Printing personalized push method and system based on user behavior analysis

Abstract

The application provides a printing personalized pushing method and system based on user behavior analysis, the printing personalized pushing method comprises the following steps: collecting printing data of a user terminal device in real time and sending the printing data to a printing terminal; a printing cloud server of the printing terminal performs data entity extraction processing on the printing data, and obtains printing entity data corresponding to the printing data; the printing entity data is written to a knowledge graph corresponding to the user, and a personal work context knowledge graph which has been completed, constructed and deployed corresponding to the user is dynamically updated; the printing cloud server infers a printing strategy through the work context knowledge graph according to the printing entity data, and recommends the printing strategy to the user terminal; the printing cloud server waits for a strategy confirmation instruction sent by the user terminal, and when the printing cloud server receives the strategy confirmation instruction of the user, controls a printer to perform a printing operation according to the printing strategy confirmed by the user. The system comprises modules corresponding to the method steps.

Description

Technical Field

[0001] This invention proposes a method and system for personalized printing recommendations based on user behavior analysis, belonging to the field of personalized printing information technology. Background Technology

[0002] In modern office settings, printing technology, as an important means of materializing information, is increasingly widely used. Currently, the mainstream printing process typically relies on users manually setting printing parameters (such as paper type, number of copies, color mode, duplex printing mode, etc.) on the terminal device and generating a print job which is then sent to the printing terminal. The printing terminal simply acts as an execution device, completing the printing operation according to the received fixed parameters.

[0003] As user printing needs become more diverse and complex, existing printing technologies are increasingly revealing several limitations: First, they lack personalized adaptation capabilities. Different users have significant differences in printing habits and document types (such as contracts, presentations, and images), but current printing terminals cannot automatically match the optimal strategy based on user history and document attributes, requiring users to manually adjust each time, which is cumbersome and prone to errors. Second, they are disconnected from the user's work context. User printing behavior is often associated with specific work scenarios (such as project reports and client communication). Current technologies only process single printing tasks in isolation, failing to link printing data with the user's schedule, document library, collaborators, and other work context information. This results in the inability to dynamically optimize printing strategies based on the user's current work status; for example, it cannot automatically recommend suitable print formats for upcoming meetings. Third, resource utilization efficiency is low. Due to the lack of analysis and prediction of user behavior, current printing technologies often suffer from over-printing and paper waste due to inappropriate formats, which is inconsistent with the development trend of green office practices.

[0004] Therefore, there is an urgent need for a method that can combine user behavior analysis and work context information to achieve personalized printing strategy push, in order to solve the problems of cumbersome operation, poor adaptability, and waste of resources in the existing technology. Summary of the Invention

[0005] This invention provides a method and system for personalized printing push based on user behavior analysis, in order to solve the technical problems existing in the prior art. The technical solution adopted is as follows: A personalized printing push method based on user behavior analysis, the personalized printing push method includes: The system collects printing data from user terminal devices in real time and sends the printing data to the printing terminal; wherein the printing terminal includes a printing cloud server and a printer; The printing terminal's printing cloud server performs data entity extraction processing on the printing data to obtain the printing entity data corresponding to the printing data; The printed entity data is dynamically updated using the entity coefficients corresponding to the entity data to update the user's personal work context knowledge graph that has been built and deployed. The printing cloud server infers a printing strategy based on the printing entity data using a working context knowledge graph, and recommends the printing strategy to the user terminal. The print cloud server waits for a policy confirmation instruction sent by the user terminal. Once the print cloud server receives the user's policy confirmation instruction, it controls the printer to perform the printing operation according to the printing policy confirmed by the user.

[0006] Furthermore, the real-time acquisition of printing data from user terminal devices and the transmission of the printing data to the printing terminal include: Real-time monitoring of multiple data streams from user terminal devices, including system event streams, browser metadata streams, document metadata streams, and Hitachi event streams; The print data is filtered from the various data streams, and the print data is encapsulated to obtain the encapsulated data packet corresponding to the user terminal. The encapsulated data packet is sent to the printing terminal.

[0007] Furthermore, the printing terminal's printing cloud server performs data entity extraction processing on the printing data to obtain the printing entity data corresponding to the printing data, including: The printing cloud server parses the received encapsulated data packets to obtain the printing data; Scan the printed data to obtain the file name and file content data corresponding to the printed data; Retrieve key information data from the file name and file content data; wherein, the key information data includes basic information and key general terms, and the basic information includes, but is not limited to, username, project name and date, etc. The basic information is used as the first entity data, and the key general terms are used as the second entity data.

[0008] Furthermore, the printed entity data is dynamically updated using the entity coefficients corresponding to the entity data to update the user's corresponding personal work context knowledge graph that has been built and deployed, including: Retrieve the first entity data and the second entity data, and obtain the entity coefficients corresponding to the first entity data and the second entity data; Map the first entity data and the second entity data to the user's corresponding personal work context knowledge graph that has been built and deployed; Determine whether there is entity data in the personal work context knowledge graph that corresponds to the first entity data and the second entity data; When there is a combination of entity data in the personal work context knowledge graph that corresponds to the first entity data and the second entity data, the node corresponding to the entity data is marked. When there is no entity data combination in the personal work context knowledge graph that corresponds to the first entity data and the second entity data, then the entity coefficients corresponding to the first entity data and the second entity data are combined to create a node and obtain the dynamically updated personal work context knowledge graph corresponding to the user.

[0009] Furthermore, the entity coefficients corresponding to the first entity data are obtained in the following ways: Retrieve the first entity data and determine the number of data types present in the first entity data; Retrieve the preset basic weight value corresponding to each data type in the first entity data, wherein the basic weight value is set based on experience, and the weight value setting range is 0-1; The entity coefficient corresponding to the first entity data is obtained based on the basic weight value corresponding to the data type of the first entity data.

[0010] Furthermore, the entity coefficients corresponding to the second entity data are obtained in the following ways: Retrieve the second entity data and determine the number of data types present in the second entity data; Retrieve the preset basic weight value corresponding to each data type in the second entity data, wherein the basic weight value is set based on experience, and the weight value setting range is 0-1; Based on each data type corresponding to the first entity data and each data type in the second entity data, obtain the association strength between each data type in the second entity data and all data types in the first entity data; The entity coefficients corresponding to the second entity data are obtained by combining the preset basic weight values ​​corresponding to each data type in the second entity data with the correlation strength between each data type in the second entity data and all data types in the first entity data.

[0011] Furthermore, based on each data type corresponding to the first entity data and each data type in the second entity data, the association strength between each data type in the second entity data and all data types in the first entity data is obtained, including: Retrieve the user's historical print data; Extract the total number of times each data type appears in the user's historical print data from the second entity data; Extract the data types contained in the first entity data corresponding to each combination that appears in the second entity data each time a data type appears; The association strength between each data type in the second entity data and all data types in the first entity data is obtained by combining the total number of times each data type in the second entity data appears in the user's corresponding historical print data with the data types contained in the first entity data that appear in the corresponding combination each time each data type in the second entity data appears.

[0012] Furthermore, by combining the entity coefficients corresponding to the first and second entity data, nodes are created to obtain a dynamically updated personal work context knowledge graph for each user, including: Retrieve the combination of the first entity data and the second entity data corresponding to each node in the user's personal work context knowledge graph; The similarity between the combination of the first entity data and the second entity data corresponding to each node in the personal work context knowledge graph and the current combination of the first entity data and the second entity data is compared to obtain the similarity value between the combination of the first entity data and the second entity data corresponding to each node in the personal work context knowledge graph and the current combination of the first entity data and the second entity data. Retrieve the entity coefficients of the first and second entity data corresponding to each node in the user's personal work context knowledge graph. By combining the entity coefficients of the first entity data and the second entity data of each node in the knowledge graph of the user's personal work context with the similarity value of the combination of the first entity data and the second entity data of each node, the comprehensive strength value of each node and the current combination of the first entity data and the second entity data is obtained. The comprehensive strength value is compared with a preset strength threshold, and nodes with comprehensive strength values ​​not lower than the preset strength threshold are selected as associated nodes. A target node is created in the personal work context knowledge graph for the combination of the current first entity data and the second entity data, and a directed edge is established between the target node and the associated node; Retrieve the entity coefficients corresponding to the first and second entity data of the target node, and the entity coefficients corresponding to the first and second entity data of each associated node; Based on the entity coefficients corresponding to the first and second entity data of the target node and the entity coefficients corresponding to the first and second entity data of each associated node, set the edge weight values ​​corresponding to the directed edges established between the target node and the associated nodes.

[0013] Furthermore, the printing cloud server infers a printing strategy based on the printing entity data using a work context knowledge graph, and recommends the printing strategy to the user terminal, including: When there is no entity data combination in the personal work context knowledge graph that corresponds to the first entity data and the second entity data, locate the associated node associated with the target node; Retrieve from the database the printing strategies corresponding to the associated nodes related to the target node, as well as the edge weights between the target node and the associated nodes; By combining edge weights with a sorting strategy, the printing strategies corresponding to the associated nodes are sorted to obtain a set of printing strategy sequences. The printing strategies corresponding to the associated nodes are recommended to the user terminal according to the order of the printing strategies in the printing strategy sequence set. When there is a combination of entity data in the personal work context knowledge graph that corresponds to the first entity data and the second entity data, the node marked in the personal work context knowledge graph is searched. After identifying the marked nodes, the corresponding printing strategies for the marked nodes are retrieved from the database and recommended to the user terminal.

[0014] A personalized printing push system based on user behavior analysis, the personalized printing push system comprising: A data acquisition module is used to collect printing data from user terminal devices in real time and send the printing data to the printing terminal; wherein, the printing terminal includes a printing cloud server and a printer; The entity extraction module is used by the printing cloud server of the printing terminal to perform data entity extraction processing on the printing data to obtain the printing entity data corresponding to the printing data. The graph update module is used to dynamically update the user's personal work context knowledge graph that has been built and deployed by the user through the entity coefficients corresponding to the printed entity data. The inference module is used by the printing cloud server to infer the printing strategy based on the printing entity data through the working context knowledge graph, and recommend the printing strategy to the user terminal. The print execution module is used for the print cloud server to wait for the policy confirmation instruction sent by the user terminal. When the print cloud server receives the user's policy confirmation instruction, it controls the printer to perform the printing operation according to the print policy confirmed by the user.

[0015] Beneficial effects of this invention: This invention proposes a personalized printing recommendation method and system based on user behavior analysis. Through user behavior analysis and dynamic updates of a knowledge graph, printing strategies are tailored to users' long-established operating habits and preferences, reducing manual parameter settings, improving ease of operation, and lowering the rate of human error. Simultaneously, by integrating printing entity data into the user's personal work context knowledge graph, the printing strategy reasoning process fully considers the user's current work scenario, such as task progress and collaboration needs, effectively improving the matching degree between strategies and actual work requirements and enhancing the scenario adaptability of the printing service. Furthermore, optimizing printing strategies effectively reduces unnecessary paper and consumable waste, thereby lowering office costs. Moreover, the dynamic updates and reasoning capabilities based on the knowledge graph transform the printing terminal from a passive execution device into an intelligent service terminal with proactive decision-making capabilities. It can adaptively adjust the strategy recommendation logic according to changes in user behavior, effectively improving the system's self-optimization capabilities. By automatically recommending suitable printing strategies, the user's operational burden is reduced while ensuring that the printing results meet actual needs, enhancing user satisfaction and reliance on the printing service. Attached Figure Description

[0016] Figure 1 This is a flowchart of the method described in this invention; Figure 2 This is a system block diagram of the system described in this invention. Detailed Implementation

[0017] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0018] This invention proposes a personalized printing push method based on user behavior analysis, such as... Figure 1 As shown, the personalized printing push method includes: The system collects printing data from user terminal devices in real time and sends the printing data to the printing terminal; wherein the printing terminal includes a printing cloud server and a printer; The printing terminal's printing cloud server performs data entity extraction processing on the printing data to obtain the printing entity data corresponding to the printing data; The printed entity data is dynamically updated using the entity coefficients corresponding to the entity data to update the user's personal work context knowledge graph that has been built and deployed. The printing cloud server infers a printing strategy based on the printing entity data using a working context knowledge graph, and recommends the printing strategy to the user terminal. The print cloud server waits for a policy confirmation instruction sent by the user terminal. Once the print cloud server receives the user's policy confirmation instruction, it controls the printer to perform the printing operation according to the printing policy confirmed by the user.

[0019] The working principle of the above technical solution is as follows: The personalized printing push method in this embodiment collects printing data from the user terminal in real time and transmits it to the printing terminal. The printing terminal's printing cloud server extracts entities from the printing data to obtain printing entity data. Using this printing entity data and corresponding entity coefficients, the constructed user's personal work context knowledge graph is dynamically updated, so that the knowledge graph continuously reflects the relationship between user behavior characteristics and work scenarios. Based on the updated knowledge graph, the printing terminal's printing cloud server infers a printing strategy that adapts to the user's behavior habits and current work context and pushes it to the user terminal. After receiving the user's confirmation instruction, the printer is controlled to execute the printing operation according to the confirmed strategy. By deeply associating printing data with the user's work context, a closed-loop personalized push is achieved from data collection to strategy execution.

[0020] The effects of the above technical solution are as follows: By analyzing user behavior and dynamically updating the knowledge graph, printing strategies can be tailored to users' long-established operating habits and preferences, reducing manual parameter settings, effectively improving operational convenience, and lowering the rate of human error. Simultaneously, integrating printing entity data into the user's personal work context knowledge graph allows the printing strategy reasoning process to fully consider the user's current work scenario, such as task progress and collaboration needs, effectively improving the matching degree between the strategy and actual work requirements and enhancing the scenario adaptability of the printing service. On the other hand, optimizing printing strategies can effectively reduce unnecessary paper and consumable waste, thereby lowering office costs. Furthermore, based on the dynamic updating and reasoning capabilities of the knowledge graph, the printing terminal transforms from a passive execution device into an intelligent service terminal with proactive decision-making capabilities, adaptively adjusting the strategy recommendation logic according to changes in user behavior, effectively improving the system's self-optimization capabilities. By automatically recommending suitable printing strategies, the user's operational burden is reduced while ensuring that the printing results meet actual needs, enhancing user satisfaction and reliance on the printing service.

[0021] One embodiment of the present invention involves real-time acquisition of printing data from a user terminal device and transmission of the printing data to a printing terminal, including: Real-time monitoring of multiple data streams from user terminal devices, including system event streams, browser metadata streams, document metadata streams, and Hitachi event streams; The print data is filtered from the various data streams, and the print data is encapsulated to obtain the encapsulated data packet corresponding to the user terminal. The encapsulated data packet is sent to the printing terminal.

[0022] The working principle of the above technical solution is as follows: This embodiment achieves comprehensive capture of user printing-related behavioral data by real-time monitoring of multiple data streams, including system event streams, browser metadata streams, document metadata streams, and Hitachi event streams from the user terminal device; based on preset printing data feature rules, such as those including print instruction identifiers and document printing parameter information, it accurately filters data directly related to the printing operation from multiple data streams; the filtered printing data is standardized and encapsulated to form a structured encapsulated data packet; and the encapsulated data packet is sent to the printing terminal in real time through a secure transmission protocol, providing raw data support for subsequent printing entity extraction and personalized strategy reasoning. The above technical solution, with accurate filtering and standardized encapsulation as its core, achieves real-time processing of printing data throughout the entire chain from generation to transmission.

[0023] The above technical solution achieves the following effects: by covering multiple data streams including systems, browsers, documents, and Hitachi events, it breaks the limitations of a single data source, accurately and effectively capturing user printing data in different operating scenarios, avoiding subsequent strategy deviations due to data omissions. The real-time monitoring mechanism ensures that printing data is captured and processed instantly upon generation, reducing data transmission latency and effectively improving the efficiency of the overall printing process. Irrelevant data is filtered out through feature rules, and data standardization is achieved through encapsulation processing, effectively reducing the interference of invalid data on subsequent entity extraction and knowledge graph updates, ensuring high quality and usability of data transmitted to the printing terminal. Simultaneously, it supports monitoring and processing of multiple data stream types, adapting to printing data generated by different operating systems, applications, and terminal devices, enhancing the universality of the data processing method described in this embodiment in complex terminal environments, and effectively reducing data acquisition failures caused by terminal differences.

[0024] In one embodiment of the present invention, the printing cloud server of the printing terminal performs data entity extraction processing on the printing data to obtain printing entity data corresponding to the printing data, including: The printing cloud server parses the received encapsulated data packets to obtain the printing data; Scan the printed data to obtain the file name and file content data corresponding to the printed data; Retrieve key information data from the file name and file content data; wherein, the key information data includes basic information and key general terms, and the basic information includes, but is not limited to, username, project name and date, etc. The basic information is designated as the first entity data, and key general terms are designated as the second entity data. These key general terms include, but are not limited to, budgets, plans, and academic papers.

[0025] The working principle of the above technical solution is as follows: In this embodiment, the printing cloud server of the printing terminal first parses the received encapsulated data packet to restore the original printing data; it retrieves the target printing data from the parsed printing data to identify the core object to be processed; it separates the file name and file content data corresponding to the printing data through scanning operations, realizing the splitting of data dimensions; for the file name, it extracts the first entity data using entity recognition, wherein the first entity data mainly targets entity information with scene-specific characteristics such as username, project name, and date; then, for the file content data, it extracts key general terms such as budget, planning, and thesis as the second entity data through keyword extraction and semantic analysis. This embodiment starts with data parsing, and through layered scanning and targeted extraction, it achieves accurate separation and extraction of different types of entity information in the printing data, providing structured entity data support for subsequent knowledge graph updates.

[0026] The above technical solution achieves the following effects: it breaks down unstructured print data into file names and content data, and further extracts entity information that can be directly used for analysis, breaking the disorder of the original data and making the data present structured characteristics, which is convenient for subsequent system processing and application. By extracting first and second entity data separately from file names and content data, entity extraction focuses on key information in different dimensions. The first entity data is used to strengthen scene association, while the second entity data is used to highlight the content theme. Entity data extracted through different dimensions of key information effectively avoids the mixing and omission of entity information, effectively improving the quality of entity data. The entity data obtained in this way serves as the core feature of print data, accurately reflecting the attributes and background of the printing task. This allows the printing terminal to associate user history and work scenarios based on entity information when reasoning about printing strategies, effectively improving the pertinence and rationality of strategy reasoning. Simultaneously, through layered scanning and targeted extraction, interference from irrelevant data on entity recognition is reduced, shortening the entity extraction processing chain. While ensuring extraction quality, processing efficiency is effectively improved, ensuring that entity data can quickly support subsequent processes such as knowledge graph updates. One embodiment of the present invention involves dynamically updating the user's completed and deployed personal work context knowledge graph using the entity coefficients corresponding to the printed entity data, including: Retrieve the first entity data and the second entity data, and obtain the entity coefficients corresponding to the first entity data and the second entity data; Map the first entity data and the second entity data to the user's corresponding personal work context knowledge graph that has been built and deployed; Determine whether there is entity data in the personal work context knowledge graph that corresponds to the first entity data and the second entity data; When there is a combination of entity data in the personal work context knowledge graph that corresponds to the first entity data and the second entity data, the node corresponding to the entity data is marked. When there is no entity data combination in the personal work context knowledge graph that corresponds to the first entity data and the second entity data, then the entity coefficients corresponding to the first entity data and the second entity data are combined to create a node and obtain the dynamically updated personal work context knowledge graph corresponding to the user.

[0027] The working principle of the above technical solution is as follows: First and second entity data are retrieved, and their corresponding entity coefficients are obtained. These two types of entity data are mapped to the constructed user's personal work context knowledge graph, establishing an initial association between the data and the graph. Through the entity matching mechanism within the graph, it is determined whether there exists an entity combination in the existing graph that is completely identical to the current first and second entity data. If an identical combination exists, the corresponding node is marked to enhance its activity in the graph. If no identical combination exists, the initial weight of the new node and its association strength with related nodes are calculated based on the entity coefficients. A new node is then created and embedded into the graph, achieving dynamic expansion of the knowledge graph. The entire process is based on entity data mapping, using matching judgments to update or create nodes, and relying on entity coefficients to ensure the accuracy and rationality of graph updates, enabling the knowledge graph to continuously reflect the user's latest work behavior characteristics.

[0028] The above technical solution achieves the following effects: By retrieving entity data in real time and updating the knowledge graph based on entity coefficients, the knowledge graph can quickly respond to entity information generated by the user's latest printing behavior, avoiding graph obsolescence due to data lag and ensuring that the graph always reflects the user's current work status, effectively improving the real-time performance of user behavior analysis. Entity combination matching effectively avoids the problem of duplicate creation of identical entity combinations and reduces graph redundancy. Simultaneously, the marking mechanism strengthens the activity of existing nodes, making the relationships between entities more closely match the actual interaction frequency and importance, effectively improving the graph's accuracy in depicting the user's work context. When creating nodes for new entity combinations, the node weights and relationships are planned based on entity coefficients, allowing new nodes to be naturally embedded into the existing graph. This supplements new entity information while maintaining the logic and relevance of the graph structure, gradually improving the graph's coverage of the user's work scenarios.

[0029] The introduction of entity coefficients allows the node update and creation process to fully consider the personalized characteristics of entities, such as the degree of user attention to specific project names and the frequency of use of key terms. This makes the structure and weight distribution of the knowledge graph more aligned with individual user behavior habits, providing more accurate basic data for subsequent personalized printing strategy reasoning. Matching judgments reduce duplicate nodes, effectively lowering the consumption of graph storage and computing resources. Simultaneously, dynamically adjusting node weights based on entity coefficients tilts graph resources towards high-value entities (such as frequently occurring project names and core terms), effectively improving the graph's ability to focus on key information and enhancing the efficiency and accuracy of subsequent strategy reasoning. In one embodiment of the present invention, the entity coefficients corresponding to the first entity data are obtained in the following manner, including: Retrieve the first entity data and determine the number of data types present in the first entity data; The system retrieves the preset base weight value for each data type in the first entity data. This base weight value is set based on experience and ranges from 0 to 1. For example, the weight of the username is set to 0.5, the weight of the project name to 1.0, and the weight of the date to 0.7. The weight setting for the username of a fixed user in the first entity data is fixed; each user has a fixed weight value for their username. This value can be determined randomly or after considering practical factors. This setting ensures that for the same user, the weight of their username remains consistent across subsequent printing processes. Dates and project names, however, change and are updated, thus introducing some variability. Compared to different printed files in historical data, such as those with identical information but different dates, files with closer dates have a higher weight, indicating that the newer date content is likely more up-to-date.

[0030] Obtain the entity coefficient corresponding to the first entity data based on the basic weight value corresponding to the data type of the first entity data. ,in, A 01 This represents the entity coefficient corresponding to the first entity data; n Indicates the number of data types present in the first entity data; W i Indicates the first entity data containing the first entity. i The basic weight values ​​corresponding to each data type; W z01 Indicates the first entity data corresponding to n The median value of the basic weight corresponding to each data type.

[0031] The working principle of the above technical solution is as follows: by retrieving the first entity data, firstly, the number of data types contained therein is determined; based on preset rules, the basic weight value corresponding to each data type is retrieved, wherein the basic weight value ranges from 0 to 1, is set by experience, and reflects the inherent importance of different types of entities; subsequently, the entity coefficient is calculated through the above formula. In this embodiment, the above technical solution is based on data type identification, uses the preset basic weight as a reference, and uses the intermediate value to balance and adjust the weight of each type, thereby realizing the quantitative representation of the comprehensive importance of the first entity data.

[0032] The above technical solution achieves the following effects: By introducing basic weights to reflect the inherent importance of different data types and combining them with intermediate values ​​to balance extreme differences in weights across types, the calculated entity coefficients not only reflect the preset importance of each entity type but also avoid excessive influence of a single high-weight or low-weight type on the results through the adjustment effect of intermediate values, thus enhancing the coefficients' true reflection of the comprehensive value of the first entity data. The basic weight values ​​can be flexibly adjusted according to actual application scenarios and experience, adapting to the differentiated needs of different user groups and work scenarios regarding the importance of entity types, making the calculation of entity coefficients more aligned with specific business logic and effectively improving the universality of the technical solution. The introduction of intermediate values ​​effectively buffers the impact of extreme weight values ​​on the overall results, reduces the impact of fluctuations in the weights of a single data type on the entity coefficients, making the coefficient results more stable and providing a reliable quantitative basis for subsequent knowledge graph updates based on entity coefficients. The quantified entity coefficients accurately characterize the importance of the first entity data, providing clear numerical references for node creation, weight adjustment, and other operations in the knowledge graph, ensuring that the weight allocation of entity data in the graph matches its actual importance, and effectively improving the rationality and accuracy of knowledge graph updates. Meanwhile, the above-mentioned technical solution in this embodiment reduces the consumption of computing resources while ensuring the quantification effect, making it easy to execute efficiently on devices such as printing terminals and effectively improving the overall system operating efficiency.

[0033] Existing technologies often rely on fixed algorithms for entity weight calculation, requiring manual parameter adjustments when user behavior patterns undergo implicit changes over time. This proposed solution, however, uses a dynamic balancing of intermediate values ​​for basic weights. This automatically mitigates the impact of occasional extreme weights on the overall coefficients without manual intervention, allowing entity coefficients to change gradually as user behavior fluctuates, creating a self-calibrating effect. This dynamic adaptability is difficult to achieve with simple weighting or fixed algorithms. Furthermore, existing entity coefficient calculation technologies often focus on the importance of a single entity, while this solution indirectly enhances the ability to mine the correlations between entity combinations by linking the number of data types with the median value. This allows subsequent knowledge graphs to more accurately capture implicit relationships between entities, and this implicit enhancement of combined semantics goes beyond the original design intention of simply calculating entity importance. Meanwhile, while complex entity weighting algorithms in existing technologies (such as those incorporating machine learning models) offer high accuracy, they are difficult to run on edge devices such as printing terminals. The technical solution in this embodiment aims only to simplify calculations, achieving a balance between lightweight and high accuracy. Specifically, the calculation of the median value in the formula does not require iterative operations; it can be completed simply by sorting and taking the median. This effectively reduces the calculation time for a single data entry while ensuring coefficient accuracy, effectively adapting to the computing power limitations of edge devices. Additionally, existing technologies are prone to coefficient distortion when encountering abnormal weight values. The existence of the median value in the solution of this embodiment forms a natural fault-tolerance mechanism. Even if the base weight of a certain data type is mistakenly set to an extreme value, the median value can still neutralize its impact through the weights of other types, minimizing the deviation of the final coefficient and effectively improving the soft fault tolerance capability for abnormal data.

[0034] In one embodiment of the present invention, the entity coefficients corresponding to the second entity data are obtained in the following manner, including: Retrieve the second entity data and determine the number of data types present in the second entity data; Retrieve the preset basic weight value corresponding to each data type in the second entity data. The basic weight value is set based on experience, and the weight value setting range is 0-1. For example, the weight of budget is set to 0.8, the weight of planning is set to 0.6, and the weight of the paper is set to 0.4. Based on each data type corresponding to the first entity data and each data type in the second entity data, obtain the association strength between each data type in the second entity data and all data types in the first entity data; The entity coefficients corresponding to the second entity data are obtained by combining the preset basic weight values ​​corresponding to each data type in the second entity data with the correlation strength between each data type in the second entity data and all data types in the first entity data.

[0035] The entity coefficients corresponding to the second entity data are obtained in the following way: Retrieve the second entity data j The correlation strength between each data type and all data types of the first entity data. AS j ; and the correlation strength AS j The second entity data contains the first jBasic weight values ​​corresponding to each data type W j After performing the product operation, the square root is taken to obtain the first element from the second entity data. j Smoothing parameters for each data type H j = ;in, W j This indicates that the second entity data contains the first j The basic weight values ​​corresponding to each data type; AS j Indicates the second entity data j The correlation strength between each data type and all data types of the first entity data; Then, using the second entity data... j The smoothing parameters corresponding to each data type are combined with the second entity data. m The median value of the basic weights corresponding to each data type is used to obtain the entity coefficient corresponding to the first entity data. ,in, A 02 This represents the entity coefficient corresponding to the first entity data; m Indicates the number of data types contained in the second entity data; W z02 Indicates the second entity data corresponding to m The median value of the basic weight corresponding to each data type.

[0036] The working principle of the above technical solution is as follows: First, by retrieving the second entity data, the number of data types it contains is determined; then, a preset basic weight value corresponding to each data type is retrieved. The preset basic weight value ranges from 0 to 1 and is set based on experience; further, the association strength between each type in the second entity data and all types in the first entity data is calculated, wherein the association strength is used to reflect the semantic and contextual relevance of the two types of entities; finally, the above parameters are integrated through a formula, with the basic weight as the core, combined with... m The intermediate value of the basic weights for each type (balancing the impact of extreme weights) is incorporated into the association strength to finally calculate the entity coefficient of the second entity data. The entire process is based on the attributes of the second entity itself, forms a cross-type linkage with the first entity through the association strength, and uses the intermediate value to stabilize the calculation result, thereby quantifying the comprehensive importance of the second entity data, and this quantification result is intrinsically related to the first entity.

[0037] The above technical solution achieves the following effect: by introducing the correlation strength with the first entity's data, the entity coefficients of the second entity not only depend on their own basic weights but also are linked to the type characteristics of the first entity. This avoids isolated quantification of the second entity, making the coefficients more closely reflect the collaborative relationship between the two types of entities in actual business scenarios, and effectively improving the coefficients' ability to represent complex working contexts. The basic weights reflect the inherent importance of each type of the second entity, the intermediate values ​​buffer extreme weight fluctuations, and the correlation strength... AS j The dynamic adaptation and the degree of association with the first entity, combined with the other three factors, allow the coefficient calculation to retain the stability of preset empirical values ​​while incorporating the flexibility of dynamic association, reducing the bias of single-dimensional quantification. Since the coefficient simultaneously associates with both the first and second entities, updating or creating knowledge graph nodes based on this coefficient can more accurately reflect the intrinsic connection between the two types of entities, avoiding the fragmentation of entity relationships in the knowledge graph, strengthening the linkage between cross-type entities in the graph, and providing a more coherent basis for entity association for subsequent strategy reasoning. Simultaneously, the association strength... AS j The introduction of this feature enables the coefficients to dynamically respond to changes in the first entity. For example, when the type of the first entity is updated, the association strength is adjusted synchronously, allowing the entity coefficients of the second entity to have dynamic adaptability. This effectively adapts to the dynamic changes in entity relationships in the user's work scenario and avoids the scenario rigidity problem caused by static weights.

[0038] Furthermore, existing technologies typically calculate entity weights in isolation, while the above-described scheme in this embodiment dynamically binds the second entity to the first entity through association strength. This allows the entity coefficient to reflect not only its own attributes but also the quantification of implicit relationships between entities. This association forms entity clusters as data accumulates, enabling the knowledge graph to automatically exhibit cross-type entity clustering characteristics, far exceeding the expected effect of simple weight aggregation. When the base weight of a certain second entity data type is abnormal (e.g., mistakenly set to an extremely high value), the introduction of association strength naturally dilutes its extreme impact; that is, if the association between this type and the first entity is weak, the association strength... AS j Low values ​​of these values ​​reduce the actual weight of the entity in the coefficients. This association verification mechanism is impossible to achieve with existing single-dimensional weight calculations, enhancing the system's tolerance to human configuration errors. Since the coefficient calculation integrates the scenario attributes of the first entity and the semantic attributes of the second entity, when a user switches work scenarios, the system can automatically adjust the weight priority of terms based on historical association strength, allowing the knowledge graph to adaptively adjust according to scenario changes. Simultaneously, by continuously calculating association strength, the system captures user-unannounced work patterns. For example, when username A is frequently associated with papers, the system automatically identifies the user's academic attributes and uses coefficient changes to influence the knowledge graph, making subsequent entity processing more aligned with the user's implicit habits.

[0039] One embodiment of the present invention involves obtaining the association strength between each data type in the second entity data and all data types in the first entity data, based on each data type corresponding to the first entity data and each data type in the second entity data, including: Retrieve the user's historical print data; Extract the total number of times each data type appears in the user's historical print data from the second entity data; Extract the data types contained in the first entity data corresponding to each combination that appears in the second entity data each time a data type appears; The association strength between each data type in the second entity data and all data types in the first entity data is obtained by combining the total number of times each data type in the second entity data appears in the user's corresponding historical print data with the data types contained in the first entity data that appear in the corresponding combination each time each data type in the second entity data appears. ,in, ASj Indicates the second entity data j The correlation strength between each data type and all data types of the first entity data; m Indicates the number of data types contained in the second entity data; N j Indicates the second entity data j The total number of times each data type appears in the user's corresponding historical print data; N ij Indicates the second entity data j The number of times the i-th data type contained in the first entity data appears simultaneously when a data type appears; W i Indicates the first entity data containing the first entity. i The basic weight values ​​corresponding to each data type; W j This indicates that the second entity data contains the first j The basic weight value corresponding to each data type.

[0040] The working principle of the above technical solution is as follows: By retrieving users' historical printing data, the association strength is calculated based on historical behavioral data. First, the total occurrence frequency of each data type in the second entity data is extracted from the historical data to establish a frequency statistics benchmark. Second, the simultaneous occurrence of each type in the corresponding first entity data is tracked each time the second entity type appears, and the co-occurrence frequency is recorded. Finally, the co-occurrence frequency, total frequency, and basic weights of the two types of entities are integrated and calculated using a formula to obtain the comprehensive association strength between the j-th type of the second entity and all types of the first entity. The entire process uses historical co-occurrence data as the core, integrating frequency statistics and weight adjustment to achieve a quantitative representation of the association strength between entity types.

[0041] The above technical solution achieves the following effects: By introducing historical co-occurrence frequency to reflect actual association frequency, and combining it with basic weights to reflect the inherent importance of entity types, it avoids the problem of misjudging high-frequency, low-value associations caused by simply relying on frequency. This ensures that the quantitative results conform to user behavior patterns and fit the contextual value of entity types, enhancing the objectivity and accuracy of association strength. Furthermore, calculating associations based on users' own historical printing data allows the results to reflect the behavioral habits of specific users, rather than general group characteristics, effectively improving the ability of association strength to characterize individual user work patterns and providing more relevant basic data for subsequent personalized strategies. Moreover, as user historical data accumulates, co-occurrence frequency is updated in real time, and the association strength calculation results can automatically adjust with changes in user behavior, enabling the quantitative representation of entity associations to dynamically adapt to the evolution of user work habits and avoiding the association lag problem caused by static rules. Additionally, through cross-entity type co-occurrence analysis of the first and second entities, the semantic hierarchy of entity relationships is enriched, providing richer association evidence for knowledge graph construction. Meanwhile, precise association strength provides a quantitative reference for setting connection weights between entity nodes, ensuring a high degree of real-time consistency between the closeness of associations between entities in the knowledge graph and the actual behavioral characteristics of users, effectively improving the quality of the graph's structured mapping of the user's working context.

[0042] One embodiment of the present invention involves creating nodes by combining entity coefficients corresponding to first entity data and second entity data to obtain a dynamically updated personal work context knowledge graph for a user, including: Retrieve the combination of the first entity data and the second entity data corresponding to each node in the user's personal work context knowledge graph; The similarity between the combination of the first entity data and the second entity data corresponding to each node in the personal work context knowledge graph and the current combination of the first entity data and the second entity data is compared to obtain the similarity value between the combination of the first entity data and the second entity data corresponding to each node in the personal work context knowledge graph and the current combination of the first entity data and the second entity data. Retrieve the entity coefficients of the first and second entity data corresponding to each node in the user's personal work context knowledge graph. By combining the entity coefficients of the first and second entity data corresponding to each node in the user's personal work context knowledge graph with the similarity value corresponding to the combination of the first and second entity data of each node, the comprehensive strength value corresponding to the combination of the first and second entity data of the current node is obtained. ,in, Y This represents the combined strength value corresponding to the combination of each node with the current first entity data and second entity data; S This represents the similarity score between the combination of the first entity data and the second entity data corresponding to each node in the personal work context knowledge graph and the current combination of the first entity data and the second entity data. A 01c This represents the absolute difference in entity coefficients between the first entity data corresponding to each node in the personal work context knowledge graph and the current first entity data. A 02c This represents the absolute difference in entity coefficients between the second entity data corresponding to each node in the personal work context knowledge graph and the current first entity data. The comprehensive strength value is compared with a preset strength threshold, and nodes with comprehensive strength values ​​not lower than the preset strength threshold are selected as associated nodes. For the combination of the current first entity data and the second entity data, a target node is created in the personal work context knowledge graph, and a directed edge is established between the target node and the associated node (from the new node to the associated node). Retrieve the entity coefficients corresponding to the first and second entity data of the target node, and the entity coefficients corresponding to the first and second entity data of each associated node; Based on the entity coefficients corresponding to the first and second entity data of the target node and the entity coefficients corresponding to the first and second entity data of each associated node, set the edge weight values ​​for the directed edges established between the target node and the associated nodes. ;in, W b Indicates the edge weight value;S g This represents the numerical similarity between the associated node and the target node. Y m This represents the overall strength value between the target node and its associated nodes.

[0043] The working principle of the above technical solution is as follows: First, the combination of the first and second entity data of existing nodes in the user's personal work context knowledge graph is retrieved and compared with the current entity data combination to obtain a similarity score. Then, combining the entity coefficients of existing nodes and the entity coefficients of the current entity data, the absolute difference in entity coefficients and the similarity are calculated to obtain the comprehensive strength score of each existing node and the current entity combination. This comprehensive strength score is compared with a preset threshold to filter out qualified associated nodes. A target node is created based on the current entity combination, and a directed edge is established between the target node and associated nodes. Finally, the weights of the directed edges are calculated and set using a dedicated formula based on the similarity and comprehensive strength scores between the associated nodes and the target node, ultimately achieving dynamic updates to the knowledge graph. The entire process relies on entity combination similarity and entity coefficient differences as its core basis, using quantitative calculations to achieve accurate embedding of new nodes and reasonable construction of association relationships.

[0044] The above technical solution achieves the following effects: By comparing similarity and calculating comprehensive strength, it ensures that target nodes only connect with existing nodes that have high relevance, reducing meaningless associations and making the relationships between nodes in the graph more consistent with the inherent logic of the entity data, thus enhancing the accurate mapping of the user's working context. The edge weight settings integrate both similarity and comprehensive strength dimensions, reflecting both the matching degree of entity combinations and the association depth of entity coefficients, making the quantification of the strength of connections between nodes more hierarchical and providing a reliable weight basis for subsequent graph-based reasoning. Simultaneously, by filtering associated nodes through preset thresholds, graph redundancy caused by new nodes connecting with low-relevance nodes is avoided, and directed edges are created based on quantified indicators, ensuring the graph maintains a clear structure during expansion and effectively improving its maintainability and query efficiency. Furthermore, based on real-time comparison and calculation between current entity data and existing nodes, it ensures that the creation of new nodes and the establishment of associations instantly reflect the latest entity data characteristics, enabling the knowledge graph to dynamically adapt to changes in the user's working context and effectively improving the timeliness of user behavior analysis. Finally, the comprehensive strength calculation effectively balances the impact of differences between entity data on the association strength by introducing the absolute difference of entity coefficients. The edge weight setting further associates similarity and comprehensive strength, so that the connection between the target node and the associated node not only conforms to the historical association pattern, but also reflects the current data characteristics, effectively improving the logical consistency between the association pattern and data characteristics reflected by the overall graph.

[0045] In one embodiment of the present invention, a printing cloud server infers a printing strategy based on the printing entity data using a working context knowledge graph, and recommends the printing strategy to the user terminal, including: When there is no entity data combination in the personal work context knowledge graph that corresponds to the first entity data and the second entity data, locate the associated node associated with the target node; Retrieve from the database the printing strategies corresponding to the associated nodes related to the target node, as well as the edge weights between the target node and the associated nodes; By combining edge weights with a sorting strategy, the printing strategies corresponding to the associated nodes are sorted to obtain a set of printing strategy sequences. The printing strategies corresponding to the associated nodes are recommended to the user terminal according to the order of the printing strategies in the printing strategy sequence set. When there is a combination of entity data in the personal work context knowledge graph that corresponds to the first entity data and the second entity data, the node marked in the personal work context knowledge graph is searched. After identifying the marked nodes, the corresponding printing strategies for the marked nodes are retrieved from the database and recommended to the user terminal.

[0046] The sorting strategy is as follows: Retrieve the edge weights between each associated node and the target node, and the total number of times the printing strategy corresponding to each associated node has been adopted by users in the historical recommendation records; Retrieve all adopted printing strategies from the user's historical adoption records, and use all adopted printing strategies other than the printing strategy corresponding to the node associated with the point as the reference printing strategy. Each reference printing strategy is compared with the printing strategy corresponding to each associated node in turn to obtain the strategy similarity value corresponding to the printing strategy of each associated node. The ranking weight of each associated node's printing strategy is obtained by combining the strategy similarity value of each associated node with the edge weight between each associated node and the target node, and the total number of times the printing strategy of each associated node has been adopted by users in the historical recommendation records. ;in, Q This represents the sorting weight corresponding to the printing strategy of each associated node; W b Indicates the edge weight value; N g This represents the total number of times the normalized printing strategy corresponding to each associated node has been adopted by users in the historical recommendation record; k Indicates the number of reference printing strategies; Sci Indicates the first i The strategy similarity value between each reference printing strategy and the printing strategy corresponding to the associated node; S cb Indicates the associated node and its corresponding k The standard deviation of the strategy similarity of each reference printing strategy; The printing strategies are sorted according to their respective sorting weights.

[0047] The working principle of the above technical solution is as follows: In this embodiment, when the printing cloud server of the printing terminal infers the printing strategy based on the personal work context knowledge graph, it handles two scenarios: When there is no entity data combination in the graph that is the same as the current first and second entity data combinations, the printing cloud server of the printing terminal retrieves the historical printing strategy corresponding to the associated nodes associated with the newly created target node, and simultaneously obtains the edge weights between the target node and each associated node. It then sorts the printing strategies of the associated nodes according to the combination sorting strategy and recommends the sorted strategies to the user terminal. When there is the same entity data combination in the graph, it directly retrieves the historical printing strategy associated with the marked node corresponding to that combination and directly recommends it to the user terminal. The entire process is based on the matching results of entity data combinations in the knowledge graph, and achieves targeted inference and recommendation of printing strategies through the edge weights of associated nodes or the historical strategies of marked nodes, ensuring the relevance of the strategy to the user's work context and historical behavior.

[0048] The above technical solution achieves the following effects: by combining edge weights with ranking strategies to filter highly relevant node strategies, or by directly calling the strategies of marked nodes with the same entity combinations, the recommended printing strategies closely align with the entity characteristics of the current printing task and the user's historical behavior patterns, reducing interference from irrelevant strategies and effectively improving the matching accuracy between strategies and actual needs. Simultaneously, for scenarios with the same entity combinations, historical strategies of marked nodes are directly reused, eliminating the complex reasoning and ranking process and shortening the response time for strategy generation and recommendation. For scenarios with new entity combinations, edge weights are used to quickly locate highly relevant strategies, avoiding the resource consumption of full-scale retrieval and effectively improving overall recommendation efficiency. Recommending based on historical strategies of related nodes in the knowledge graph essentially continues the user's printing preferences in similar work scenarios, allowing strategy recommendations to reflect the user's individual behavioral habits; while the reuse of marked node strategies directly preserves the user's fixed choice for specific entity combinations, thereby enhancing the stability of personalized services. Furthermore, the ranking relationship obtained by combining edge weights with the ranking strategy presents the priority of printing strategies, providing users with clear selection guidance and effectively reducing the decision-making cost for users among multiple strategies. Simultaneously, strategy recommendation is based on a work context knowledge graph, effectively improving the relevance of recommendation results to the user's current work scenario, thereby reducing the frequency of secondary adjustments due to insufficient strategy adaptability. Moreover, as entity associations in the knowledge graph are dynamically updated, the edge weights of associated nodes adjust according to changes in user behavior, enabling the strategy recommendation logic to automatically adapt to the evolution of user printing habits. This implicit optimization of the recommendation model can be achieved without manual intervention, effectively enhancing the long-term applicability of the system.

[0049] On the other hand, existing technologies typically recommend printing strategies based on only one or a few factors (such as print file type or past user choices). However, the ranking strategy in this embodiment integrates multiple dimensions, including edge weights, the total number of times a printing strategy has been adopted by the user, and strategy similarity scores, enabling it to more accurately adapt to complex and ever-changing work scenarios. For example, in actual work, in addition to common file types, there may be a complex interplay of factors such as different work projects and different collaborators. Edge weights reflect the correlation between these factors and the current printing task, the total number of adoptions reflects the user's historical preferences, and strategy similarity captures the similarity between different printing strategies when dealing with similar work scenarios. Through multi-dimensional comprehensive consideration, even in the face of entirely new and complex work contexts, it can more accurately recommend suitable printing strategies, thereby effectively improving the accuracy and precision of printing strategy recommendations.

[0050] In practical work, printing entity data may be ambiguous or incomplete. This ranking strategy leverages the relationships between nodes in a knowledge graph (i.e., edge weights) and similarity comparisons with other reference printing strategies. Even with incomplete information, it can infer a more reasonable ranking of printing strategies by using related information and similar strategies. In contrast, existing technologies may suffer from poor recommendation performance due to insufficient information in such situations, failing to accurately recommend suitable printing strategies. Furthermore, this embodiment combines edge weight values, the total number of times a normalized printing strategy is adopted, strategy similarity values, and the standard deviation of strategy similarity values. This allows for adaptive and dynamic adjustment of the ranking weight of each associated node's printing strategy based on different work contexts and user behavior patterns. Existing technologies lack this dynamic adjustment mechanism, making it difficult to make flexible and accurate ranking adjustments based on real-time changes. For example, when a company's business direction changes, user needs for printing strategies will also change. This mathematical model can adjust the ranking weight of printing strategies in a timely manner based on new user adoption data and changes in the similarity between different strategies. Existing ranking methods often cannot adapt to such changes promptly, resulting in lagging recommendation strategies. Meanwhile, the introduction of the standard deviation of strategy similarity can effectively reduce the impact of abnormal reference printing strategies on the ranking weights. When calculating strategy similarity, there may be anomalies where individual reference printing strategies have extremely high or low similarity to the printing strategies of associated nodes. The standard deviation of strategy similarity can balance and correct these anomalies, making the ranking weights more reflective of the true superiority or inferiority of strategies.

[0051] This invention proposes a personalized printing push system based on user behavior analysis, such as... Figure 2 As shown, the personalized printing push system includes: A data acquisition module is used to collect printing data from user terminal devices in real time and send the printing data to the printing terminal; wherein, the printing terminal includes a printing cloud server and a printer; The entity extraction module is used by the printing cloud server of the printing terminal to perform data entity extraction processing on the printing data to obtain the printing entity data corresponding to the printing data. The graph update module is used to dynamically update the user's personal work context knowledge graph that has been built and deployed by the user through the entity coefficients corresponding to the printed entity data. The inference module is used by the printing cloud server to infer the printing strategy based on the printing entity data through the working context knowledge graph, and recommend the printing strategy to the user terminal. The print execution module is used for the print cloud server to wait for the policy confirmation instruction sent by the user terminal. When the print cloud server receives the user's policy confirmation instruction, it controls the printer to perform the printing operation according to the print policy confirmed by the user.

[0052] The working principle of the above technical solution is as follows: The personalized printing push method in this embodiment collects printing data from the user terminal in real time and transmits it to the printing terminal. The printing terminal extracts entities from the printing data to obtain printing entity data. Using this printing entity data and corresponding entity coefficients, the constructed user's personal work context knowledge graph is dynamically updated, so that the knowledge graph continuously reflects the relationship between user behavior characteristics and work scenarios. Based on the updated knowledge graph, the printing terminal infers a printing strategy that adapts to the user's behavior habits and current work context and pushes it to the user terminal. After receiving the user's confirmation instruction, the printing operation is executed according to the confirmed strategy. By deeply associating printing data with the user's work context, a closed-loop personalized push is achieved from data collection to strategy execution.

[0053] The effects of the above technical solution are as follows: By analyzing user behavior and dynamically updating the knowledge graph, printing strategies can be tailored to users' long-established operating habits and preferences, reducing manual parameter settings, effectively improving operational convenience, and lowering the rate of human error. Simultaneously, integrating printing entity data into the user's personal work context knowledge graph allows the printing strategy reasoning process to fully consider the user's current work scenario, such as task progress and collaboration needs, effectively improving the matching degree between the strategy and actual work requirements and enhancing the scenario adaptability of the printing service. On the other hand, optimizing printing strategies can effectively reduce unnecessary paper and consumable waste, thereby lowering office costs. Furthermore, based on the dynamic updating and reasoning capabilities of the knowledge graph, the printing terminal transforms from a passive execution device into an intelligent service terminal with proactive decision-making capabilities, adaptively adjusting the strategy recommendation logic according to changes in user behavior, effectively improving the system's self-optimization capabilities. By automatically recommending suitable printing strategies, the user's operational burden is reduced while ensuring that the printing results meet actual needs, enhancing user satisfaction and reliance on the printing service.

[0054] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for personalized print push notifications based on user behavior analysis, characterized in that, The personalized printing push method includes: S1: Real-time acquisition of printing data from user terminal devices and transmission of the printing data to the printing terminal; wherein, the printing terminal includes a printing cloud server and a printer; S2: The printing cloud server of the printing terminal performs data entity extraction processing on the printing data to obtain the printing entity data corresponding to the printing data; S3: Dynamically update the user's completed and deployed personal work context knowledge graph by using the entity coefficients corresponding to the printed entity data; S4: The printing cloud server infers a printing strategy based on the printing entity data through a work context knowledge graph, and recommends the printing strategy to the user terminal; S5: The print cloud server waits for the policy confirmation instruction sent by the user terminal. When the print cloud server receives the user's policy confirmation instruction, it controls the printer to perform the printing operation according to the printing policy confirmed by the user. Step S3 includes: retrieving first entity data and second entity data, and obtaining entity coefficients corresponding to the first entity data and second entity data; mapping the first entity data and second entity data to the user's corresponding completed and deployed personal work context knowledge graph, wherein the first entity data is basic information and the second entity data is key general terms; Determine whether there is entity data in the personal work context knowledge graph that corresponds to the first entity data and the second entity data; When there is a combination of entity data in the personal work context knowledge graph that corresponds to the first entity data and the second entity data, the node corresponding to the entity data is marked. When there is no entity data combination in the personal work context knowledge graph that corresponds to the first entity data and the second entity data, then the node is created by combining the entity coefficients corresponding to the first entity data and the second entity data to obtain the dynamically updated personal work context knowledge graph corresponding to the user. Specifically, node creation is performed by combining the entity coefficients corresponding to the first entity data and the second entity data to obtain the dynamically updated personal work context knowledge graph corresponding to the user, including: retrieving the combination of the first entity data and the second entity data corresponding to each node in the personal work context knowledge graph corresponding to the user. The similarity between the combination of the first entity data and the second entity data corresponding to each node in the personal work context knowledge graph and the current combination of the first entity data and the second entity data is compared to obtain the similarity value between the combination of the first entity data and the second entity data corresponding to each node in the personal work context knowledge graph and the current combination of the first entity data and the second entity data. Retrieve the entity coefficients of the first and second entity data corresponding to each node in the user's personal work context knowledge graph. By combining the entity coefficients of the first entity data and the second entity data of each node in the knowledge graph of the user’s personal work context, and the similarity value of the combination of the first entity data and the second entity data of each node, the comprehensive strength value of each node and the current combination of the first entity data and the second entity data is obtained. The comprehensive strength value is compared with a preset strength threshold, and nodes with comprehensive strength values ​​not lower than the preset strength threshold are selected as associated nodes; a target node is created in the personal work context knowledge graph for the current combination of first entity data and second entity data, and a directed edge is established between the target node and the associated node; the entity coefficients corresponding to the first entity data and second entity data of the target node and the entity coefficients corresponding to the first entity data and second entity data of each associated node are retrieved; Based on the entity coefficients corresponding to the first and second entity data of the target node and the entity coefficients corresponding to the first and second entity data of each associated node, set the edge weight values ​​corresponding to the directed edges established between the target node and the associated nodes.

2. The personalized printing push method according to claim 1, characterized in that, Real-time acquisition of printing data from user terminal devices and transmission of the printing data to the printing terminal, including: Real-time monitoring of multiple data streams from user terminal devices, including system event streams, browser metadata streams, document metadata streams, and Hitachi event streams; The print data is filtered from the various data streams, and the print data is encapsulated to obtain the encapsulated data packet corresponding to the user terminal. The encapsulated data packet is sent to the printing terminal.

3. The personalized printing push method according to claim 1, characterized in that, The printing terminal's printing cloud server performs data entity extraction processing on the printing data to obtain the printing entity data corresponding to the printing data, including: The printing cloud server parses the received encapsulated data packets to obtain the printing data; Scan the printed data to obtain the file name and file content data corresponding to the printed data; Retrieve key information data from the file name and file content data; wherein, the key information data includes basic information and key general terms, and the basic information includes, but is not limited to, username, project name and date; The basic information is used as the first entity data, and the key general terms are used as the second entity data.

4. The personalized printing push method according to claim 1, characterized in that, The entity coefficients corresponding to the first entity data are obtained in the following ways: Retrieve the first entity data and determine the number of data types present in the first entity data; Retrieve the preset basic weight value corresponding to each data type in the first entity data, wherein the basic weight value is set based on experience, and the weight value setting range is 0-1; The entity coefficient corresponding to the first entity data is obtained based on the basic weight value corresponding to the data type of the first entity data.

5. The personalized printing push method according to claim 4, characterized in that, The entity coefficients corresponding to the second entity data are obtained in the following ways: Retrieve the second entity data and determine the number of data types present in the second entity data; Retrieve the preset basic weight value corresponding to each data type in the second entity data, wherein the basic weight value is set based on experience, and the weight value setting range is 0-1; Based on each data type corresponding to the first entity data and each data type in the second entity data, obtain the association strength between each data type in the second entity data and all data types in the first entity data; The entity coefficients corresponding to the second entity data are obtained by combining the preset basic weight values ​​corresponding to each data type in the second entity data with the correlation strength between each data type in the second entity data and all data types in the first entity data.

6. The personalized printing push method according to claim 5, characterized in that, Based on each data type corresponding to the first entity data and each data type in the second entity data, obtain the association strength between each data type in the second entity data and all data types in the first entity data, including: Retrieve the user's historical print data; Extract the total number of times each data type appears in the user's historical print data from the second entity data; Extract the data types contained in the first entity data corresponding to each combination that appears in the second entity data each time a data type appears; The association strength between each data type in the second entity data and all data types in the first entity data is obtained by combining the total number of times each data type in the second entity data appears in the user's corresponding historical print data with the data types contained in the first entity data that appear in the corresponding combination each time each data type in the second entity data appears.

7. The personalized printing push method according to claim 1, characterized in that, The printing cloud server infers a printing strategy based on the printing entity data using a work context knowledge graph, and recommends the printing strategy to the user terminal, including: When there is no entity data combination in the personal work context knowledge graph that corresponds to the first entity data and the second entity data, locate the associated node associated with the target node; Retrieve from the database the printing strategies corresponding to the associated nodes related to the target node, as well as the edge weights between the target node and the associated nodes; By combining edge weights with a sorting strategy, the printing strategies corresponding to the associated nodes are sorted to obtain a set of printing strategy sequences. The printing strategies corresponding to the associated nodes are recommended to the user terminal according to the order of the printing strategies in the printing strategy sequence set. When there is a combination of entity data in the personal work context knowledge graph that corresponds to the first entity data and the second entity data, the node marked in the personal work context knowledge graph is searched. After identifying the marked nodes, the corresponding printing strategies for the marked nodes are retrieved from the database and recommended to the user terminal.

8. A personalized printing push system based on user behavior analysis, characterized in that, The personalized printing push system includes: A data acquisition module is used to collect printing data from user terminal devices in real time and send the printing data to the printing terminal; wherein, the printing terminal includes a printing cloud server and a printer; The entity extraction module is used by the printing cloud server of the printing terminal to perform data entity extraction processing on the printing data to obtain the printing entity data corresponding to the printing data. The graph update module is used to dynamically update the user's personal work context knowledge graph that has been built and deployed by the user through the entity coefficients corresponding to the printed entity data. The inference module is used by the printing cloud server to infer the printing strategy based on the printing entity data through the working context knowledge graph, and recommend the printing strategy to the user terminal. The print execution module is used for the print cloud server to wait for the policy confirmation instruction sent by the user terminal. When the print cloud server receives the user's policy confirmation instruction, it controls the printer to perform the printing operation according to the print policy confirmed by the user. The graph update module includes: retrieving first entity data and second entity data, and obtaining the entity coefficients corresponding to the first entity data and second entity data; mapping the first entity data and second entity data to the user's corresponding completed and deployed personal work context knowledge graph, wherein the first entity data is basic information and the second entity data is key general terms; Determine whether there is entity data in the personal work context knowledge graph that corresponds to the first entity data and the second entity data; When there is a combination of entity data in the personal work context knowledge graph that corresponds to the first entity data and the second entity data, the node corresponding to the entity data is marked. When there is no entity data combination in the personal work context knowledge graph that corresponds to the first entity data and the second entity data, then the node is created by combining the entity coefficients corresponding to the first entity data and the second entity data to obtain the dynamically updated personal work context knowledge graph corresponding to the user. Specifically, nodes are created by combining the entity coefficients corresponding to the first and second entity data, resulting in a dynamically updated personal work context knowledge graph for each user, including: Retrieve the combination of the first entity data and the second entity data corresponding to each node in the user's personal work context knowledge graph; The similarity between the combination of the first entity data and the second entity data corresponding to each node in the personal work context knowledge graph and the current combination of the first entity data and the second entity data is compared to obtain the similarity value between the combination of the first entity data and the second entity data corresponding to each node in the personal work context knowledge graph and the current combination of the first entity data and the second entity data. Retrieve the entity coefficients of the first and second entity data corresponding to each node in the user's personal work context knowledge graph. By combining the entity coefficients of the first entity data and the second entity data of each node in the knowledge graph of the user’s personal work context, and the similarity value of the combination of the first entity data and the second entity data of each node, the comprehensive strength value of each node and the current combination of the first entity data and the second entity data is obtained. The comprehensive strength value is compared with a preset strength threshold, and nodes with comprehensive strength values ​​not lower than the preset strength threshold are selected as associated nodes; a target node is created in the personal work context knowledge graph for the current combination of first entity data and second entity data, and a directed edge is established between the target node and the associated node; the entity coefficients corresponding to the first entity data and second entity data of the target node and the entity coefficients corresponding to the first entity data and second entity data of each associated node are retrieved; Based on the entity coefficients corresponding to the first and second entity data of the target node and the entity coefficients corresponding to the first and second entity data of each associated node, set the edge weight values ​​corresponding to the directed edges established between the target node and the associated nodes.

Images