelligent push system can obtain the chain push decision feature corresponding to the knowledge relationship chain. This feature is a high-dimensional vector that integrates all the needs and preferences of users under a specific interactive behavior chain. It can provide the intelligent push system with more comprehensive and in-depth user behavior pattern information, and provide strong support for the intelligent push system to formulate more accurate and personalized push strategies.

[0158] Taking the purchase process of an e-commerce platform as an example, the intelligent push system obtains a knowledge relationship chain containing knowledge entities such as "browse products", "click on product images", "view product details", "add to shopping cart" and "purchase products". By embedding features of these knowledge entities, the intelligent push system can obtain the following chain push decision features: [User ID: XYZ123, browse product frequency: 10 times a day, click product image frequency: 5 times a day, average browsing time: 30 seconds, view product details frequency: 100 times in the past 30 days, average browsing time: 60 seconds, purchase history: 2 pairs of sports shoes, 1 pair of casual shoes, purchase intention: high]. This feature vector provides the intelligent push system with comprehensive demand and preference information about user XYZ123 in the purchase process, and provides strong support for the intelligent push system to formulate more accurate and personalized push strategies.

[0159] In practical applications, the generation process of chain push decision features involves complex feature embedding technology. For example, for each knowledge entity in the knowledge relationship chain, the system may generate a feature vector containing multiple dimensions based on the user's historical behavior and instant interaction data. These dimensions may include the user's click frequency, browsing time, purchase history, preference category, etc. By associating these feature vectors with knowledge entities, the system can more comprehensively understand the user's needs and preferences under specific interactive behaviors.

[0160] Furthermore, the present invention will generate an intelligent push strategy for the initial screening of potential push targets based on the characteristics of each chain push decision. This step is to convert all the previous analysis and mining results into a specific push strategy. For example, for a potential push target that has been initially screened (such as user XYZ123), the system may formulate an intelligent push strategy containing the following elements based on its chain push decision characteristics: [Push type: personalized product recommendation; Push timing: user active time period (2pm to 4pm); Push content: user preferred sports shoe categories and latest discount information; Push method: through user's commonly used interaction channels (such as APP push and email); Push expected effect: expected click-through rate to increase by 25%, add-to-cart rate to increase by 20%, purchase conversion rate to increase by 15%].

[0161] By introducing knowledge relationship chains and chain push decision features, the present invention can more deeply understand the user's interactive behavior patterns and preferences, and formulate more accurate and personalized push strategies accordingly. This method not only takes into account the user's immediate interactive behavior, but also fully mines the user's historical interactive behavior data, thereby achieving a more comprehensive grasp of user needs.

[0162] At the same time, the present invention also provides a wealth of data analysis and mining tools to help enterprises extract valuable information and patterns from massive amounts of user interaction data. These tools not only support the construction and update of knowledge relationship chains, but also enable real-time monitoring and optimization of chain push decision features to ensure the continued effectiveness and competitiveness of push strategies.

[0163] It can be seen that by combining the knowledge relationship chain and the chain push decision-making characteristics, the present invention provides enterprises with a new, more accurate and personalized intelligent push strategy formulation method.

[0164] Furthermore, a readable storage medium is provided, on which a program is stored, and the program implements the above method when executed by a processor.

[0165] Those skilled in the art can clearly understand that, for the convenience and simplicity of description, the specific working process of the system and device described in the above exemplary embodiment can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here. In the several embodiments provided by the present invention, it should be understood that the disclosed system, device and method can be implemented in other ways. The device embodiments described above are merely schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some communication interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

Claims

1. A method for processing page user big data applied to intelligent push, characterized in that: The method is performed by an intelligent push system, and the method includes: Acquire page user interaction data of the preliminarily screened potential push targets, wherein the page user interaction data includes at least one to-be-mined interactive behavior event of the preliminarily screened potential push targets, and the to-be-mined interactive behavior event includes a plurality of operation response image information; Based on the page user interaction data, multiple operation response image information included in each of the interactive behavior events to be mined are associated to obtain at least one associated image information set corresponding to each of the interactive behavior events to be mined, and the associated image information set is used to characterize the user interest association result of at least two operation response image information of the interactive behavior event to be mined; Generate a push demand relationship topology of each interactive behavior event to be mined based on the associated image information set; wherein the push demand relationship topology is a graphical representation method for displaying the user's demand association and transfer path between different interactive behavior events; Performing push decision feature mining on the push demand relationship topology of each of the interactive behavior events to be mined, and obtaining basic push decision features corresponding to each of the interactive behavior events to be mined; Determine the target push decision feature of each of the interactive behavior events to be mined based on the basic push decision feature, and generate the intelligent push strategy for preliminarily screening potential push targets based on the target push decision feature of each of the interactive behavior events to be mined; Determining the target push decision feature of each interactive behavior event to be mined based on the basic push decision feature includes: Obtaining the scenario push involvement weight of each of the basic push decision features, wherein the scenario push involvement weight is used to characterize the correlation between the push matching viewpoint corresponding to the basic push decision feature and the ideal matching viewpoint of the corresponding interactive behavior event to be mined; When the scene push involvement weight of the basic push decision feature is not less than the preset push involvement weight, the basic push decision feature is determined as the target push decision feature of the corresponding interactive behavior event to be mined; When the scene push involvement weight of the basic push decision feature is less than the preset push involvement weight, and the number of the operation response image information included in the corresponding interactive behavior event to be mined is two, the basic push decision feature is determined as the target push decision feature of the corresponding interactive behavior event to be mined; When the scene push involvement weight of the basic push decision feature is less than the preset push involvement weight, and the number of the operation response image information included in the corresponding interactive behavior event to be mined is at least three, the associated image information set of the interactive behavior event to be mined is adjusted to obtain an associated image adjustment set, and the target push decision feature of the corresponding interactive behavior event to be mined is determined based on the associated image adjustment set; The step of adjusting the associated image information set of the interactive behavior event to be mined to obtain an adjusted associated image set includes: Acquire multiple initial associated image information sets corresponding to the interactive behavior events to be mined, and the number of operation response image information in each of the initial associated image information sets; Determine the initial associated image information sets whose number is greater than the set adjustment number as the associated image adjustment sets of the interactive behavior event to be mined; The step of obtaining the scene push-related weights of each of the basic push decision features includes: According to each of the basic push decision features, push analysis is performed on the corresponding interactive behavior events to be mined, and push matching opinions corresponding to each of the basic push decision features are obtained; Obtaining the ideal matching viewpoints of each interactive behavior event to be mined, comparing and analyzing each ideal matching viewpoint with the corresponding push matching viewpoint, and obtaining the commonality evaluation coefficient corresponding to each basic push decision feature; Determine each of the commonality evaluation coefficients as the scene push involvement weight of the corresponding basic push decision feature; Determining the target push decision feature of the corresponding interactive behavior event to be mined based on the associated image adjustment set includes: Generating a corresponding push demand relationship adjustment topology of the interactive behavior event to be mined according to the associated image adjustment set; Mining the push decision features of the push demand relationship adjustment topology to obtain the push decision optimization features corresponding to the interactive behavior events to be mined; When the scene push involvement weight of the push decision optimization feature is not less than the preset push involvement weight, the push decision optimization feature is determined as the target push decision feature of the corresponding interactive behavior event to be mined; When the scene push involvement weight of the push decision optimization feature is less than the preset push involvement weight, the associated image adjustment set is adjusted again for at least one round to obtain the target associated image information set corresponding to each round of adjustment, and a target push decision feature is generated based on the target associated image information set, until the scene push involvement weight of the target push decision feature is determined to be greater than the preset push involvement weight, the target push decision feature whose scene push involvement weight is greater than the preset push involvement weight is determined as the corresponding target push decision feature of the interactive behavior event to be mined; Among them, the ideal matching viewpoint refers to the optimal push effect or state that is expected to be achieved by the interactive behavior events to be mined under the enterprise operation and management goals; the push matching viewpoint reflects the push adaptability and effect expectations of the basic push decision-making characteristics for the interactive behavior events to be mined in the corresponding scenarios.

2. The method according to claim 1, characterized in that The step of associating a plurality of operation response image information included in each of the to-be-mined interactive behavior events based on the page user interaction data to obtain at least one associated image information set corresponding to each of the to-be-mined interactive behavior events, includes: For each interactive behavior event to be mined in the page user interaction data, the following steps are performed respectively: When the number of the operation response image information included in the interactive behavior event to be mined is two, associating the two operation response image information of the interactive behavior event to be mined to obtain the associated image information set; When the number of the operation response image information included in the interactive behavior event to be mined is at least three, for the at least three operation response image information included in the interactive behavior event to be mined, user interest association of at least two operation response image information is implemented multiple times to obtain multiple initial associated image information sets, and based on the multiple initial associated image information sets, the associated image information set of the interactive behavior event to be mined is determined.

3. The method according to claim 2, characterized in that The step of determining the associated image information set of the interactive behavior event to be mined based on the plurality of initial associated image information sets includes: Acquire the number of operation response image information in each of the initial associated image information sets; The initial associated image information sets whose number is greater than the set number are determined as the associated image information sets of the interactive behavior events to be mined.

4. The method according to claim 1, characterized in that The step of generating a push demand relationship topology of each interactive behavior event to be mined based on the associated image information set includes: The following steps are implemented for each interactive behavior event to be mined: For each of the associated image information sets corresponding to the interactive behavior events to be mined, respectively generating a push demand relationship topology corresponding to each of the associated image information sets; According to the push demand relationship topology corresponding to each of the associated image information sets, a push demand relationship topology of the interactive behavior event to be mined is generated.

5. The method according to claim 4, characterized in that The step of generating a push demand relationship topology corresponding to each of the associated image information sets corresponding to the interactive behavior event to be mined includes: For each of the associated image information sets corresponding to the interactive behavior events to be mined, obtaining push demand relationship topologies corresponding to each of the operation response image information used to create the associated image information sets; The push demand relationship topologies corresponding to the operation response image information are merged to obtain the push demand relationship topology of the associated image information set.

6. The method according to claim 4, characterized in that The generating of the push demand relationship topology of the interactive behavior event to be mined according to the push demand relationship topology corresponding to each of the associated image information sets includes: Acquire the number of relationship topologies of the push demand relationship topologies corresponding to the associated image information set; When the number of the relationship topologies is one, determining the push demand relationship topology corresponding to the associated image information set as the push demand relationship topology of the interactive behavior event to be mined; When there are multiple relationship topologies, the push demand relationship topologies corresponding to the associated image information sets are merged to obtain the push demand relationship topology of the interactive behavior event to be mined.

7. The method according to claim 1, characterized in that The page user interaction data further includes at least one historical interaction behavior event of the initially screened potential push target, and the historical interaction behavior event includes one operation response image information; After obtaining the page user interaction data for preliminary screening of potential push targets, the method further includes: obtaining the push demand relationship topology of each of the historical interactive behavior events; performing push decision feature mining on the push demand relationship topology of each of the historical interactive behavior events to obtain the target push decision feature corresponding to each of the historical interactive behavior events; The method of generating the intelligent push strategy for initially screening potential push targets based on the target push decision features of each of the interactive behavior events to be mined includes: utilizing the target push decision features of each of the interactive behavior events to be mined, and the target push decision features corresponding to each of the historical interactive behavior events, to generate the intelligent push strategy for initially screening potential push targets.

8. An intelligent push system, characterized in that: It comprises a processor and a memory; the processor and the memory are communicatively connected, and the processor is used to read a computer program from the memory and execute it to implement the method described in any one of claims 1 to 7.

9. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and the computer program implements the method described in any one of claims 1 to 7 when running.

Images