Page configuration and dynamic display method and device, equipment and medium

Abstract

The invention relates to the technical field of data analysis, and discloses a page configuration and dynamic display method and device, equipment and a medium, and the method comprises the steps: receiving a page template file, and extracting page element information to form a page template resource library; generating a page display strategy through visual configuration interaction; collecting user portrait data and user behavior data, and inputting the data into the personalized recommendation model to obtain a page template combination; dynamically generating a page code based on the page display strategy and the page template combination, and rendering and displaying a target page; and recording interaction behavior data and feeding back the interaction behavior data to the personalized recommendation model and the page display strategy for updating and optimization. The method can be applied to business scenes such as financial science and technology, page display has continuous optimization capability through page template management, strategy configuration, user feature matching, dynamic generation and behavior feedback, and display flexibility and matching accuracy are improved.

Description

Technical Field

[0001] This invention relates to the field of data analysis technology, and in particular to a method, apparatus, device, and medium for page configuration and dynamic display. Background Technology

[0002] In the fintech sector, as insurance business gradually moves online, application pages generally use preset templates for display, with page structure, element layout, and displayed content fixed during the design phase. These pages present a highly consistent format across different users, making it difficult to tailor them to different customer groups. Users lack content that matches their individual needs, negatively impacting the application experience.

[0003] The current configuration of insurance application pages typically relies on front-end code modifications or manual editing of template files. The positional relationships of page elements and the display logic need to be implemented programmatically by technical personnel, making it difficult for business personnel to directly participate in page layout and rule configuration. When business strategies or market demands change, page updates require a development and deployment process, resulting in a long adjustment cycle and insufficient flexibility in page configuration.

[0004] Meanwhile, there is a lack of effective linkage between the page display logic and user profile and behavior data. Page content is determined before user access, making it impossible to dynamically match appropriate page templates or element styles based on user characteristics. Furthermore, user interactions on the page are not continuously recorded and fed back into the page configuration logic, making it difficult to continuously optimize page display strategies based on actual interaction data. Summary of the Invention

[0005] The main objective of this invention is to provide a method, apparatus, device, and storage medium for page configuration and dynamic display, aiming to solve the technical problem that existing page display methods fix the page template and display logic before display, making it impossible to dynamically generate and continuously optimize the page display content based on user profile data and user behavior data.

[0006] To achieve the above objectives, the present invention provides a page configuration and dynamic display method, comprising: The system receives page template files in various formats, each containing page element information, and performs categorized storage and version control operations on the page template files to generate a page template resource library. The page display strategy is generated by responding to configuration operations on the page element information through a visual configuration interface. Collect user profile data and user behavior data of the target user, input the user profile data and user behavior data into a preset personalized recommendation model to perform feature matching, and obtain a combination of page templates corresponding to the target user; Based on the page display strategy and the page template combination, page code is dynamically generated by calling page element information from the page template resource library, and rendering operations are performed to display the target page. Record the interactive behavior data generated by the target user on the target page, and feed the interactive behavior data back to the personalized recommendation model and the page display strategy to update the parameters of the personalized recommendation model and optimize the page display strategy.

[0007] Furthermore, to achieve the above objectives, the present invention provides a page configuration and dynamic display device, comprising: The page template management module is used to receive page template files in various formats, which contain page element information, and to perform classification, storage and version control operations on the page template files to generate a page template resource library. The strategy generation module is used to respond to configuration operations on the page element information through a visual configuration interaction interface and generate a page display strategy. The recommendation matching module is used to collect user profile data and user behavior data of target users, input the user profile data and user behavior data into a preset personalized recommendation model to perform feature matching, and obtain a combination of page templates corresponding to the target user. The page generation module is used to dynamically generate page code by calling page element information from the page template resource library according to the page display strategy and the page template combination, and to perform rendering operations to display the target page. The feedback optimization module is used to record the interaction behavior data generated by the target user on the target page, and feed the interaction behavior data back to the personalized recommendation model and the page display strategy to update the parameters of the personalized recommendation model and optimize the page display strategy.

[0008] Furthermore, to achieve the above objectives, the present invention also provides a computer device, the computer device including a memory, a processor, and a page configuration and dynamic display program stored in the memory and executable on the processor, wherein when the page configuration and dynamic display program is executed by the processor, it implements the steps of the page configuration and dynamic display method as described above.

[0009] Furthermore, to achieve the above objectives, the present invention also provides a computer-readable storage medium storing a page configuration and dynamic display program, wherein the page configuration and dynamic display program, when executed by a processor, implements the steps of the page configuration and dynamic display method as described above.

[0010] Beneficial Effects: This invention relates to the field of data analysis technology and discloses a method, apparatus, device, and medium for page configuration and dynamic display. The method includes: receiving page template files and extracting page element information to form a page template resource library; generating page display strategies through visual configuration interaction; collecting user profile data and user behavior data and inputting them into a personalized recommendation model to obtain page template combinations; dynamically generating page code based on the page display strategy and page template combinations and rendering the target page; recording interaction behavior data and feeding it back to the personalized recommendation model and page display strategy for updates and optimization. This invention can be applied to business scenarios such as fintech. Through page template management, strategy configuration, user feature matching, dynamic generation, and behavior feedback, it enables continuous optimization of page display, improving display flexibility and matching accuracy. Attached Figure Description

[0011] The present invention will be further described below with reference to the accompanying drawings and embodiments. In the accompanying drawings: Figure 1 This is a schematic diagram of an application environment for a page configuration and dynamic display method according to an embodiment of the present invention; Figure 2 This is a flowchart illustrating an embodiment of the page configuration and dynamic display method of the present invention; Figure 3 A schematic diagram of functional modules of a preferred embodiment of the page configuration and dynamic display device of the present invention; Figure 4 This is a schematic diagram of the structure of a computer device according to an embodiment of the present invention; Figure 5 This is another structural schematic diagram of a computer device according to one embodiment of the present invention. Detailed Implementation

[0012] It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of the invention.

[0013] The page configuration and dynamic display method provided in this embodiment of the invention can be applied to, for example, Figure 1In this application environment, the client communicates with the server via a network. The server can receive page template files from the client and extract page element information to form a page template resource library; generate page display strategies through visual configuration interaction; collect user profile data and user behavior data and input them into a personalized recommendation model to obtain page template combinations; dynamically generate page code and render the target page based on the page display strategy and page template combinations; record interaction behavior data and feed it back to the personalized recommendation model and page display strategy for updates and optimization. This invention can be applied to business scenarios such as fintech. Through page template management, strategy configuration, user feature matching, dynamic generation, and behavior feedback, it enables continuous optimization of page display, improving display flexibility and matching accuracy. The client can be, but is not limited to, various personal computers, laptops, smartphones, tablets, and portable wearable devices. The server can be implemented using a dedicated server or a server cluster consisting of multiple servers. The invention will be described in detail below through specific embodiments.

[0014] Please see Figure 2 , Figure 2 This is a flowchart illustrating an embodiment of the page configuration and dynamic display method provided by the present invention. It should be noted that although a logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than that shown here.

[0015] like Figure 2 As shown, the page configuration and dynamic display method proposed in this invention includes the following steps: S10, receive page template files in multiple formats, the page template files containing page element information, perform classification storage and version control operations on the page template files, and generate a page template resource library; In this embodiment, receiving page template files in multiple formats involves the unified access and processing of page description files generated from different design tools or front-end frameworks. These multiple formats include structured markup files, style files, and component description files. Page template files contain page element information, which is represented as independently identifiable and reusable page units, carrying position, size, and display attributes, providing a data foundation for subsequent management. Categorizing and storing page template files involves establishing category directories or tag indexes based on business or functional identifiers in the page element information, enabling different templates to be distinguishable at the storage layer. Version control involves generating unique version identifiers through file content verification values ​​and time information, and recording the template's change history. Generating a page template resource library involves uniformly registering the storage path, page element information, category identifiers, and version identifiers of page template files in an index structure, enabling template resources to be searchable and traceable.

[0016] Page template files can be stored in object storage, and page element information and index relationships can be stored in a database; alternatively, a tag system can be used for categorized management. Version identifiers can be generated using hash calculations combined with time information, or changes can be stored using difference records. Page element information can be extracted from the template file and written into the index structure by a parser to improve subsequent retrieval efficiency.

[0017] This embodiment classifies and manages the page element information in the page template file, enabling the page template to have clear structured management capabilities and historical traceability.

[0018] S20, respond to configuration operations for the page element information through a visual configuration interaction interface, and generate a page display strategy; In this embodiment, the visual configuration interface presents a graphical editing environment that showcases the page structure. Page element information is displayed as components, allowing the user to directly perceive the layout relationships of page units. Responding to configuration operations on page element information involves real-time capture of drag-and-drop, position adjustment, and attribute adjustment actions, transforming these interactions into structured data. Position information generated during configuration operations describes the coordinate relationships of page elements within the display area, while attribute information describes the display style and conditions of page elements. The arrangement and hierarchy relationships between page elements are recorded as the page layout structure through the configuration process. Association rules between the page layout structure and user attribute dimensions are established during configuration, forming a parsable and executable page display strategy.

[0019] A visual configuration interface can be built using a front-end editor, and drag-and-drop and attribute modification behaviors can be captured through an event listener mechanism. Alternatively, the coordinate and hierarchical parameters of page elements can be recorded using a real-time data structure, and the association rules can be persistently saved in the form of a configuration file.

[0020] This embodiment uses a graphical configuration method to lay out and set rules for page elements, enabling the page display rules to have intuitive configuration capabilities and structured expression capabilities.

[0021] S30, collect user profile data and user behavior data of the target user, input the user profile data and user behavior data into a preset personalized recommendation model to perform feature matching, and obtain a combination of page templates corresponding to the target user; In this embodiment, the target user refers to the entity currently initiating the access request, which can be uniquely associated with the access identifier and historical records. User profile data originates from a long-term accumulated set of attribute information, including stable attributes such as age range, interest tags, risk preferences, and usage frequency, as well as dynamic attributes such as recent activity and access frequency. User behavior data comes from operation trajectory records during the access process, reflecting the user's actual operational preferences and content focus in the current session. The collection process reads corresponding data from the user database and log system through a data interface, and performs standardized formatting, missing value imputation, and outlier removal on the read data to ensure its computability. User profile data and user behavior data are converted into a structured feature representation form using unified encoding rules, and then processed through feature vectorization to form an input feature vector. This input feature vector can fully express the comprehensive feature state of the target user at the current access time. A preset personalized recommendation model is used to establish the association between user features and page templates. The feature matching process involves calculating the similarity and evaluating the relevance between the input feature vector and the page template feature vector. The page template combination is obtained by sorting and filtering the matching results. It represents a set of page templates that match the characteristics of the target user. This set is used for template calls in the subsequent page generation process.

[0022] This embodiment uses matching calculations between user characteristics and page templates to enable page template selection to reflect the user's current real needs and improve the fit between page content and user preferences.

[0023] S40, based on the page display strategy and the page template combination, page code is dynamically generated by calling page element information from the page template resource library, and rendering operation is performed to display the target page; In this embodiment, the page display strategy describes the mapping relationship between the page layout structure and user profile data, indicating the arrangement and display rules of page elements on the page. The page template combination represents a set of page templates matching the target user characteristics, containing multiple callable template identifiers. The page template resource library stores categorized and version-controlled page template files and corresponding page element information, which are presented as independently callable structural units. The page layout structure is parsed according to the page display strategy, and the template identifier is parsed according to the page template combination. The corresponding page element information is retrieved and called from the page template resource library using the template identifier. The called page element information is embedded according to the coordinate positions and hierarchical order defined in the page layout structure, forming page code with a complete structural relationship. The page code is presented as structured text that can be parsed by the terminal, describing the complete content displayed on the page. The rendering operation transforms the page code into a visual interface. By parsing the page code and loading associated resources, the page elements are presented on the terminal interface according to the predetermined structure, ultimately forming the target page.

[0024] This embodiment combines page layout rules with template calling mechanisms to ensure structural consistency and content flexibility in the page code generation process, enabling the rapid creation of page display results that conform to user characteristics.

[0025] S50, record the interaction behavior data generated by the target user on the target page, and feed the interaction behavior data back to the personalized recommendation model and the page display strategy to update the parameters of the personalized recommendation model and optimize the page display strategy.

[0026] In this embodiment, the target user's clicks, pauses, scrolling, and inputs on the target page are continuously collected and form interaction behavior data, which is a structured record with time and page position identifiers. This interaction behavior data is used to depict the target user's actual browsing trajectory and content of interest on the target page. The collected interaction behavior data is transmitted to the processing module and organized and categorized according to a preset data structure, enabling it to directly participate in model calculations and rule analysis. After receiving the interaction behavior data, the personalized recommendation model extracts and transforms features to obtain feedback features that reflect changes in user preferences. Based on these feedback features, the parameters of the personalized recommendation model are adjusted, changing the model's ability to match page template combinations. The page display strategy also receives interaction behavior data, statistically analyzes the attention levels of page elements and page layout usage reflected in the data, and adjusts the mapping relationship between the page layout structure and user profile data based on the statistical results, thereby changing the display rules during subsequent page generation.

[0027] This implementation, by collecting and utilizing real interactive behaviors, enables the model matching capability and page display rules to be dynamically adjusted according to changes in user behavior, thereby improving the consistency between the page display content and the user's focus.

[0028] In one embodiment, step S10 above includes: S101, receive page template files of various formats uploaded through the file transfer interface, and perform file format validity verification on the page template files to obtain files that pass the verification; S102, the parser is used to identify and extract page element information from the verified file, and the business category metadata is determined based on the page element information; S103, match the corresponding archive directory according to the business classification metadata, write the verified file into the archive directory, and obtain the storage path of the verified file in the archive directory; S104, determine the content verification code of the verified file, generate a unique version identifier based on the content verification code and the file upload timestamp, and record the change log corresponding to the verified file; S105, construct an index database containing the storage path, the unique version identifier, the page element information, and the change log, and generate a page template resource library.

[0029] In this embodiment, the file transfer interface serves as the upload channel for page template files. The various formats of page template files are manifested in different file encapsulation methods and syntactic organization methods. When the file transfer interface receives uploaded page template files of various formats, it writes the page template file along with basic attributes such as the upload session identifier, transmission verification information, and file size into the receiving buffer to support subsequent verification processes in checking the integrity of the received data. File format validity verification revolves around the format identifier, syntactic structure, and key paragraph integrity of the page template file. The format identifier is used to determine if the page template file belongs to a set of parsable target formats. Syntactic structure verification is used to determine whether the internal tags, tag levels, and quoted paragraphs of the page template file meet the parsing constraints of the parser. Key paragraph integrity verification is used to determine whether there are any missing or truncated core paragraphs related to page element information in the page template file. Files that pass verification are fixed as subsequent parsing objects. In subsequent processing, files that pass verification serve as input sources for the extraction and storage of page element information, ensuring that subsequent extraction and storage are not interfered with by files that fail verification.

[0030] The parser performs structured recognition on verified files. The recognition process focuses on locating renderable components, style definitions, interaction binding descriptions, and resource references within the template. The extraction process extracts this content as page element information in a field-based format. Page element information includes descriptions of combinable elements, sets of element attributes, element hierarchy relationships, and resource reference identifiers, enabling it to support both classification and indexing. Business category metadata is derived from the page element information. This metadata expresses the archive category corresponding to the page template file. The source of business category metadata can be business tags declared in the page element information, or element type distribution and page structure characteristics. Once formed, the business category metadata serves as one of the primary keys for archiving and retrieval in subsequent directory matching.

[0031] A mapping table is established between the archive directory and business category metadata. This table maps business category metadata to specific archive directory paths. The matching process determines the archive directory based on the match results of the business category metadata in the mapping table. When a validated file is written to the archive directory, persistent writing and permission attribute assignment are performed on the written object. Persistent writing is used to create a stable storage copy, and permission attribute assignment is used to limit the subsequent access subjects and access methods. The storage path is used to identify the location of validated files in the archive directory. The storage path is used as a location field in the subsequent index database, enabling page template files to be directly located and loaded.

[0032] The Content Checksum (CQS) is used to represent the content fingerprint of a verified file. The CQS is calculated from the file content and is sensitive to changes in the file content, used to distinguish different content versions. The file upload timestamp represents the time when the upload occurred, serving as a time-dimension distinguishing factor in version generation. The Unique Version Identifier (UPI) is generated by combining the CQS and the file upload timestamp. The UPI uniquely identifies version instances of the same page template file uploaded at different times or with different content states. The Change Log records version change information corresponding to the verified file. The Change Log includes at least the UPI, the change time, the change type identifier, and the associated storage path, enabling version evolution to be tracked and audited.

[0033] The index database is constructed with storage path, unique version identifier, page element information, and change log as core fields. Its data structure is designed around querying and locating: the storage path locates the page template file entity, the unique version identifier locates the version instance, page element information supports element-level retrieval and category filtering, and the change log supports version rollback and difference tracking. After generation, the index database is associated with the archive directory. The index database provides the retrieval entry point, while the archive directory provides file entity storage. Together, they constitute the page template resource library, enabling it to simultaneously store template files, locate versions, perform element-level retrieval, and track changes.

[0034] This embodiment excludes page template files that do not meet parsing constraints from subsequent processes by validating file format, reducing the risk of parsing errors and resource library pollution; it extracts page element information and generates business category metadata through a parser, giving page template files searchable semantic tags and a structured content foundation; it achieves location-based storage of template files through archive directories and storage paths, enabling subsequent calls to directly obtain stable file entry points; it distinguishes and locates versions through content verification codes, file upload timestamps, and unique version identifiers, and forms a traceable version evolution record in conjunction with change logs; and it unifies storage paths, unique version identifiers, page element information, and change logs into a search structure through an index database, enabling the page template resource library to simultaneously support category search, version backtracking, and template location, improving the controllability and scalability of template management.

[0035] In one embodiment, step S20 above includes: S201, Render the visual configuration interaction interface, and load and display the page element information in the component library area of ​​the visual configuration interaction interface; S202, detect drag-and-drop layout actions and attribute adjustment actions in the canvas area of ​​the visual configuration interaction interface for the page element information, and identify the drag-and-drop layout actions and attribute adjustment actions as configuration operations. S203, determine the coordinate position parameters and stacking order of page elements according to the configuration operation, and generate the page layout structure; S204, responding to the logical binding instruction for the page layout structure, select the attribute dimension of the user profile data and set the display threshold condition, and establish a mapping relationship between the user profile data and the page layout structure; S205, the page layout structure and the mapping relationship are encapsulated into a structured configuration file to generate a page display strategy.

[0036] In this embodiment, the visual configuration interface is used to carry out page configuration input and configuration result expression. The interface rendering action corresponds to interface resource loading, interface layout initialization, and interactive area registration. The component library area carries the presentation of page element information. The page element information is organized into optional component items. Each item contains an element identifier, element type, default attribute set, and editable attribute set. The default attribute set is used for initial display, and the editable attribute set limits the range of attributes that can be modified. When the component library area loads page element information, it reads the data items of the page element information and performs structured expansion to generate a list of component items. At the same time, it establishes an index relationship between component items and page element information for subsequent drag-and-drop layout actions to locate target elements. When the component library area displays page element information, it renders the visual representation of the items and attaches an interaction handle. The interaction handle is used to trigger drag-and-drop layout actions and attribute adjustment actions.

[0037] The canvas area is used to display the page layout editing state, detecting and merging events corresponding to drag-and-drop layout actions and attribute adjustment actions. Drag-and-drop layout actions are identified through a sequence of pointer press, move, and release events. This sequence includes the starting position, movement trajectory, and release position, with the release position determining the placement area of ​​page elements. Attribute adjustment actions are identified through attribute panel input events. These events include attribute keys, attribute values, and submission status. Attribute keys are used to locate the attribute category, and attribute values ​​express the new attribute state. Recognizing drag-and-drop layout actions and attribute adjustment actions as configuration operations involves uniformly encapsulating the event sequences and attribute panel input events. The encapsulation result includes page element identifiers, operation type identifiers, operation parameter sets, and operation time information. The operation parameter set includes target coordinate information and target container information for drag-and-drop layout actions, and a set of attribute key-value pairs for attribute adjustment actions. Configuration operations are written to the configuration operation queue or configuration operation record area for subsequent layout structure generation calculations.

[0038] Coordinate position parameters express the spatial position of page elements in the canvas coordinate system. The coordinate system can be relative or absolute; relative coordinates are referenced to the parent container, while absolute coordinates are referenced to the canvas origin. Stacking order expresses the occlusion and rendering order of page elements. Stacking order can be expressed using hierarchical values ​​or ordered lists. Hierarchical values ​​are used for quick comparisons, while lists represent stable orders. Determining coordinate position parameters and stacking order based on configuration operations involves parsing information such as target placement, alignment, and hierarchy adjustment instructions from the configuration operations. Placement is converted into coordinate position parameters, alignment into constraint parameters, and hierarchy adjustment instructions into updated stacking order. Page layout structure expresses the overall organizational relationship of page elements within the page. The page layout structure includes a set of page element nodes and a set of node relationships. The node set consists of page element identifiers, coordinate position parameters, size parameters, and attribute parameters. The node relationship set consists of parent-child relationships, adjacency relationships, and stacking order. When the page layout structure is generated, the configuration operations in the configuration operation record area are aggregated according to the page element identifier to generate the final parameter status of each page element, and the parameter status is written into the data object of the page layout structure.

[0039] Logical binding instructions are used to establish associations between the attribute dimensions of user profile data and page element nodes or sets of nodes in the page layout structure. These instructions are generated by user interface interaction and include target node selection information, attribute dimension selection information, and display threshold condition configuration items. The attribute dimensions of user profile data describe the value space of user characteristic fields. Attribute dimensions can correspond to category dimensions, interval dimensions, or Boolean dimensions. Category dimensions are represented by an enumeration set, interval dimensions by interval boundaries, and Boolean dimensions by true / false states. Display threshold conditions describe the set of judgment conditions that trigger display. This set consists of comparison operators, target thresholds, and combination relations. Comparison operators limit the judgment type, target thresholds limit value boundaries or value sets, and combination relations limit the conjunction or disjunction of multiple conditions. Establishing the mapping relationship between user profile data and the page layout structure is manifested by binding attribute dimensions and display threshold conditions to target nodes in the page layout structure. The mapping relationship records include attribute dimension identifiers, display threshold conditions, target node identifiers, and scope identifiers. The scope identifier limits the mapping relationship to acting on a single node or a set of nodes.

[0040] The structured configuration file is used to solidify the page layout structure and mapping relationships. The configuration file uses a field-based structure, which includes layout field paragraphs and mapping field paragraphs. The layout field paragraphs contain the set of nodes and node relationships of the page layout structure, while the mapping field paragraphs contain the set of record items for the mapping relationships. Encapsulated actions correspond to serialization processing and consistency checks. Consistency checks verify whether the node identifiers in the page layout structure match the target node identifiers in the mapping relationships, and also verify whether the attribute dimension identifiers referenced by the display threshold conditions exist in the user profile data field definitions. The page display strategy is formed by the structured configuration file. During subsequent execution, the page display strategy can be parsed into layout constraints and display conditions, thereby driving the combination and display of page elements.

[0041] This embodiment loads and displays page element information in the component library area, detects drag-and-drop layout actions and attribute adjustment actions in the canvas area and forms configuration operations, determines coordinate position parameters and hierarchical stacking order based on configuration operations to generate page layout structure, writes the attribute dimensions and display threshold conditions of user profile data into the mapping relationship under the drive of logical binding instructions, and encapsulates the page layout structure and mapping relationship into a structured configuration file to form a page display strategy. The page display strategy obtains a parsable layout expression and condition expression, the configuration result can be verified for consistency and saved in a structured form, and the generation process of the page display strategy has a traceable parameter source and a stable data organization form.

[0042] In one embodiment, step S30 above includes: S301, identify the target user who initiated the access request, extract user profile data associated with the target user from the user database, and read the user behavior data of the target user from the log system; S302, Perform feature vectorization processing on the user profile data and the user behavior data to construct an input feature vector representing the characteristics of the target user; S303, input the input feature vector into a preset personalized recommendation model, and use the personalized recommendation model to determine the matching probability value between the input feature vector and each template in the page template resource library; S304, sort and filter each template in the page template resource library according to the matching probability value, and use the filtered templates as the page template combination corresponding to the target user.

[0043] In this embodiment, the target user is identified by the request identifier of the access request. The request identifier includes at least one of the following: session identifier, terminal identifier, account identifier, and device fingerprint information. The request identifier carried in the access request is used to locate the user record of the target user in the user database. The user database is used to persistently store user profile data associated with the target user. The user profile data is used to express the target user's stable attributes and inductive attributes. Stable attributes include fields such as age range, regional information, language preference, and account level. Inductive attributes include fields such as preference tag set, historical selection preferences, and risk preference segmentation. When extracting user profile data, field selection and field validity checks are performed on the user profile data. Field selection is determined based on the input field definition of the personalized recommendation model. Field validity checks include missing field identification, field format checks, and field value range checks. Missing fields are filled with default values ​​or supplemented by backtracking through historical records. Field format checks are used to unify character encoding, time format, and enumeration encoding. Field value range checks are used to filter outliers and record the source of correction.

[0044] Personalized recommendation models receive input feature vectors and output matching probability values. The model structure can be divided into three parts: an input encoding layer, a feature fusion layer, and a matching output layer. The input encoding layer maps different fields in the input feature vector to a unified representation space. The input feature vector typically contains both user profile fields and user behavior fields. User profile fields include enumerated and continuous fields. Enumerated fields are converted into dense vectors through an embedding mapping table, while continuous fields are normalized and mapped to continuous values, which are then concatenated with the dense vector. User behavior fields include event sequence summaries, frequency statistics, and dwell time statistics. Event sequence summaries are converted into fixed-dimensional vectors through a sequence encoding unit. The sequence encoding unit can employ gated recurrent units, convolutional sequence units, or attention sequence units. The output of the sequence encoding unit is aligned with the dense vector of the user profile in the same dimensional space. The input encoding layer outputs a user representation vector, which serves as the main input to the subsequent fusion layer.

[0045] The feature fusion layer fuses the user representation vector and the template representation vector. The template representation vector comes from the template feature vectors of various templates in the page template resource library. The template feature vector is derived from the page element information of the template. The derivation process converts the page element type distribution, layout structure summary, theme style fields, etc., into template-side input vectors, which are then mapped to template representation vectors by the template encoder. The fusion method can adopt a dual-branch encoding structure, where the user encoder outputs the user representation vector and the template encoder outputs the template representation vector. The two generate a matching representation vector through an interaction unit. The interaction unit can use vector concatenation and multilayer perceptron, or a bidirectional gating interaction unit, or an attention interaction unit to enhance the selective influence of user behavior fields on template element fields. The feature fusion layer outputs a matching representation vector, which retains the joint information of user profile fields, user behavior fields, and template element fields.

[0046] The matching output layer maps the matching representation vector to matching probability values. It contains linear mapping units and probability mapping units. The linear mapping unit outputs the matching score, while the probability mapping unit compresses the matching score to a range of probability values ​​and forms the matching probability values. To ensure comparability of matching probability values ​​across different templates, a temperature coefficient or a piecewise mapping unit can be added to the probability mapping unit. The temperature coefficient controls the sharpness of the distribution, and the piecewise mapping unit reduces the compression of the output range by extreme samples. The matching probability values ​​obtained at the output end correspond one-to-one with each template in the page template resource library, forming a candidate set. Subsequently, page template combinations can be selected based on the matching probability values.

[0047] The training process revolves around sample construction, target definition, parameter updates, and evaluation and validation. Sample construction is based on historical access records. Each sample contains an input feature vector and a template identifier. Positive samples come from the target user's interaction behavior data such as clicks, dwell times, and submissions related to templates on the target page. Negative samples come from the set of templates exposed but not triggered within the same time window or the set of unselected templates from similar users. The number of negative samples can be set to a multiple of the number of positive samples to stabilize the training distribution. Target definition adopts either binary classification or ranking objectives. Binary classification objectives define interaction behaviors as positive or negative labels, while ranking objectives form relative preference pairs between multiple templates within the same session. During training, the training set, validation set, and test set are time-segmented. Time segmentation is used to avoid information leakage, the validation set is used to select the model version, and the test set is used for final evaluation. Parameter updates use either batch updates or incremental updates. Batch updates are used for offline retraining, while incremental updates are used to quickly absorb new samples when the distribution of interaction behaviors changes. The training process needs to record the model version identifier, training data version identifier, and feature dictionary version identifier. The feature dictionary version identifier is used to constrain the field order of the input feature vectors and the consistency with the embedding mapping table.

[0048] In fintech business scenarios, page displays are typically organized around product explanations, risk warnings, protection terms, rights descriptions, and insurance information entry. The model needs to be able to differentiate user preferences regarding information density, prompt intensity, and interaction complexity. Model input data can be a combination of user profile field sets and user behavior field sets. The user profile field set includes fields such as account level, risk preference segments, common terminal types, common language preferences, and historical purchase preference tags. The user behavior field set includes fields such as page browsing identifier sequences, key control trigger sequences, dwell time distribution, back and jump sequences, and entry interruption points. To ensure stable alignment of the input with domain semantics, the input field set needs to define field encoding methods and missing data handling methods. Encoding methods map categorical fields to fixed-dimensional expressions and continuous fields to comparable interval expressions. Missing data handling fills in uncollected fields with default states and adds validity markers. Model output data can be a template matching result set, containing the correspondence between template identifiers and matching scores. Matching scores are used to compare the suitability of different templates for the current user, and template identifiers are used to locate template element sets in the template resource library. The way the output is consumed by the business can be written as selecting a template combination based on the matching result set and using it for page generation. The matching result set can also be recorded as a recommendation decision record for subsequent auditing and backtracking.

[0049] The log system stores user behavior data for the target users. This data represents the observable behaviors of the target users during interactions. User behavior data includes event items such as page access events, click events, scrolling events, dwell events, and submission events. Each event item contains fields such as event type, event timestamp, page identifier, element identifier, and interaction parameters. When reading user behavior data, log records are filtered based on the target user's request identifier and a time window. The time window is determined by the occurrence time of the access request. The filtered results are sorted along the event timestamp dimension, and the sorted results are used to form a behavior sequence that can be used for feature extraction. After reading the behavior sequence, deduplication and anomaly filtering are performed. Deduplication merges duplicate reported events based on event identifier or timestamp thresholds. Anomaly filtering removes or corrects incomplete fields, invalid page identifiers, and abnormal timestamps. Correction actions are recorded in the data quality tag field, which is used for weight adjustment or missing data handling during subsequent feature vectorization processing.

[0050] Feature vectorization is used to convert user profile data and user behavior data into a unified numerical representation. Feature vectorization of user profile data includes discrete field encoding, continuous field normalization, and multi-valued field aggregation. Discrete field encoding maps enumerated values ​​to dense vectors or one-hot vectors; continuous field normalization maps interval values ​​to standardized values; and multi-valued field aggregation maps the tag set to a fixed-dimensional statistical vector. Feature vectorization of user behavior data includes event sequence binning statistics, time decay weighting, and sequence summary generation. Binning statistics maps the frequency and duration of event types within a time window to statistical features; time decay weighting assigns higher weights to behaviors closer to the access request; and sequence summary generation compresses the behavior sequence into a fixed-length representation, which can include fields such as the type code of the most recent event, the type distribution of several recent events, and the number of visits to key page identifiers. The input feature vector is used to represent the characteristics of the target user. The input feature vector is composed of user profile vector and behavior vector according to a preset concatenation rule. The concatenation rule includes dimension alignment, missing position padding and fixed field order. Dimension alignment is used to ensure that the input dimensions received by the personalized recommendation model are consistent. Missing position padding is used to keep the vector length stable in the case of missing fields. Fixed field order is used to ensure that the position of the same field in the input feature vector is constant to support the stabilization of model parameters.

[0051] The personalized recommendation model estimates the matching degree between the input feature vector and each template in the page template resource library. The model receives the input feature vector as input and generates matching probability values ​​corresponding to each template in the library as output. The matching probability value expresses the similarity or selection bias between the input feature vector and the template features. The matching probability value can be a normalized probability distribution or a comparable confidence score. The normalized probability distribution satisfies the summation constraint, and the confidence score satisfies the monotonic comparability constraint. Before participating in matching, each template in the page template resource library forms a template feature representation. This representation is derived from at least one of the following: page element information extracted from the page template file, business classification metadata, and historical template performance information. The template feature representation and the input feature vector are compared in the same feature space. The comparison result is mapped to a matching probability value, forming a list of matching probability values. The index of this list corresponds one-to-one with the template identifier in the page template resource library to support subsequent sorting and filtering.

[0052] The sorting and filtering function selects a set of templates from the page template resource library that meet the display requirements. Sorting is based on the matching probability value, and the sorting rules include descending order and stable sorting. Stable sorting maintains a fixed order of template identifiers to reduce display fluctuations when matching probabilities are the same. Filtering criteria include threshold filtering and quantity filtering. Threshold filtering removes templates with matching probabilities below a preset threshold, while quantity filtering retains a preset number of the highest-matching templates. Threshold and quantity filtering can be used together to simultaneously meet quality and scale constraints. The filtering results are output as a set of template identifiers or a set of template objects and named a page template combination. The page template combination represents the template selection results corresponding to the target user. Each template in the page template combination corresponds to a template identifier in the page template resource library, used for subsequent retrieval of page element information and generation of page code based on the template identifier.

[0053] This embodiment identifies target users, extracts user profile data from the user database, reads user behavior data from the log system and forms a behavior sequence. After field validation and deduplication filtering, the user profile data and user behavior data are processed into feature vectorization to construct an input feature vector. The input feature vector is input into a personalized recommendation model to obtain a matching probability value corresponding to each template in the page template resource library. The matching probability value is used to drive each template in the page template resource library to complete sorting and filtering and output a page template combination. The page template combination obtains a template selection result consistent with the target user's characteristics, while maintaining stable input dimensions and a clear correspondence with template identifiers, thereby providing a reusable selection basis for subsequent page element information calls and page display.

[0054] In one embodiment, step S40 above includes: S401, parse the page display strategy, obtain the page layout structure, and parse the page template combination to obtain the template identifier; S402, retrieve and call the corresponding page element information from the page template resource library according to the template identifier; S403, embed the called page element information into the corresponding position in the page layout structure to generate page code; S404, The rendering engine is used to parse the page code and load associated resources to perform rendering operations and generate the rendered page; S405, the rendered page is displayed as the target page on the user terminal.

[0055] In this embodiment, the page display strategy carries the configuration constraints of the page layout structure and page element information. Parsing the page display strategy refers to the semantic decoding and structured reconstruction of the configuration file. Semantic decoding covers the parsing of field names, field types, value ranges, and constraint expressions. Structured reconstruction transforms layout-related fields into the page layout structure. The page layout structure is used to express the spatial carrying relationship of page element information. The page layout structure includes at least a layout container identifier, container geometric parameters, container hierarchical order, and placeholder rules. Geometric parameters are expressed as coordinate position parameters and size parameters, hierarchical order is expressed as hierarchical stacking order, and placeholder rules are used to define the target areas where page element information can be embedded and the conflict handling method. The page template combination is used to express the template selection result. Parsing the page template combination refers to the decoding and splitting of the combination data. The splitting yields a set of template identifiers and combination order information. The template identifiers serve as search keys for locating the page template resource library, and the combination order information is used to constrain the splicing order, area allocation, or coverage priority when multiple templates coexist.

[0056] The page template resource library centrally stores index data of page template files and page element information. Retrieval from the page template resource library based on template identifiers points to queries and filters in the index database. Query conditions must include at least the template identifier and version conditions. Version conditions limit the set of available versions corresponding to the template identifier, and filtering operations remove records that do not meet state constraints or compatibility constraints. The page element information is then called to extract and load the search results, extracting and covering the structure fields and resource reference fields of the page element information. Structure fields describe component types, attribute sets, and event binding definitions, while resource reference fields form a list of associated resources. After the page element information is loaded, the assembly stage begins. The assembly stage merges the page element information according to template identifiers and assembly order information. Merging rules handle conflicts between components with the same name, attribute overriding, and default value filling. Default value filling completes the renderable attribute set when some attribute fields are missing from the page element information.

[0057] Embedding page element information into the page layout structure binds page element information to the target location within the layout structure. The target location is determined by the layout container identifier and coordinate position parameters. The embedding process includes position matching, hierarchical writing, and constraint validation. Position matching determines the correspondence between page element information and the layout container identifier. Hierarchical writing writes the page element information into the hierarchical stacking order of the page layout structure. Constraint validation checks whether the size parameters of the page element information and the geometric parameters of the layout container meet the placeholder rules. If constraint validation fails, an adjustment strategy is triggered, including fine-tuning of coordinate position parameters, scaling of size parameters, or replacement with placeholder components. The choice of adjustment strategy is determined by the constraint fields in the page display strategy. Page code generation serializes the embedded page layout structure and page element information into executable page description text. The serialization process outputs a page structure segment, a style segment, and a behavior segment. The page structure segment expresses the hierarchical structure and placeholder relationships, the style segment expresses visual and layout attributes, and the behavior segment expresses interactive events and trigger actions. The resource reference fields in the page code simultaneously form an associated resource list, which is used for subsequent loading.

[0058] The rendering engine parses the page code, performing syntax parsing and structural construction. Syntax parsing breaks down the page structure into node sets and parent-child relationships, while structural construction maps the node sets to internal tree structures and preserves the binding references between style and behavior segments. Loading associated resources involves locating, acquiring, and validating resources based on the associated resource list. Resource location parses resource identifiers and paths; acquisition covers both local reads and network requests; and validation covers resource integrity and version consistency checks. Version consistency checks prevent rendering discrepancies caused by mismatches between associated resources and page code descriptions. After loading associated resources, the rendering engine performs rendering operations, including style calculation, layout calculation, and drawing submission. Style calculation merges style segments with style definitions from associated resources to obtain the final style set for each node. Layout calculation calculates the final coordinate position and size parameters of nodes based on the geometric parameters of the page layout structure. Drawing submission converts the visual attributes of nodes into drawing instructions and forms a layer set. The layer set enters compositing, where layers are overlaid and clipped according to their stacking order. Clipping rules are provided by the page layout structure's placeholder rules. The compositing result is the rendered page.

[0059] The rendered page, displayed on the user's terminal, refers to the display context in which the rendered page is output to the user's terminal. The display context is determined by the user's terminal's window identifier and display area parameters. The display action includes frame submission and display refresh. Frame submission submits the composited layer set to the user's terminal's display buffer, while display refresh drives the user's terminal to render the content of the display buffer onto the screen area. The display action also preserves event bindings for page element information, keeping the target page interactive on the user's terminal. This interactive state is constrained by the event types and response actions defined in the behavior section.

[0060] This embodiment obtains the page layout structure by parsing the page display strategy and obtains the template identifier by parsing the page template combination, so that the page generation process has a reusable layout expression and a searchable template positioning basis. The page element information is retrieved and called from the page template resource library through the template identifier, and then the page element information is embedded into the page layout structure to generate page code, so that the page structure, style and behavior can be assembled according to configuration constraints and form an executable description. The rendering engine parses the page code and loads the associated resources to perform rendering operations to generate the rendered page, and then displays the rendered page on the user terminal as the target page, so that the configuration result can be stably converted into a visual presentation and maintain interactive usability.

[0061] In one embodiment, step S404 includes: S4041, The rendering engine is used to perform lexical and syntactic analysis on the page code, construct a document object model tree, and identify the associated resources referenced in the page code; S4042, Initiate a resource acquisition request based on the resource attribute type of the associated resource to load the associated resource; S4043, During the loading of associated resources, for associated resources that are static resources, the local cache system is searched, and if the search is successful, the associated resources are read. S4044, During the loading of associated resources, for associated resources that are not on the first screen, register a visible area listening event, and perform asynchronous loading when the associated resources enter the visible area of ​​the screen; S4045, Construct a rendering tree based on the document object model tree and the loaded associated resources, determine the geometric layout coordinates of the page elements based on the rendering tree, and perform drawing; S4046 performs composite processing on the drawn layers to generate the rendered page.

[0062] In this embodiment, the rendering engine is used to convert page code into a displayable rendered page. The rendering engine can be deployed in a terminal-side runtime environment or an embedded runtime environment. The rendering engine includes at least one of the following: a parsing submodule, a resource scheduling submodule, a layout calculation submodule, a drawing submodule, and a compositing submodule. Page code is used to describe page structure, style references, and script references. Page code can be a combination of markup language text, script fragments, and style references. The reference information carried in the page code is used to locate associated resources. Parsing the page code employs lexical analysis and syntax analysis. Lexical analysis is used to segment the character stream into tag units and identify tag names, attribute names, attribute values, and text nodes. Syntax analysis is used to construct a tree structure based on the hierarchical nesting relationship of the tag units and to verify syntax constraints. The Document Object Model (DOM) tree is used to represent the node hierarchy and node attribute set of the page structure. The node attribute set includes at least one of the following: node type, attribute key-value pairs, text content, and parent-child relationship pointers. The construction process of the DOM tree includes node creation, node attachment, and node attribute writing. Node attachment is completed based on a stack-based matching of start and end tags, and node attribute writing is completed based on attribute name resolution and attribute value decoding. Associated resources are used to represent external objects referenced in the page code. Associated resources include at least one of the following: style files, script files, font files, image files, audio / video files, and data fragment files. Associated resources are identified through reference tags in the page code. Reference tags include at least one of the following: link reference tags, script reference tags, media reference tags, or resource preloading tags. The identification result forms an associated resource list, which includes at least one of the following fields: resource address, resource type, resource priority, and dependency relationship. Resource priority is determined by the attribute fields of the reference tag, page visibility requirements, or rendering blocking rules. Dependency relationships are used to represent the loading order constraints of scripts and styles, as well as the pre-required resource constraints for script execution.

[0063] Resource attribute types are used to distinguish the scheduling methods of different associated resources. Resource attribute types are determined by a combination of file type, reference attribute, blocking attribute, and visibility attribute. The file type originates from the resource address extension or the media type field in the response header. The reference attribute originates from the attribute field carried by the reference tag in the page code. The blocking attribute indicates whether the resource affects layout calculation or rendering timing. The visibility attribute indicates whether the page element corresponding to the resource is in the initial visible area. Resource acquisition requests are used to obtain the content entity of associated resources from resource providers. Resource providers include at least one of the following: local file system, built-in resource package, cache storage medium, or network server. Resource acquisition requests include at least one of the following: resource address, request header field, cache verification field, and priority field. Loading associated resources is executed through the resource scheduling submodule. The resource scheduling submodule allocates loading queues to associated resources based on resource attribute types. Loading queues include at least one of the following: blocking queues and non-blocking queues. Blocking queues ensure the availability of style resources and critical script resources required before layout calculation, while non-blocking queues improve the speed of the first screen rendering and delay the acquisition of non-critical resources. A local caching system is introduced during loading. This system stores the content entities and metadata of associated resources. The local caching system can include at least one of the following: memory caching, persistent caching, or a hierarchical caching structure. The hierarchical caching structure prioritizes low-overhead storage media before searching high-capacity storage media. Cache retrieval is performed on associated resources that are static resources. Static resources represent resource objects that are frequently reused across pages and have low update frequency. The identification criteria for static resources include at least one of the following: common path characteristics of the resource address, resource version tag characteristics, or historical hit statistics. Cache retrieval locates cache entries using resource address and version information. Version information comes from at least one of the following: version parameters carried in the resource address, entity tags carried in the resource response header, or content checksum fields. Retrieval hit conditions include the existence of the cache entry, version consistency, and expiration date. After a successful retrieval, the content entity of the associated resource is read, and the result is populated back into the resource table of the resource scheduling submodule. The resource table provides directly referenceable resource handles for subsequent layout calculations and drawing.

[0064] Asynchronous loading is performed on associated resources that are not part of the initial screen display. These non-initial screen resources represent resource objects outside the initial rendering area or resource objects in a delayed display area. Identification criteria for non-initial screen resources include at least one of the following: the initial position of the page element in the page layout structure, the visibility attribute of the page element, or the preset attribute fields of a placeholder element. The visible area listener event triggers resource scheduling updates when scrolling, scaling, or layout changes cause changes in the visible area boundary. This event is registered by the rendering engine to the event system or observer system. The observer system monitors the intersection state of the bounding box of a page element with the visible area boundary. Entering the visible area indicates that the bounding box of a page element meets the intersection threshold condition with the visible area boundary. This threshold condition can be defined by the intersection area ratio or the boundary distance threshold. When the intersection threshold condition is met, a resource acquisition request is submitted to the resource scheduling submodule, and network retrieval or local retrieval is performed in an independent task queue. This independent task queue is used to avoid blocking the main execution thread of the rendering engine. After asynchronous loading is complete, the content entity of the associated resource is written to the local cache system and the resource table is updated. The update action includes resource handle update, availability status mark update and dependency status update. The dependency status update is used to allow subsequent script execution or style recalculation to be triggered.

[0065] The rendering tree represents the set of visible nodes participating in rendering. It is jointly determined by the document object model tree and the loaded associated resources. The style content in the loaded associated resources is used to calculate the style attribute set of the nodes. The style attribute set includes at least one of the following: display attributes, box model attributes, font attributes, color attributes, and cascading order attributes. Display attributes determine whether a node enters the rendering tree, box model attributes determine the node size calculation rules, and cascading order attributes determine the layer allocation and composition order. Building the rendering tree involves style matching, style calculation, and visible node filtering. Style matching determines the applicable rule set based on selector rules and the node attribute set. Style calculation outputs the final style attribute set based on cascading and inheritance rules. Visible node filtering excludes invisible nodes and nodes not participating in rendering to reduce the rendering load. Geometric layout coordinates are used to express the position and size of page elements in the rendering space. Geometric layout coordinates include at least one of the following: horizontal position, vertical position, width, height, and baseline information. Geometric layout coordinates are calculated by the layout calculation submodule based on the rendering tree and style attribute set. Layout calculation includes measurement, constraint solving, and layout result backfilling. Measurement is used to determine the inherent size of text and replacement elements. Constraint solving is used to handle parent-child layout relationships, floating rules, and positioning rules. Layout result backfilling is used to write the geometric layout coordinates into the rendering tree node for the drawing submodule to read.

[0066] The drawing process converts render tree nodes into a sequence of displayable graphics instructions or pixel buffers. The drawing submodule reads the render tree and geometric layout coordinates to generate a drawing list. This list contains at least one of the following fields: drawing order, clipping region, drawing style, and image sampling parameters. Drawing can be performed in layers to support subsequent compositing. Layered drawing assigns nodes with independent transformation, opacity, or animation properties to separate layers, which are stored as off-screen buffers or texture objects. Compositing combines the drawn layers according to their stacking order and blending rules to create the final display result. Blending rules include at least one of opacity blending, masking rules, and transformation matrix applications. The compositing output forms the rendered page, which can be represented as framebuffer content, texture compositing results, or a handle to a visual output object. The rendered page is then provided to the terminal display component for display refresh.

[0067] This embodiment uses a rendering engine to perform lexical and syntactic analysis on the page code and construct a document object model tree. The page structure obtains a computable node hierarchy representation, and the associated resources referenced in the page code are identified as a schedulable resource list. By initiating resource acquisition requests based on resource attribute types and performing priority retrieval of static resources in the local cache system, the overhead of repeated loading is reduced, and the resource availability status can be quickly backfilled. By registering visible area listener events for non-first-screen resources and performing asynchronous loading when entering the visible area of ​​the screen, the rendering tree construction and geometric layout coordinate calculation obtain more controllable resource readiness conditions. By constructing a rendering tree based on the document object model tree and the loaded associated resources and performing drawing and compositing processing, the rendered page forms a stable visual output result. The structure, style, and resource dependencies required for page display are organized and updated within the same execution flow.

[0068] In one embodiment, step S50 above includes: S501, Deploy a front-end monitoring probe on the target page, and use the front-end monitoring probe to capture the target user's click events, scroll depth and dwell time in real time, and generate interactive behavior data; S502, determine the conversion contribution of the target user based on the interaction behavior data, construct a model reward signal based on the conversion contribution, and adjust the parameters of the personalized recommendation model using the model reward signal; S503, Analyze the conversion rate performance of the interaction behavior data, and adjust the mapping relationship between the user profile data and the page layout structure defined in the page display strategy according to the conversion rate performance, so as to optimize the page display strategy.

[0069] In this embodiment, the target user refers to the interactive object that initiated the access request and is identified as the same entity. The identification source can be a login state identifier, device identifier, session identifier, or internal user key value mapped from the account system. Determining the target user is used to aggregate scattered page operation records to the same user dimension. The target page refers to a page view instance that has been presented by the terminal and allowed user operation. The source of the target page can be a page route identifier, page instance identifier, or rendering container identifier. Determining the target page is used to limit the interactive behavior data to a traceable page range, avoiding the mixing of events from different pages into the same statistical scope. Interactive behavior data is used to express the set of observable operations and measurable states generated by the target user on the target page. Interactive behavior data can take the form of a structured event sequence, a time window statistical table, or a combination of both. The structured event sequence includes at least one of the following fields: event type, event time, event location, and event payload. The time window statistical table includes at least one of the following fields: window start and end time, cumulative number of times, cumulative duration, and maximum scroll depth.

[0070] Front-end monitoring probes are monitoring components deployed in the target page's runtime environment. Deployment includes loading probe code, completing initialization configuration, and establishing event subscription and reporting channels. The implementation source of front-end monitoring probes can be browser script injection, terminal SDK integration, or rendering container plugin extension. Initialization configuration includes at least one of the following: sampling ratio, event filtering rules, reporting batch size, and reporting trigger conditions. Click events express the target user's trigger action on interactive controls. Click event collection sources can be control listener callbacks or event bubbling captures. The payload field of a click event can include at least one of the following: control identifier, control position, page area identifier, and trigger method field. Scroll depth measures the target user's browsing progress on the target page. Scroll depth can be derived from the ratio calculated from the scroll container's displacement and content height, or from the maximum reachable segment obtained using segmented anchor points. Scroll depth distinguishes between browsing the first screen and browsing deeper content areas. Dwell time measures the effective time a target user spends on a target page. Dwell time can be derived from a time period constrained by page visibility status, foreground / background switching events, and interaction event timestamps. Dwell time calculations exclude background suspension time and focus-free time to reduce noise. Interaction behavior data is generated by a front-end monitoring probe that uniformly encapsulates click events, scroll depth, and dwell time. Encapsulation includes event alignment, timestamp standardization, field normalization, and missing field filling. Timestamp standardization can use server time calibration or a relative time representation of the session start point. Field normalization maps control identifiers reported by different terminals to a unified set of control identifiers.

[0071] Conversion contribution is used to map interactive behavior data into a measure of contribution to the target business conversion. The source of conversion contribution is the event sequence and statistics in the interactive behavior data. The form of conversion contribution can be a score, level, or probability value. Conversion contribution construction includes defining conversion goals and contribution calculation rules. Conversion goals can consist of page submission actions, key control trigger actions, or confirmation actions. Contribution calculation rules can combine the number of key control hits in click events, the state of scroll depth reaching a threshold, and the state of dwell time exceeding a threshold for weighted summation. Alternatively, path-based attribution rules can be used to map click events of different controls to different contribution weights. Model reward signals are used to convert conversion contribution into feedback input that the personalized recommendation model can accept. The source of model reward signals is the numerical result of conversion contribution. Model reward signals can be scalar rewards, segmented rewards, or vector rewards. Scalar rewards are used to directly drive the direction of parameter updates, segmented rewards are used to reduce the impact of extreme values, and vector rewards are used to simultaneously express feedback from multiple target dimensions. The construction of the model reward signal includes scale normalization and noise suppression. Scale normalization can map the conversion contribution of different pages and different session lengths to a unified range. Noise suppression can eliminate abnormal click events caused by automatic refresh, accidental clicks, or script triggers based on anomaly detection rules.

[0072] The parameters of a personalized recommendation model represent the set of adjustable variables used to generate recommendation results within the model. These parameters can originate from initial parameters obtained through offline training or from continuously updated online parameters. The model reward signal adjusts the parameters of the personalized recommendation model, including parameter update triggering, update sample organization, and update write control. Parameter update triggering can be achieved when the model reward signal reaches a threshold or when a batch condition is met. Update sample organization can combine interaction behavior data, user profile data, and conversion contribution into training sample records. Update write control can use version number incrementing and rollback point recording to ensure parameter traceability. The expression of parameter adjustment is not limited to specific learning rules; gradient updates, weighted estimation, incremental fitting, or rule-based parameter correction are allowed, as long as the model reward signal constrains the direction or magnitude of parameter updates.

[0073] Conversion rate performance measures the conversion effectiveness of interactive behavior data within a statistically valid framework. Sources of conversion rate performance can include the ratio of sessions that completed conversions to the total number of sessions, the ratio of users achieving conversion goals to the total number of users accessing the platform, or segmented conversion rates based on user profile data. Conversion rate performance analysis includes maintaining consistency in metrics and grouped statistics. Maintaining consistency includes unifying conversion goals, time windows, and deduplication rules. Grouped statistics can be performed based on the attribute dimensions of user profile data. The mapping relationship between user profile data and page layout structure defined in the page display strategy describes the binding rules that trigger different page layout structure configurations under different user profile data conditions. The mapping relationship originates from the strategy configuration file or rule set generated by the visual configuration interface. Adjusting the mapping relationship based on conversion rate performance involves locating the rule items to be adjusted and generating replacement rule items. Locating the rule items to be adjusted can be obtained by comparing the conversion rate performance of different rule items across different target audiences. Generating replacement rule items can be done by adjusting display threshold conditions, adjusting attribute dimension selection, adjusting page layout structure references, or adjusting rule priorities. The adjustment results are written into the page display strategy so that subsequent page generation uses the updated mapping relationship.

[0074] For example, taking online health insurance application as an example, operations staff prepare multiple application page template files. These template files cover different page layouts, color themes, font styles, and button styles, and include page elements such as health disclosure prompts, coverage explanations, premium calculation entry areas, and application information filling areas. The template management module receives these page template files, performs format validity checks, and after successful validation, the parser identifies and extracts page element information. Based on this information, it determines business classification metadata, such as categorizing templates into different archive directories according to health insurance product lines, application process stages, and user type preferences. After the page template files are written to the archive directory, their storage path is obtained. The content verification code and file upload timestamp are used to generate a version identifier, and corresponding change logs are recorded. The index database uniformly registers the storage path, version identifier, page element information, and change logs, forming a page template resource library. In actual operation, the page template resource library is used to centrally manage template versions. When operators adjust button styles or text layout during an event, they can locate historical versions by version identifiers, track the scope of changes by change logs, and quickly retrieve the target template set for the health insurance application page by business category metadata.

[0075] For flexible configuration of health insurance application pages, the template configuration module provides a visual configuration interface. The component library area loads and displays page element information, while the canvas area is used for drag-and-drop layout actions and attribute adjustments. Operators drag the health declaration component into the first screen area, place the policyholder information entry component in the next screen, present the coverage description component in a collapsed area, and adjust attributes such as color theme, font style, and button style. Drag-and-drop layout actions and attribute adjustments are recognized as configuration operations, used to determine the coordinate position parameters and stacking order of page elements, generating the page layout structure. To display different page styles for different customer groups, the interface provides logical binding instructions. Operators select attribute dimensions from user profile data and set display threshold conditions. For example, binding the regional dimension to the intensity of health declaration prompts, the age range dimension to the expansion method of the coverage description, and the occupation dimension to the display position of the risk warning component, thereby establishing a mapping relationship between user profile data and the page layout structure. The page layout structure and mapping relationship are encapsulated into a structured configuration file and a page display strategy is generated. The page display strategy is used to constrain the organization order, area placement and display conditions of page elements when the page is generated in subsequent pages.

[0076] When a user enters the health insurance application page, the personalized recommendation module identifies the target user who initiated the access request, extracts user profile data associated with the target user from the user database, and reads the target user's user behavior data from the log system. User profile data may include age range, geographic information, occupation category, historical insurance preference tags, etc., while user behavior data may include the frequency of browsing health insurance product pages, preference for clicking on the coverage description, dwell time on the health declaration page, and interruption points on the application information filling page. The user profile data and user behavior data are processed into feature vectors to form input feature vectors. These feature vectors are then input into a preset personalized recommendation model to obtain matching probability values. These matching probability values ​​correspond to various templates in the page template resource library, and after sorting and filtering, page template combinations are formed. Significant differences can be observed in the health insurance scenario: first-time visitors whose browsing preferences focus on the coverage description are more likely to be matched with template combinations that provide more comprehensive information; repeat visitors who are sensitive to process efficiency are more likely to be matched with template combinations that offer a simpler initial screen and shorter filling paths; younger users are more likely to be matched with button styles and color themes that offer stronger interaction; and older users are more likely to be matched with template combinations that provide clearer fonts and more focused prompts.

[0077] The dynamic rendering module retrieves and calls corresponding page element information from the page template resource library based on the page display strategy and page template combination. It embeds this information into the corresponding position in the page layout structure, dynamically generates page code, and executes rendering operations to display the target page. During rendering, the rendering engine parses the page code and identifies associated resources. For static resources, it prioritizes searching the local cache system. For non-first-screen resources, it registers visible area listener events and performs asynchronous loading when they enter the visible area of ​​the screen. Subsequently, it combines the document object model tree with the loaded associated resources to construct a rendering tree, determines the geometric layout coordinates of page elements, and performs drawing. After layer compositing, it generates the rendered page and displays it on the user's terminal. Health insurance application pages often include resources such as terms and conditions icons, risk warning pop-up styles, and health declaration questionnaire controls. Caching reduces the overhead of repeated loading, and asynchronous loading of non-first-screen resources reduces first-screen waiting time. This ensures stable page rendering performance and interactive response, reducing lag and crashes for users when viewing coverage and filling out disclosure items.

[0078] User interaction data on the target page is captured in real time by a front-end monitoring probe. Click events, scroll depth, and dwell time are encapsulated as interaction data and processed in the feedback mechanism. This data is used to determine the conversion contribution of target users. For example, actions such as completing a health declaration questionnaire, entering a premium calculation, completing insurance information, and submitting confirmation are mapped to different contribution weights. This conversion contribution is further used to construct a model reward signal, which is used to adjust the parameters of the personalized recommendation model, making subsequent matching probabilities more closely match actual conversion behavior. The conversion rate of the interaction data is used to evaluate the effectiveness of the page display strategy. For example, if a template combination under a certain user profile dimension increases the completion rate of the health declaration questionnaire but decreases the submission rate, the mapping relationship can be adjusted by moving the risk warning component to the back or changing the default expansion method of the coverage description, thereby optimizing the page display strategy. In health insurance operations, there are often differences between promotional periods and off-peak periods. During promotional periods, the efficiency of processes is more important, while during off-peak periods, the understanding of information is more important. Continuous feedback from interactive behavior data enables personalized recommendation models and page display strategies to absorb these changes in sync, making it easier for users who subsequently enter the insurance application page to see a page style and content organization that matches their own profile and behavior.

[0079] This embodiment uses a front-end monitoring probe to uniformly collect and encapsulate click events, scroll depth, and dwell time. The interaction behavior data forms a structured expression with alignable time and field definitions. The conversion contribution can be stably constructed based on the interaction behavior data and further converted into model reward signals, providing constrained update inputs for the parameters of the personalized recommendation model. By statistically analyzing the conversion rate performance formed by the interaction behavior data and adjusting the mapping relationship in the page display strategy accordingly, the binding rules between user profile data and page layout structure can be updated with actual performance, making subsequent page displays more aligned with the interactive feedback of the target user.

[0080] In one embodiment, a page configuration and dynamic display device is provided, which corresponds one-to-one with the page configuration and dynamic display method in the above embodiments. (Refer to...) Figure 3 , Figure 3 This is a schematic diagram of the functional modules of a preferred embodiment of the page configuration and dynamic display device of the present invention. The modules include a page template management module 10, a strategy generation module 20, a recommendation matching module 30, a page generation module 40, and a feedback optimization module 50. Detailed descriptions of each functional module are as follows: The page template management module 10 is used to receive page template files in various formats, the page template files containing page element information, perform classification storage and version control operations on the page template files, and generate a page template resource library; Strategy generation module 20 is used to generate page display strategies by responding to configuration operations on the page element information through a visual configuration interaction interface. Recommendation matching module 30 is used to collect user profile data and user behavior data of target users, input the user profile data and user behavior data into a preset personalized recommendation model to perform feature matching, and obtain a combination of page templates corresponding to the target user; The page generation module 40 is used to dynamically generate page code by calling page element information from the page template resource library according to the page display strategy and the page template combination, and to perform rendering operations to display the target page. The feedback optimization module 50 is used to record the interaction behavior data generated by the target user on the target page, and feed the interaction behavior data back to the personalized recommendation model and the page display strategy to update the parameters of the personalized recommendation model and optimize the page display strategy.

[0081] For specific limitations regarding page configuration and dynamic display devices, please refer to the foregoing limitations on page configuration and dynamic display methods, which will not be repeated here. Each module in the aforementioned page configuration and dynamic display device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in the computer device in hardware form, or stored in the memory of the computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0082] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 4 As shown. The computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides determination and control capabilities. The memory includes non-volatile and / or volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The network interface is used to communicate with external clients via a network connection. When the computer program is executed by the processor, it implements the functions or steps of a page configuration and dynamic display method on the server side.

[0083] In one embodiment, a computer device is provided, which may be a client, and its internal structure diagram may be as follows: Figure 5 As shown, the computer device includes a processor, memory, network interface, display screen, and input devices connected via a system bus. The processor provides determination and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The network interface is used to communicate with an external server via a network connection. When executed by the processor, the computer program implements client-side functions or steps of a page configuration and dynamic display method.

[0084] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to perform the following steps: The system receives page template files in various formats, each containing page element information, and performs categorized storage and version control operations on the page template files to generate a page template resource library. The page display strategy is generated by responding to configuration operations on the page element information through a visual configuration interface. Collect user profile data and user behavior data of the target user, input the user profile data and user behavior data into a preset personalized recommendation model to perform feature matching, and obtain a combination of page templates corresponding to the target user; Based on the page display strategy and the page template combination, page code is dynamically generated by calling page element information from the page template resource library, and rendering operations are performed to display the target page. Record the interactive behavior data generated by the target user on the target page, and feed the interactive behavior data back to the personalized recommendation model and the page display strategy to update the parameters of the personalized recommendation model and optimize the page display strategy.

[0085] In one embodiment, a computer-readable storage medium is provided, which may be non-volatile or volatile, and a computer program is stored thereon, which, when executed by a processor, performs the following steps: The system receives page template files in various formats, each containing page element information, and performs categorized storage and version control operations on the page template files to generate a page template resource library. The page display strategy is generated by responding to configuration operations on the page element information through a visual configuration interface. Collect user profile data and user behavior data of the target user, input the user profile data and user behavior data into a preset personalized recommendation model to perform feature matching, and obtain a combination of page templates corresponding to the target user; Based on the page display strategy and the page template combination, page code is dynamically generated by calling page element information from the page template resource library, and rendering operations are performed to display the target page. Record the interactive behavior data generated by the target user on the target page, and feed the interactive behavior data back to the personalized recommendation model and the page display strategy to update the parameters of the personalized recommendation model and optimize the page display strategy.

[0086] It should be noted that the functions or steps that can be implemented by the computer-readable storage medium or computer device described above can be referred to the relevant descriptions on the server side and client side in the foregoing method embodiments. To avoid repetition, they will not be described one by one here.

[0087] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0088] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0089] It should be noted that any AI models, software tools, or components not belonging to this company appearing in the embodiments of this application are merely illustrative examples and do not represent actual use. All user personal information involved in the embodiments of this application has been authorized (with the knowledge and consent) by the relevant parties or has been fully authorized by all parties, and the executing entity may obtain it through various public, legal, and compliant means. The collection, storage, use, processing, transmission, provision, and disclosure of the information, data, and signals involved all comply with the relevant laws and regulations of the relevant countries and regions, and do not violate public order and good morals.

[0090] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A method for page configuration and dynamic display, characterized in that, Includes the following steps: The system receives page template files in various formats, each containing page element information, and performs categorized storage and version control operations on the page template files to generate a page template resource library. The page display strategy is generated by responding to configuration operations on the page element information through a visual configuration interface. Collect user profile data and user behavior data of the target user, input the user profile data and user behavior data into a preset personalized recommendation model to perform feature matching, and obtain a combination of page templates corresponding to the target user; Based on the page display strategy and the page template combination, page code is dynamically generated by calling page element information from the page template resource library, and rendering operations are performed to display the target page. Record the interactive behavior data generated by the target user on the target page, and feed the interactive behavior data back to the personalized recommendation model and the page display strategy to update the parameters of the personalized recommendation model and optimize the page display strategy.

2. The page configuration and dynamic display method as described in claim 1, characterized in that, The system receives page template files in various formats, each containing page element information. It then performs categorized storage and version control operations on these page template files to generate a page template resource library, including: The system receives page template files in various formats uploaded via a file transfer interface, performs file format validity checks on the page template files, and obtains files that pass the checks. The parser is used to identify and extract page element information from the verified files, and business category metadata is determined based on the page element information. Match the corresponding archive directory according to the business category metadata, write the verified file into the archive directory, and obtain the storage path of the verified file in the archive directory; Determine the content verification code of the verified file, generate a unique version identifier based on the content verification code and the file upload timestamp, and record the change log corresponding to the verified file; Construct an index database containing the storage path, the unique version identifier, the page element information, and the change log, and generate a page template resource library.

3. The page configuration and dynamic display method as described in claim 1, characterized in that, The page display strategy is generated by responding to configuration operations on the page element information through a visual configuration interface, including: Render the visual configuration interaction interface, and load and display the page element information in the component library area of ​​the visual configuration interaction interface; Detect drag-and-drop layout actions and attribute adjustment actions in the canvas area of ​​the visual configuration interaction interface for the page element information, and identify the drag-and-drop layout actions and attribute adjustment actions as configuration operations; Based on the configuration operation, the coordinate position parameters and stacking order of page elements are determined, and the page layout structure is generated; In response to the logical binding instruction for the page layout structure, select the attribute dimensions of the user profile data and set the display threshold conditions to establish a mapping relationship between the user profile data and the page layout structure; The page layout structure and the mapping relationship are encapsulated into a structured configuration file to generate a page display strategy.

4. The page configuration and dynamic display method as described in claim 1, characterized in that, Collect user profile data and user behavior data of the target user, input the user profile data and user behavior data into a preset personalized recommendation model to perform feature matching, and obtain a combination of page templates corresponding to the target user, including: Identify the target user who initiated the access request, extract user profile data associated with the target user from the user database, and read the user behavior data of the target user from the log system; The user profile data and the user behavior data are subjected to feature vectorization processing to construct an input feature vector representing the characteristics of the target user; The input feature vector is input into a preset personalized recommendation model, and the personalized recommendation model is used to determine the matching probability value between the input feature vector and each template in the page template resource library; The templates in the page template resource library are sorted and filtered according to the matching probability value, and the filtered templates are used as the page template combination corresponding to the target user.

5. The page configuration and dynamic display method as described in claim 1, characterized in that, Based on the page display strategy and the page template combination, page code is dynamically generated by calling page element information from the page template resource library, and rendering operations are performed to display the target page, including: The page display strategy is parsed to obtain the page layout structure, and the page template combination is parsed to obtain the template identifier; Based on the template identifier, retrieve and call the corresponding page element information from the page template resource library; The page element information is embedded into the corresponding position in the page layout structure to generate page code; The rendering engine is used to parse the page code and load associated resources to perform rendering operations, generating the rendered page. The rendered page is then displayed on the user's terminal as the target page.

6. The page configuration and dynamic display method as described in claim 5, characterized in that, The rendering engine parses the page code and loads associated resources to perform rendering operations, generating the rendered page, including: The rendering engine is used to perform lexical and syntactic analysis on the page code, construct a document object model tree, and identify the associated resources referenced in the page code; Initiate a resource acquisition request based on the resource attribute type of the associated resource to load the associated resource; During the loading of associated resources, for associated resources that are static resources, the local cache system is searched, and if the search is successful, the associated resources are read. During the loading of associated resources, for associated resources that are not on the first screen, a visible area listener event is registered, and asynchronous loading is performed when the associated resources enter the visible area of ​​the screen. A rendering tree is constructed based on the document object model tree and the associated resources that have been loaded. The geometric layout coordinates of the page elements are determined based on the rendering tree and the drawing is performed. The drawn layers are composited to generate the rendered page.

7. The page configuration and dynamic display method as described in claim 1, characterized in that, Recording the interactive behavior data generated by the target user on the target page, and feeding the interactive behavior data back to the personalized recommendation model and the page display strategy to update the parameters of the personalized recommendation model and optimize the page display strategy, including: A front-end monitoring probe is deployed on the target page to capture the target user's click events, scroll depth, and dwell time in real time, generating interactive behavior data. The conversion contribution of the target user is determined based on the interaction behavior data, and a model reward signal is constructed based on the conversion contribution. The parameters of the personalized recommendation model are then adjusted using the model reward signal. Analyze the conversion rate performance of the interaction behavior data, and adjust the mapping relationship between user profile data and page layout structure defined in the page display strategy based on the conversion rate performance, so as to optimize the page display strategy.

8. A page configuration and dynamic display device, characterized in that, The page configuration and dynamic display device includes: The page template management module is used to receive page template files in various formats, which contain page element information, and to perform classification, storage and version control operations on the page template files to generate a page template resource library. The strategy generation module is used to respond to configuration operations on the page element information through a visual configuration interaction interface and generate a page display strategy. The recommendation matching module is used to collect user profile data and user behavior data of target users, input the user profile data and user behavior data into a preset personalized recommendation model to perform feature matching, and obtain a combination of page templates corresponding to the target user. The page generation module is used to dynamically generate page code by calling page element information from the page template resource library according to the page display strategy and the page template combination, and to perform rendering operations to display the target page. The feedback optimization module is used to record the interaction behavior data generated by the target user on the target page, and feed the interaction behavior data back to the personalized recommendation model and the page display strategy to update the parameters of the personalized recommendation model and optimize the page display strategy.

9. A computer device, characterized in that, The computer device includes a memory, a processor, and a page configuration and dynamic display program stored in the memory and executable on the processor. When the page configuration and dynamic display program is executed by the processor, it implements the steps of the page configuration and dynamic display method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The storage medium stores a page configuration and dynamic display program, which, when executed by a processor, implements the steps of the page configuration and dynamic display method as described in any one of claims 1-7.

Images