A link graph processing method and device, a storage medium and an electronic device

By transforming inter-page and cross-level jump sequences into graph edges and anchors, and combining this with an incremental update mechanism, the problems of data discontinuity and resource waste in the user interaction link graph are solved, achieving efficient link graph processing.

CN123115473APending Publication Date: 2026-11-17ZHEJIANG E COMMERCE BANK CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611299589.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-25
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing technologies for generating and maintaining user interaction link graphs suffer from data gaps and high maintenance costs. Furthermore, the full batch processing mode of traditional graph topology structures leads to resource waste and low timeliness.

Method used

By transforming inter-page navigation paths and cross-level navigation bridging sequences into directed edges and interface anchors in graph theory, the physical topology of the macro-routing flow skeleton and the micro-operation subtree within the page is cascaded and stitched together, and an incremental update mechanism is used for local topology maintenance.

Benefits of technology

It eliminates data gaps in macro and micro behavioral flow, enables automated reconstruction of link graphs, reduces resource waste, and improves timeliness and resource utilization efficiency.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123115473A_ABST
    Figure CN123115473A_ABST
Patent Text Reader

Abstract

This specification discloses a link graph processing method, apparatus, storage medium, and electronic device. The method includes: acquiring full SPM data of an application page and page user operation sequence data; parsing the full SPM data to determine inter-page jump path sequences and cross-level jump bridging sequences from click points to pages; determining an intra-page operation topology sequence based on the page user operation sequence data; using the inter-page jump path sequences as directed edges for topology bridging, and attaching the intra-page operation topology sequence as a local substructure to the interface nodes corresponding to the cross-level jump bridging sequences to generate an initial user interaction link graph; monitoring page operation behavior data of the application page, and performing incremental update processing of the topology behavior branches of the user interaction link graph based on the page operation behavior data. The embodiments of this specification eliminate behavioral link gaps without manual configuration, achieving low-computing-power real-time freshness updates of the user interaction link graph.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This specification relates to the field of computer technology, and in particular to a link graph processing method, apparatus, storage medium, and electronic device. Background Technology

[0002] With the increasing complexity of mobile internet and various application software, analyzing user interaction behavior on application pages has become a core technical means to drive product performance optimization and human-computer interaction upgrades. In this process, the user interaction link graph, as a high-dimensional topological data carrier, plays a crucial role. By abstracting user interaction touchpoints, page transitions, and micro-operations within a page into a graph structure composed of topological nodes and directed edges, the link graph can intuitively and comprehensively reconstruct the entire online user behavior flow. This allows for the subsequent accurate location of system interaction bottlenecks, capture of abnormal churn points, and measurement of the conversion efficiency of key behaviors by leveraging the connectivity and weight distribution of the user interaction link graph's graph structure. This provides indispensable data flow support for robust tuning of complex distributed systems and reconstruction of the front-end visual control matrix. Summary of the Invention

[0003] This specification provides a link graph processing method, apparatus, storage medium, and electronic device, the technical solution of which is as follows: Firstly, this specification provides a link graph processing method, the method comprising: Obtain full SPM data and user action sequence data for the application page; The full SPM data is analyzed to determine the page jump path sequence and the cross-level jump bridging sequence from the click point to the page, and the page operation topology sequence is determined based on the page user operation sequence data; Using the page jump transition sequence as the topology bridge directed edge, the page operation topology sequence as the local substructure is mounted to the interface node corresponding to the cross-level jump bridge sequence to generate an initial user interaction link graph. Monitor page operation behavior data of the application page, and perform incremental update processing of the topology behavior branches of the user interaction link graph based on the page operation behavior data.

[0004] Secondly, this specification provides a link graph processing apparatus, the apparatus comprising: The data acquisition module is used to acquire full SPM data for the application page and page user operation sequence data; The sequence parsing module is used to parse the full SPM data to determine the page jump path sequence and the cross-level jump bridge sequence from the click point to the page, and to determine the page operation topology sequence based on the page user operation sequence data; The graph generation module is used to use the page jump transition sequence as the topology bridge directed edge, and to attach the page operation topology sequence as a local substructure to the interface node corresponding to the cross-level jump bridge sequence to generate an initial user interaction link graph. The incremental update module is used to monitor page operation behavior data of the application page and perform incremental update processing of the topology behavior branches of the user interaction link graph based on the page operation behavior data.

[0005] Thirdly, this specification provides a computer storage medium storing at least one instruction adapted for loading by a processor and executing method steps of one or more embodiments of this specification.

[0006] Fourthly, this specification provides a computer program product storing at least one instruction adapted to be loaded by a processor and to execute the method steps of one or more embodiments of this specification.

[0007] Fifthly, this specification provides an electronic device that may include: a processor and a memory; wherein the memory stores a computer program adapted to be loaded by the processor and to execute the method steps of one or more embodiments of this specification.

[0008] The beneficial effects of the technical solutions provided in some embodiments of this specification include at least the following: In one or more embodiments of this specification, the above-mentioned technical means solve the problems of data fragmentation and the inability to keep graph topology fresh in real time in user interaction link analysis. By transforming the jump paths between pages and the cross-level jump bridging sequences into directed edges and interface anchors in graph theory, the physical topology cascading and stitching of the macro-routing flow skeleton and the micro-operation subtree inside the page is realized. This eliminates the data fragmentation caused by the separation between macro and micro behavioral flow lines in traditional log analysis, and achieves automated reconstruction of the interaction link graph with zero manual rule configuration. At the same time, relying on the streaming dynamic incremental update mechanism for local topology behavioral branches, the system can perform local low-overhead topology maintenance only on the behavioral branches that have mutated. This completely avoids the high computational resource waste and performance load caused by the traditional graph reconstruction mechanism performing a global full recalculation of massive historical logs, and improves the timeliness, freshness and resource utilization efficiency of the link graph processing and the big data engine. Attached Figure Description

[0009] To more clearly illustrate the technical solutions in this specification or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this specification. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0010] Figure 1 This is a schematic diagram of a link graph processing system provided in this specification. Figure 2 This is a flowchart illustrating a link diagram processing method provided in this manual; Figure 3 This is a schematic diagram of a data acquisition process provided in this manual; Figure 4 This is a schematic diagram of a link diagram generation process provided in this manual; Figure 5 This is a flowchart illustrating an incremental update process provided in this manual; Figure 6 This is a schematic diagram of a link graph processing device provided in this specification; Figure 7 This is a schematic diagram of the structure of an electronic device provided in this specification. Detailed Implementation

[0011] The technical solutions in this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this specification, and not all embodiments. Based on the embodiments in this specification, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this specification.

[0012] In the description of this specification, it should be understood that the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance. In the description of this specification, it should be noted that, unless otherwise expressly specified and limited, "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or devices. Those skilled in the art can understand the specific meaning of the above terms in this specification based on the specific circumstances. Furthermore, in the description of this specification, unless otherwise stated, "multiple" means two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, and B alone. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship.

[0013] In related technologies, the user interaction link graph generation and maintenance mechanism has significant shortcomings in engineering practice. On the one hand, existing log analysis technologies typically physically isolate the macroscopic page navigation flow from the microscopic in-page operation flow. When constructing a panoramic link graph, it either heavily relies on a large number of manual rules for node mapping configuration, resulting in high maintenance costs and a lack of flexible scalability, or it lacks effective cross-level stitching links, leading to a severe data gap between the macroscopic flow and the microscopic in-page operations. On the other hand, traditional graph topology structures often adopt an offline big data full batch processing computing mode. When online front-end UI elements change or real-time behavioral characteristics drift, the graph computing engine must perform a full recalculation of massive historical logs to re-render the global graph matrix. This recalculation paradigm not only causes a serious waste of server cluster computing resources but also has extremely low timeliness, failing to meet the timeliness requirements of dynamic, incremental, and fresh-keeping updates of complex behavioral link graphs with low computing power overhead.

[0014] The present specification will now be described in detail with reference to specific embodiments.

[0015] Please see Figure 1 This is a schematic diagram of a link graph processing system provided in this specification. Figure 1 As shown, the link graph processing system may include at least a client cluster and a service platform 100.

[0016] The client cluster may include at least one client, such as Figure 1 As shown, it specifically includes client 1 corresponding to user 1, client 2 corresponding to user 2, ..., client n corresponding to user n, where n is an integer greater than 0.

[0017] Each client in a client cluster can be an electronic device with communication capabilities, including but not limited to: wearable devices, handheld devices, personal computers, tablets, in-vehicle devices, smartphones, computing devices, or other processing devices connected to a wireless modem. Electronic devices may have different names in different networks, such as: user equipment, access terminal, user unit, user station, mobile station, mobile station, remote station, remote terminal, mobile device, user terminal, electronic device, wireless communication device, user agent or user device, cellular phone, cordless phone, personal digital assistant (PDA), and electronic devices in 5G networks or future evolved networks.

[0018] The service platform 100 can be a standalone server device, such as a rack-mount, blade, tower, or cabinet-type server device, or a workstation, mainframe, or other hardware device with strong computing power; or it can be a server cluster composed of multiple servers. The servers in the service cluster can be composed in a symmetrical manner, wherein each server is functionally and hierarchically equivalent in the transaction chain, and each server can provide services independently. The independent provision of services can be understood as not requiring the assistance of other servers.

[0019] It should be noted that the service platform 100 establishes a communication connection with at least one client in the client cluster via a network for interactive communication. This network can be a wireless network or a wired network. Wireless networks include, but are not limited to, cellular networks, wireless LANs, infrared networks, or Bluetooth networks. Wired networks include, but are not limited to, Ethernet, universal serial bus (USB), or controller area networks. In one or more embodiments of the specification, technologies and / or formats including Hyper Text Markup Language (HTML), Extensible Markup Language (XML), etc., are used to represent data exchanged over the network (such as target compressed packets). Furthermore, conventional encryption technologies such as Secure Socket Layer (SSL), Transport Layer Security (TLS), Virtual Private Network (VPN), and Internet Protocol Security (IPsec) can be used to encrypt all or some links. In other embodiments, customized and / or dedicated data communication technologies can be used to replace or supplement the aforementioned data communication technologies.

[0020] In one or more embodiments of this specification, the service platform 100 and at least one client in the client cluster can establish a communication connection through a data transmission network, and complete the data interaction during the link graph processing based on the communication connection.

[0021] The client can run a target application (such as a mobile app, H5 application, mini-program, or web page). When an end user performs continuous interactive operations such as clicking, swiping, route navigation, or form input on the application page presented by the client, the log collection suite (SDK) embedded in the client will capture and record the atomic-level event tracking data generated during the interaction in real time. Based on this, the client cluster continuously outputs full Super Location Model (SPM) data, page user operation sequence data, and real-time page operation behavior data generated in subsequent operation and maintenance phases to the outside world through a high-concurrency thread pool.

[0022] The data transmission network is used to receive massive amounts of unstructured interactive logs reported frequently by the client cluster. It uses mechanisms such as FIFO (First In First Out) pipelines to rectify, smooth out peaks, and buffer heterogeneous logs to ensure the stability of data output from downstream service platforms.

[0023] Service platform 100 is the core of calculation and control in this link graph processing method. Service platform 100 acquires the full SPM data for the application page and the page user operation sequence data; it parses the full SPM data to determine the inter-page jump path sequence and the cross-level jump bridging sequence from the click point to the page, and determines the operation topology sequence within the page based on the page user operation sequence data; it uses the inter-page jump path sequence as the directed edge of the topology bridging, and attaches the operation topology sequence within the page as a local substructure to the interface node corresponding to the cross-level jump bridging sequence to generate the initial user interaction link graph; it monitors the page operation behavior data of the application page, and performs incremental update processing of the topology behavior branches of the user interaction link graph based on the page operation behavior data.

[0024] The link graph processing system embodiments provided in this specification and the link graph processing methods described in one or more embodiments belong to the same concept. The execution entity corresponding to the link graph processing method in one or more embodiments of this specification can be the aforementioned service platform 100; the execution entity corresponding to the link graph processing method in one or more embodiments of this specification can also be the electronic device corresponding to the client, depending on the actual application environment. The implementation process of the link graph processing system embodiments can be detailed in the following method embodiments, and will not be repeated here.

[0025] based on Figure 1 The schematic diagram shown below illustrates the link graph processing method provided by one or more embodiments of this specification.

[0026] Please see Figure 2 This specification provides a flowchart illustrating a link graph processing method according to one or more embodiments. This method can be implemented using a computer program and can run on a link graph processing device based on the von Neumann architecture. The computer program can be integrated into an application or run as a standalone utility application. The link graph processing device can be a service platform.

[0027] Specifically, the link graph processing method includes: S102: Obtain the full SPM data and page user operation sequence data for the application page; An application page refers to an independent graphical interface entity that is presented in the target application, mini-program, webpage, or window view running on a terminal device and is available for human-computer interaction.

[0028] Full SPM data refers to a collection of logs captured based on the unified collection standard of the Super Position Model, which contains the origin codes of all interactions triggered by the user within the application interface. It is usually composed of multi-level concatenated strings representing transaction categories, page identifiers, control components, and click indexes, and has strong spatial positioning characteristics.

[0029] Page user operation sequence data refers to a set of continuous user interaction events arranged in chronological order according to timestamps within the boundaries of a single application page. It fully records the micro-level non-jump behaviors of users within a specific application page, such as swiping, hovering, long pressing, text focusing, and form filling.

[0030] Indicatively, the service platform sends a data retrieval request for the target application page to the log centralized storage server via the data bus by invoking a pre-set distributed data retrieval operator in the background. This request carries the target time window to be analyzed and the application's version serial number. In response to this request, the log centralized storage server packages the full SPM data block containing multi-level cascaded location codes, which was asynchronously reported and classified by the client cluster and stored on disk, and the page user operation sequence data file, which is grouped and packaged by session identifiers and records the micro-touchpoint dependencies of users within a single page, into data entities and sends them back to the service platform via the network interface in the form of a structured document or compressed binary stream. After obtaining the structured document or compressed binary stream, the data receiving unit of the service platform calls the memory parsing thread to perform multi-threaded deserialization processing, thereby restoring the atomic-level full SPM dataset and page user operation sequence dataset distributed by session for subsequent topology graph construction in memory.

[0031] S104: Analyze the full SPM data to determine the page jump path sequence and the cross-level jump bridging sequence from the click point to the page, and determine the page operation topology sequence based on the page user operation sequence data; The page navigation sequence refers to the set of directed linear paths formed when a user switches between different independent application pages in a macroscopic topology. This sequence reflects the coarse-grained flow state at the page level and constitutes the skeleton structure of the final link graph.

[0032] Cross-level jump bridging sequence refers to a composite directed sequence of behaviors used to establish a physical mapping relationship between macro-level page nodes and micro-level operation nodes within the page. It accurately captures the process of a user directly triggering a route jump from a specific micro-level click position on the current page to the next target page, and is the core topological bridging edge that eliminates the data gap between the macro-level traffic flow and the operation within the page.

[0033] An in-page operation topology sequence refers to a micro-directed graph topology sequence with directional, dependent, or closed-loop characteristics, formed by a series of non-cross-page micro-interactions (such as clicking buttons, swiping the screen, checking checkboxes, etc.) triggered by the user within the physical boundary or visual window of a single application page.

[0034] In a schematic manner, for the full SPM data, the cascading strings of the super position model are segmented and processed using a preset cascading delimiter to extract the first-level cascading field representing the unique identifier of the application page. Then, the unique page identifiers are linearly concatenated according to the timestamp order within the same session identifier, thereby determining the page jump sequence reflecting the macroscopic page view flow status. Simultaneously, the processor deeply extracts the lower-level cascading fields representing micro-control components and click index information from the full SPM data, as well as the first-level cascading field of the next page that is temporally adjacent. When the lower-level cascading field corresponding to a certain click position is identified, it directly leads to the page unique identifier... When switching between pages, the click identifier, the current source page identifier, and the next target page identifier are associated and mapped, thereby determining the cross-level jump bridging sequence between the click position and the page that crosses macro and micro boundaries. For the page user operation sequence data, the set of atomic operation events triggered by the same user in a single application page is grouped and isolated according to the session identifier. The interaction behavior attributes and occurrence timestamps of each atomic operation event are extracted. By calculating the directed dependency relationship between adjacent operation events, a local topology matrix is ​​constructed with specific operation behaviors as network element nodes and the order of behavior as directed edges, thereby determining the page operation topology sequence that reflects the micro behavior rules within the page.

[0035] For example, in a real-world application scenario, suppose we need to analyze the user's workflow within the application's transfer function. The service platform's parsing engine begins to deconstruct the data read into the buffer in step S102. For the full SPM data, the parsing engine first identifies the user's route flow from the "Transfer Homepage" to the "Recipient Information Input Page" and then to the "Transfer Result Page" through multi-level concatenated string segmentation, thus determining the macroscopic page jump sequence of "Transfer Homepage - Recipient Information Input Page - Transfer Result Page." Simultaneously, the parsing engine discovers that after the user clicks the "Quick Transfer Button" micro-component on the "Transfer Homepage," the system directly jumps to the "Transfer Result Page (One-click transfer to frequently used contacts)." At this point, the system binds the "Transfer Homepage," the "Quick Transfer Button Component Identifier," and the "Transfer Result Page" with directed edges, thereby determining the cross-level jump bridging sequence between the click point and the page.

[0036] On the other hand, regarding the user operation sequence data on the page, the parsing engine focuses on the intra-page behavior of the "payee information input page." It discovers that the user performed a continuous, non-cross-page interaction within the input page: "enter card number - select bank - enter amount - click confirm." The engine abstracts these micro-actions into graph nodes and connects them into local subtrees based on the chronological order of the operations, thereby determining the intra-page operation topology sequence within that single page. Through the refined parsing in step S104, the massive atomic behavior logs are successfully transformed into directed sequence entities with rigorous topological logic, providing a precise data skeleton and data subtrees for the subsequent physical suspension stitching of the integrated functional link graph.

[0037] S106: Using the page jump transition sequence as a topology bridge directed edge, the page operation topology sequence as a local substructure is mounted to the interface node corresponding to the cross-level jump bridge sequence to generate an initial user interaction link graph. A topological bridging directed edge refers to a data edge in a graph data structure that connects different topological levels and has a clearly defined flow direction attribute. In this embodiment, it refers to a directed topological edge that is transformed from a coarse-grained inter-page jump trajectory sequence mapping and is used to connect macroscopic page vertices.

[0038] Local substructures refer to local graph structures or subtree data entities generated by graph theory convergence and multidimensional aggregation of the topological sequence of operations within a page. They fully represent the micro-interaction network element relationships triggered by the user within the physical boundary of a single application page.

[0039] A UI node refers to a graph vertex entity contained in a cross-level jump bridge sequence, corresponding to a specific human-computer interaction component or physical trigger point. This node serves as a physical insertion anchor point in the macroscopic skeleton graph, embedded as a microscopic local substructure within the topology network.

[0040] The initial user interaction link graph refers to a panoramic behavioral topology directed graph data entity generated in system memory, which combines macro view jumps and micro click behaviors within the page, by physically attaching the local substructures of micro-page operations to the interface nodes corresponding to the cross-level jump bridging sequence, and then performing full topology stitching with the jump line sequence between pages as directed edges.

[0041] Indicatively, a globally unique vertex handle is first allocated to each application page in the page navigation sequence within the cache, and the routing relationship between adjacent pages is instantiated as directional edges with traffic weights and directions, thereby constructing a macroscopic skeleton directed graph network in memory composed of the topological bridging directed edges; the service platform's graph computing engine extracts the unique identifiers of the components contained in the cross-level navigation bridging sequence, performs hash addressing retrieval in the macroscopic skeleton directed graph network, accurately locates and activates the corresponding interface node as the target physical mounting anchor point; then, the page content determined in the previous step is read... The topological sequence is manipulated and defined as a local substructure with independent topological boundaries. By modifying the pointers in the adjacency list in physical memory, the root node pointer of the local substructure is directly bound to the child node list handle of the interface node, thereby completing the seamless dangling stitching of the micro-topological subtree to the macro-skeleton vertices at the physical memory stack level. Finally, the graph topology validity verification operator is called to perform a deep scan of connectivity and loop dependencies on the stitched global graph data structure. After confirming that there are no isolated nodes and the topological flow is completely self-consistent, the initial user interaction link graph with both macro- and micro-behavioral flow characteristics is formally solidified in the cache.

[0042] For example, in a real-world application scenario, assume that step S104 has successfully parsed and determined the interaction data of an e-commerce application. Specifically, the parsed page navigation sequence is "product display page - shopping cart page - order checkout page," and the parsed intra-page operation topology sequence is the directed flow of the micro-behavior performed by the user within the shopping cart page: "selecting product A - modifying quantity - triggering a coupon pop-up." Simultaneously, the determined cross-level navigation bridging sequence clarifies that the user navigated to the order checkout page by clicking the "Go to Checkout" button (the corresponding interface node) at the bottom of the shopping cart page.

[0043] During the specific execution of step S106, the service platform's graph computing engine begins to stitch and reconstruct the graph entities. First, the engine directly converts the macroscopic routing path of "product display page - shopping cart page - order checkout page" into the skeleton of the graph structure, with each page jump serving as a directed edge connecting topological bridges.

[0044] Subsequently, the engine treats the entire topological sequence of operations within the shopping cart page—"selecting product A - modifying quantity - triggering coupon pop-up"—as a local substructure. At this point, the graph computing engine locates the interface node corresponding to the "go to checkout" button within the shopping cart page, which is pointed to by this cross-level jump bridging sequence.

[0045] Next, a mounting operation is performed in memory. By establishing pointer associations, the aforementioned micro-level local substructures (i.e., single-page operation subtrees) are suspended as a whole below the "Go to Checkout" interface node. In this way, the previously fragmented macro-level page navigation flow and micro-level page operation flow are physically stitched together at the "Go to Checkout" button, thus automatically reconstructing a high-fidelity, panoramic initial user interaction flow diagram in the background computing engine. This diagram not only clearly shows the flow skeleton between pages but also deeply reveals which specific button the user triggered within the shopping cart page and what micro-level operation flow they experienced before jumping to the next page, eliminating the gap between macro and micro data.

[0046] S108: Monitor page operation behavior data of the application page, and perform incremental update processing of the topology behavior branches of the user interaction link diagram based on the page operation behavior data.

[0047] Page operation behavior data refers to the log stream of user behavior events captured in real time and reported asynchronously by online applications during operation. It fully records the latest interaction touchpoints, state transitions, asynchronous request calls or abnormal interruptions triggered by users within the physical boundaries of a specific application page, in order to characterize the current dynamic interaction characteristics of the system.

[0048] Topological behavior branches refer to local subgraph topological networks in the user interaction link graph, which consist of some specific operation vertices, page nodes, and directed edges directly connected to them. They represent the discrete behavior selection paths of users in a specific interface scenario.

[0049] Incremental update processing refers to a low-computational-cost mechanism for preserving the topology of a streaming graph. For example, it can be achieved by calculating the topological residual between the real-time behavioral flow and the historical static graph baseline, and performing dynamic node hot-plugging, directed edge insertion / deletion, or flow weight accumulation only on the local topological behavioral branches that have mutated, without destroying or recalculating the global existing graph structure.

[0050] In a schematic manner, the service platform monitors and acquires page operation behavior data triggered by user online operations, reflecting the latest changes in the front-end UI components. A single-page behavior state machine buffer is established in memory, mapping the incoming page operation behavior data into a dynamic state transition probability vector in real time. Through matrix subtraction and tensor product operations with the global basic adjacency matrix representing the user interaction link graph, a high-dimensional topological behavior change residual matrix is ​​generated online. The non-zero elements in this residual matrix directly represent the local variation state of the topological structure. The corresponding physical coordinate axis index and primitive attribute identifier are extracted from the residual matrix to directly address and lock the specific topological behavior branch within the user interaction link graph that currently requires preservation maintenance. Finally, using a local matrix fragment overlay update method, algebraic summation and directed edge pointer rewriting are performed only on the submatrix interval corresponding to the locked topological behavior branch. This achieves real-time erasure, dynamic insertion, and incremental accumulation of edge weights within the video memory canvas and topological entities, ensuring that the user interaction link graph completes the incremental update processing of the topological behavior branch while avoiding a full recalculation of the global graph matrix.

[0051] This specification addresses the issues of data fragmentation and the inability to maintain graph topology in real-time during user interaction link analysis using the aforementioned technical methods. By transforming inter-page jump paths and cross-level jump bridging sequences into directed edges and interface anchors in graph theory, it achieves physical topological cascading and stitching between the macroscopic routing flow skeleton and the microscopic operation subtrees within the page. This eliminates the data fragmentation caused by the disconnect between macroscopic and microscopic behavioral flows in traditional log analysis, achieving automated, zero-manual rule configuration reconstruction of the interaction link graph. Furthermore, relying on a streaming dynamic incremental update mechanism for local topological behavioral branches, the system can perform local, low-overhead topology maintenance only on mutated behavioral branches. This completely avoids the high computational resource waste and performance load caused by the traditional graph reconstruction mechanism's global full recalculation of massive historical logs, improving the timeliness, freshness, and resource utilization efficiency of the big data engine in link graph processing.

[0052] Please see Figure 3 , Figure 3 This is a schematic diagram of a data acquisition process proposed in one or more embodiments of this specification. Specifically, the acquisition of full SPM data and user operation sequence data for an application page can be performed as follows: S202: Collect raw application page stream logs reported by the client; The raw application page logs refer to the collection of unstructured or semi-structured atomic behavioral events captured in real time and continuously reported by the online client cluster while it is running, based on a preset log collection SDK. These logs completely record all raw triggering events of the front-end interface in the form of incremental data streams.

[0053] S204: Extract user access event features, element exposure event features, and element click event features from the original flow logs of the application page, and perform feature aggregation based on the user access event features, element exposure event features, and element click event features to obtain full SPM data; User access event features refer to the time-series data features extracted from the raw stream logs that characterize the user's actions of entering, exiting, navigating routes, and staying on various independent application pages.

[0054] Element exposure event characteristics refer to the visual presentation data characteristics triggered and recorded by the underlying rendering engine of the system when a specific interactive component, advertising window, button or image in the application page enters the physical screen visible area of ​​the terminal device and meets the preset dwell time and area ratio thresholds.

[0055] Element click event characteristics refer to the core interactive trigger data characteristics of users applying click actions to specific foreground control components within the application page through physical interaction methods such as finger touch, mouse triggering, or keyboard typing.

[0056] Feature aggregation can be a data fusion processing mechanism for multi-source heterogeneous features. It can be achieved by extracting shared association keys (such as user identifiers, session identifiers, and absolute timestamps) carried in different event features, and aligning, concatenating, and structurally encapsulating discrete features in spatial and temporal dimensions.

[0057] S206: Perform page interaction parsing on the user page behavior in the original flow log of the application page to obtain page user operation sequence data.

[0058] Page interaction parsing can be an algorithmic process that delves into the physical boundaries of a single page and performs temporal reorganization and semantic recognition of non-jump micro-operation behaviors triggered by the user, such as swiping span, component focusing, text filling, and pop-up wake-up.

[0059] Indicatively, S202-S206 are executed. The service platform collects raw application page stream logs asynchronously reported by the client cluster in batches and loads the collected stream log files into a memory buffer. Using a feature extraction engine and pre-configured event type parsing rules, the platform scans the raw application page stream logs and retrieves and extracts corresponding user access event features, element exposure event features, and element click event features from them using a pattern matching operator. Then, the aforementioned discrete feature entities are obtained. Using a globally unique session identifier and a high-precision timestamp as a combined primary key, the user access event features and element exposure event features belonging to the same session and the same time period are grouped together. The system performs horizontal connections and field completion on the features and element click event features. Based on this spatial and temporal multidimensional alignment, it performs feature aggregation processing, thereby aggregating and generating structured full SPM data in distributed storage nodes. At the same time, it reads the original flow logs of the application page in parallel, filters out cross-page jump actions through routing tags, focuses on the local behavior data of users triggered within a single page, such as window scrolling, form input, and long press response, and uses a temporal state machine to perform in-depth page interaction parsing on the user page behavior, extracts the sequential dependency order of each local behavior, and encapsulates and outputs page user operation sequence data that is isolated from each session and page.

[0060] In this specification, based on S202-S206, raw flow logs are collected in real time and aggregated with access, exposure, and click features to generate full SPM data. Simultaneously, interactive parsing of page behaviors is performed to obtain operation sequence data. This achieves standardized extraction of heterogeneous logs and precise decoupling of macro and micro dimensions, effectively avoiding data loss and temporal sequence disorder, and providing a high-fidelity, multi-dimensional aligned basic data source for subsequent integrated link graph construction.

[0061] Optionally, to specifically perform page interaction parsing on the user page behavior in the original flow log of the application page as described in S206 to obtain page user operation sequence data, the following methods can be used as a reference: Step C2: Identify page tracking events from the original stream logs of the application page, and perform spatial topological sorting on the page tracking events according to the DOM order in the current DOM topology tree structure of the application page to construct a first data sequence; Page event tracking refers to logging events that occur during application runtime and are bound to specific front-end page structures, component lifecycles, or specific view areas to mark when a user touches or activates a spatial structure boundary in the macro page layout.

[0062] The DOM (Document Object Model) topology tree structure refers to a directed tree graph data structure that organizes all visual control components, layout tags, and underlying interactive elements in the current rendering form of the target application page in a hierarchical manner according to their nesting relationships and spatial cascading order.

[0063] DOM order refers to the absolute logical order in which primitive nodes in the DOM topology tree structure are arranged according to depth-first traversal, breadth-first traversal, or preset front-end rendering pipeline layout rules.

[0064] Spatial topological sorting refers to a graph theory-based node-oriented sorting mechanism that is independent of timestamps and is entirely based on the relative geometric positions and hierarchical containment relationships of user behavior within the front-end UI or document structure tree.

[0065] The first data sequence refers to an ordered data chain with spatial topological completeness, which is output by aligning multiple embedded events that occur within a specific page in order according to their logical hierarchy or spatial arrangement in the DOM topology tree structure.

[0066] Step C4: Identify event tracking points from the original flow logs of the application page, obtain the timestamp difference between the current event tracking point and the adjacent event tracking points, and if the timestamp difference is within the preset time sequence verification window, then the current event tracking point is sequentially included in the second data sequence. Event tracking refers to atomic-level operation event log nodes that are captured and recorded by the underlying operating system or browser kernel when a user performs specific physical interaction actions (such as clicking, long pressing, swiping, double-tapping) on ​​the front-end graphical interface.

[0067] The timing verification time window refers to the upper limit of the critical continuous time difference threshold, which is pre-set inside the stream computing engine or batch processing state machine to filter out occasional noise, long-tail non-interactive states, or session interruption behaviors.

[0068] The second data sequence refers to a time-series data chain composed of a series of valid event tracking points that satisfy strong continuity dependencies in the time dimension and have passed the time sequence verification time window validity check, arranged in the order of their absolute occurrence timestamps.

[0069] Step C6: Perform sequence synthesis processing based on the first data sequence and the second data sequence to obtain page user operation sequence data.

[0070] Sequence synthesis processing refers to a graph theory-based multidimensional algorithmic process that combines a topological data stream with spatial physical arrangement characteristics with a temporal data stream with temporal continuity characteristics through cross-alignment, matrix transformation, mask reconstruction, or spatiotemporal normalization.

[0071] Indicatively, in steps C2-C6, the service platform first identifies and filters all page event tracking events directly associated with the current page container by reading event classification tags from the original flow logs of the application page that has been retrieved and loaded into the stack. At the same time, it loads the standard interface structure primitive configuration file corresponding to the current application page and dynamically restores and reconstructs the DOM topology tree structure of the current application page in memory. Furthermore, the unique handle of the component carried by each page event is extracted, and it is mapped and bound to the network element node in the DOM topology tree structure of the document object model. The topology sorting operator is called to perform spatial topology sorting on the page events in strict accordance with the absolute DOM order of each node in the tree structure, thereby transforming the discrete macroscopic position points into a linear directed linked list with spatial position dependency to construct the first data sequence. Synchronously, from the same application page's original flow logs, event points representing user atomic operations are filtered out through behavioral attribute matching. A time-series analysis operator is used to sequentially read the absolute physical occurrence time recorded by two adjacent event points. The timestamp difference between the current event point and the adjacent event point is obtained through subtraction. This difference is then compared with a pre-configured time-series verification window. If the timestamp difference is less than or equal to the preset time-series verification window, it is confirmed that these two micro-operations belong to the same continuous behavior flow without session interruption. The current event point is then appended to the cache queue following the time growth direction to construct a second data sequence. Finally, the first data sequence, serving as the spatial dimension baseline, and the second data sequence, serving as the time dimension baseline, are simultaneously input into the spatiotemporal fusion matrix model. A globally unique session identifier is used for cross-axis cross-indexing, accurately assigning micro-event points on the timeline to their corresponding spatial topology DOM slots. Based on this spatiotemporal normalized sequence comprehensive processing, a page user operation sequence data with both spatial hierarchy and temporal context is finally reconstructed and output.

[0072] In this specification, based on steps C2-C6, a spatiotemporal normalization mechanism is constructed through comprehensive processing of the execution sequence using DOM topology sorting and temporal window verification. This method effectively eliminates temporal noise caused by repeated scrolling or long pauses by users, avoids behavioral distortion in pure temporal analysis, and significantly improves the spatial completeness and causal certainty of page user operation sequence data.

[0073] Optional, such as Figure 4 As shown, Figure 4This is a flowchart illustrating the process of generating a link graph. Specifically, it involves using the cross-level jump bridging sequence as the topology bridging directed edges, and attaching the in-page operation topology sequence as a local substructure to the page nodes corresponding to the inter-page jump bridging sequence to generate an initial user interaction link graph. The following method can be used as a reference: S302: Encapsulate the full SPM data, the inter-page jump path sequence, the cross-level jump bridging sequence, and the intra-page operation topology sequence into a general graph data structure; The general graph data structure refers to a composite object container allocated in memory space for standardizing and accommodating multi-source heterogeneous graph theory causal data. Through a unified metadata specification, it integrates macroscopic page views, microscopic click operations, and cross-level jump logic into a graph data base that can be directly called by a unified topology operator.

[0074] The long behavioral path refers to the original single-chain interaction trajectory triggered by the user in the target application page, which is continuous and uninterrupted in time series but crosses the boundaries of multiple different local physical windows or components.

[0075] S304: In the general graph data structure, the operation topology sequence within the page is taken as a long behavioral path, and the path is cut according to the click position breakpoint in the long behavioral path to obtain multiple spatial local micro operation sub-sequences; A click breakpoint refers to a boundary marker vertex in the long behavioral flow where the user triggers a specific concrete cross-page flow interaction component (such as a submit button or external link element), causing the current single-page micro-interaction to terminate and triggering a hierarchical transition state.

[0076] The path cutting refers to the process by which the graph calculation engine calls the sequence segmentation algorithm, using the click position breakpoint as the physical segmentation scale, to cut and discretize a one-dimensional behavioral long line into multiple local sub-paths with spatial local convergence characteristics.

[0077] Micro-operation subsequences refer to atomic-level continuous operation behaviors that are spatially confined to a single specific page or local window component and obtained through path cutting, and are sometimes sequential chains.

[0078] S306: In the general graph data structure, the cross-level jump bridging sequence is used to perform cross-application page topology mounting concatenation processing on each of the micro-operation sub-sequences to generate an initial user interaction link graph.

[0079] Topology mounting and serial processing refers to using cross-level routing dependencies as topological links, and in reverse binding and suspending the scattered micro-operation sub-sequences as local subgraphs to the corresponding touch interface nodes at both ends of the macro-jump directed edge, thereby completing the computational logic of the graph structure.

[0080] Indicatively, in steps S302-S306, the service platform's graph topology construction node first assigns unified pointer handles to the full SPM data, inter-page jump path sequences, cross-level jump bridging sequences, and in-page operation topology sequences generated in the preceding steps within the physical memory stack. Through object-oriented attribute mapping and multi-dimensional field fusion and encapsulation, a high-dimensional general graph data structure is instantiated within the graph database. Within this general graph data structure, the input in-page operation topology sequence is defined as a global long-form path. A path scanning operator is invoked to search along this long-form path for all component identifiers marked with page switching trigger attributes or network request call attributes, establishing them as click breakpoints at spatial boundaries. Then, a path segmentation method is used to strictly follow these click breakpoints to define the long-form path. The line performs physical path cutting, thereby discretizing the long single-chain data into multiple spatially isolated micro-operation subsequences that each converge within their corresponding page. Subsequently, the control graph computing engine activates the adjacency list reconstruction operator within the general graph data structure, extracts the mapping pointers defined in the cross-level jump bridging sequence that connect macro-routes and micro-controls, and uses the cross-level jump bridging sequence as a graph theory stitching link. By modifying the out-degree and in-degree pointers of the adjacency list in physical memory, each of the aforementioned micro-operation subsequences is treated as a local subtree and precisely associated with the interface node below the corresponding application page in the page jump line sequence. Thus, the topology mounting and chaining process across application pages is completed within the general graph data structure, reconstructing and generating the initial user interaction link graph.

[0081] In this specification, multi-source heterogeneous link features are uniformly encapsulated into a general graph data structure. Click breakpoints are introduced within this structure to perform path cutting on long behavioral lines. Then, cross-level jump bridging sequences are used to perform topology mounting and concatenation. This achieves accurate decomposition and reverse stitching of long temporal behaviors into local page space graph elements. While reducing the computational overhead of panoramic topology map data assembly, it also achieves a high-fidelity reconstruction generation effect with seamless alignment of macro and micro behavioral links.

[0082] Optionally, the specific execution of the cross-application page topology mounting and concatenation processing of each of the micro-operation sub-sequences using the cross-level jump bridging sequence to generate an initial user interaction link graph can refer to the following methods: Step D2: Based on the cross-level jump bridging sequence, determine the target mounting parent page node corresponding to each of the micro-operation sub-sequences in the general graph data structure; The target mounting parent page node refers to the parent graph vertex entity within the general graph data structure, which is used to receive, anchor, and suspend a specific micro-operation sub-sequence. Physically, it corresponds to the specific trigger interface location or specific control component of the cross-level jump behavior within the source application page.

[0083] Indicatively, the object access interface within the general graph data structure is first invoked to read the cross-level jump bridging sequence obtained from the preceding steps into the cache. Then, the unique identifier of the source application page, the micro-touchpoint component encoding triggering the jump, and the corresponding session lifetime context features encapsulated in the cross-level jump bridging sequence are deeply extracted. The extracted component encoding and page unique identifier are used as a joint retrieval key to perform a precise addressing comparison in the pre-set global node index table or hash mapping matrix within the general graph data structure to find the micro-touchpoint. The component is physically bound to the graph vertex handle; the control graph calculation engine, based on the found graph vertex handle, combined with the source page attributes and event trigger timestamps of each micro-operation subsequence, performs topological causality verification, and in the general graph metadata structure, accurately determines the graph vertex that has a direct pre-trigger relationship with the micro-operation subsequence in business flow logic as the target mounting parent page node of each micro-operation subsequence, and allocates a corresponding mounting preparation handle in memory for the target mounting parent page node, so as to facilitate the physical association of the subsequent execution subtree pointer.

[0084] Step D4: Based on the target mounted parent page node, associate each of the micro-operation sub-sequences as a local operation subtree with the corresponding target mounted parent page node, and perform graph node deduplication and caching processing on the same interactive elements in the micro-operation sub-sequences within the general graph data structure to build the interactive association relationship between sibling operation nodes; Local operation subtrees refer to subgraphs or tree-like data structures with independent topological boundaries generated by topological reconstruction of micro-operation subsequences within a specific page. They fully represent the topological relationships of micro-interactive network elements triggered by the user within a single application page.

[0085] Graph node deduplication and caching refers to a graph element optimization mechanism that, within memory space or a graph database, merges and normalizes the unique identifiers of identical interactive element nodes generated due to repeated user triggering within the same application page or local operation subtree, and temporarily stores their topology handles in a high-speed cache to optimize storage overhead and computational performance.

[0086] The interactive relationship between peer operation nodes refers to the directed topological edge connection relationship established between various control component vertices that are at the same level or in parallel interaction within a single application page or local operation subtree, based on the absolute or relative sequence of the user's physical operations.

[0087] Schematic, the physical memory address handle of the target mounted parent page node, which has been locked in step D2, is extracted. By calling the underlying memory pointer assignment operator, the root node pointer of the local operation subtree corresponding to the micro-operation subsequence is directly written to the tail of the child node linked list of the target mounted parent page node, thereby completing the physical association mounting of the micro-operation subtree to the macro-skeleton node at the memory stack level. The control graph calculation engine activates the traversal scanning thread, scanning all newly introduced interactive element nodes layer by layer along the directed edges of the local operation subtree. Whenever an interactive element node is scanned, its globally unique component identifier is extracted and compared with the node identifier set stored in the cache using a fast hash function. If the same component identifier is matched, the element is determined to be... When a user repeatedly clicks on or triggers redundant elements, the graph primitive release operator is invoked to destroy the currently duplicated node, and the direction of the preceding directed edge is modified to converge and redirect to the unique node remaining in the cache. This completes the deduplication and caching of graph nodes within the general graph data structure. After the deduplication and merging action is completed, the temporal topology association operator is further invoked to extract the vertices of each independent interactive element at the same logical level in the same local operation subtree after deduplication. According to the absolute physical timestamp order in their original logs, a directed topology edge representing the sequential flow of behavior is dynamically instantiated between two adjacent interactive element vertices. By assigning the corresponding interaction frequency weight value to the directed edge, the interaction association relationship between sibling operation nodes is successfully constructed.

[0088] Step D6: Connect the end operation node of the application page to which the target parent page node belongs to the target through directed edge logic to the root node of the next application page within the general graph data structure to obtain the full topology network element structure. The end operation node refers to the critical graph vertex at the temporal or flow boundary in the local operation subtree corresponding to a specific application page. It marks the last interaction action performed by the user within the current page, and this action directly triggers the routing jump state to the next page.

[0089] The root node of a page refers to the starting source vertex in a graph theory network that represents the global virtual boundary of a specific application page. It serves as the top-level common ancestor of all micro-operation child nodes within that application page.

[0090] The full topology network structure refers to a composite topology network formed within a general graph data structure by globally and physically stitching together the local operation subtrees of each application page and the routing jump control logic across application pages through directed edges, thus containing complete graph theory network element connectivity relationships.

[0091] Indicatively, the system first uses algorithms such as depth-first or breadth-first search to perform connected path retrieval along the local operation subtrees associated with the target parent page node, extracting the out-degree and timestamp features of each vertex. This accurately identifies the end operation node triggered by the user in the current application page, which is in a state of edge truncation in terms of temporal sequence and graph theory flow. Simultaneously, the control graph computing engine retrieves the metadata configuration information of the next target application page from the general graph metadata structure and pulls up and activates the virtual top-level representation of the next application page in physical memory space. The parent node serves as the page root node. The directed edge instantiation operator is invoked to create a cross-page topology directed edge containing page jump status code, flow direction attribute, and initial flow weight in the cache stack space. By directly modifying the data structure of the graph adjacency list in memory, the out-degree pointer address of the end operation node is directly pointed to and written to the in-degree storage handle of the page root node, thereby eliminating the physical flow boundary between macro-routing flow and single-page micro-action. Finally, within the general graph data structure, a full topology network element structure containing cross-page full network element connectivity is obtained through concatenation.

[0092] Step D8: Perform tree structure transformation on the full topology network element structure to generate and output the initial user interaction link diagram.

[0093] Tree structure transformation refers to the process of calling a preset graph theory transformation algorithm to prune and hierarchically reorganize the global full topology network structure containing complex network dependencies, multiple loops or dense edge relationships according to the convergence rules of the main flow or flow direction, and reconstruct it into a high-fidelity data graph with clear tree branch topology characteristics.

[0094] Indicatively, the entire topology network structure is read, and a global connectivity graph scanning thread is activated to obtain the out-degree, in-degree, and flow weight distribution of all vertices and directed edges in the entire network topology. The control graph calculation engine calls the strongly connected component identification operator within the general graph data structure to accurately locate all closed-loop topology paths in the entire topology network structure. Based on the access frequency of the raw logs aggregated on each directed edge, the maximum flow pruning algorithm in related technologies is used to remove backtracking edges or non-backbone redundant edges below a preset flow threshold, ensuring physical connectivity between the global entry page and the terminal operation node. Under the premise of connectivity constraints, the graph topology structure is broken down and its dimensionality reduced. Then, the global session start page in the full topology network structure is used as the top-level root node. The directed paths after the breakdown and pruning are arranged in order of logical progression depth as a global trunk, single-page subtrees and component leaf nodes, thereby completing the tree structure transformation of the full topology network structure in memory. Finally, the tree topology entity mapping generated by the transformation, which has both macro and micro causal relationships, is encapsulated into a standard initial user interaction link graph object, and serialized and output to the trunk snapshot storage area of ​​the high-performance graph database for permanent storage.

[0095] In this specification, steps D2-D8 utilize a cross-level bridging mechanism to precisely anchor and suspend fine-grained single-page micro-operation subtrees to coarse-grained macro-routing vertices. By performing local node deduplication and caching optimization within the graph data structure, a fully automated, causal closed-loop panoramic interactive link network assembly with zero manual rule configuration is achieved. Simultaneously, a cross-page directed edge logical connection mechanism between the end operation node and the page root node, along with the final tree structure transformation operator, is introduced. This successfully converges the densely interwoven original mesh data stream into a hierarchical, tree-like link topology graph entity that conforms to human cognitive models. This reduces the storage and rendering computational load on the graphics engine while improving the accuracy and high-fidelity restoration capability of user panoramic behavior trajectory analysis.

[0096] Optionally, after generating the initial user interaction chain diagram in step D8, you can also refer to the following methods: 1) When a single-function full graph generation instruction is received, the target application category and the number of hotspot path truncations carried by the single-function full graph generation instruction are determined. The full page topology network corresponding to the target application category is extracted from the user interaction link graph. The full page topology network is then pruned according to the number of hotspot path truncations to obtain and output the first user interaction link graph of the single-function full dimension. A single-function full-scale graph generation command refers to a trigger signal issued by a management terminal or business analyst, instructing the graph calculation engine to construct a macroscopic full-scale topology link based on the global dashboard traffic characteristics of a specific independent business category.

[0097] The target application category refers to the independent business domain identifier in a multi-functional application, which is divided according to the core business logic or functional modules (such as the wealth management category, credit category, and personal account category in financial applications).

[0098] The number of hotspot path truncations refers to the threshold boundary of the number of high-frequency edges that are pre-set inside the graph computing engine and used to filter dense networks, so as to limit the output graph structure to retain only the top N core interaction paths in terms of traffic.

[0099] The full page topology network refers to a local functional subgraph formed by all page vertices and directed edges that have a direct mapping relationship with the target application category, extracted from the global base of the user interaction link graph by extraction operators.

[0100] The first user interaction link graph refers to a graph data entity that, after being filtered by major categories and pruned based on hot traffic, can intuitively and faithfully present the macro-frequency backbone flow characteristics of a specific business segment.

[0101] In a schematic manner, when the service platform's data interface receives a single-function full graph generation instruction sent by an external terminal or generated by the local terminal, it first calls the instruction parsing module to extract the target application category identifier and the number of hotspot path truncations carried in the instruction. The control graph calculation engine uses the target application category identifier as the core retrieval key to perform a global topology subgraph deep scan in the generated user interaction link graph, physically stripping all graph page vertices and directed edges belonging to this functional domain, thereby extracting the full page topology network corresponding to the target application category from the cache. The topology pruning operator is called to sort the traffic access counts on each directed edge in the full page topology network in reverse order, strictly retaining the high-frequency core edges that rank at the top and whose number is equal to the number of hotspot path truncations, and physically truncating and destroying low-frequency edge behavior branches that exceed this number, thereby completing the network pruning process and obtaining and outputting the first user interaction link graph of the single-function full dimension. 2) When a single-user link graph generation instruction is received, the target user identifier and target anchoring event carried by the single-user link graph generation instruction are determined. Based on the target user identifier and target anchoring event, the target user event sequence chain is determined. The target user event sequence chain is used as a topology boundary mask to perform network path pruning processing on the user interaction link graph, and a second user interaction link graph with a fine dimension of single user is obtained and output.

[0102] The single-user link graph generation instruction is a trigger signal that indicates the execution of precise drill-down backtracking and private chain reconstruction for a specific single user's micro-personalized interaction trajectory.

[0103] The target user identifier refers to a string or numeric code (such as a user identification code ID) used globally and uniquely to identify a specific physical user in an application system.

[0104] Target anchoring events refer to specific core business nodes or abnormal truncation behavior logs triggered by users during interaction (such as payment failure events or real-name authentication rejection events), which are used as benchmark focus anchors for micro-trajectory tracking.

[0105] The target user event sequence chain refers to a linear directed linked list of an ordered set of human-computer interaction events triggered by a specific user around the target anchor event in a temporal sequence.

[0106] Topological boundary masking refers to a topological matrix masking operator that uses linear event chains as boundary constraints to perform dynamic path projection, filtering, and intersection pruning on high-dimensional dense graph networks.

[0107] The second user interaction link graph refers to a refined personal behavior graph data entity generated by cropping the global graph using a topological boundary mask, which faithfully reproduces the micro-personalized movement characteristics of a specific single user.

[0108] In a schematic manner, when the data interface receives a single-user link graph generation command sent by an external terminal or generated locally, it parses out the target user identifier and target anchoring event carried in the command. Using the target user identifier as an index, it retrieves the full interaction log of the specific user within a preset period from a distributed non-relational database. By focusing on timestamps, it identifies consecutive ordered event nodes before and after the target anchoring event, thereby determining the target user event sequence chain. Subsequently, the control graph calculation engine instantiates the target user event sequence chain as a topology boundary mask and projects it onto the user interaction link graph as a physical pruning boundary. By performing primitive intersection calculations and modifying the pointers of non-target user paths in the adjacency table, it performs network path pruning processing on all redundant branches not covered by the sequence chain. Finally, it extracts graph entities containing only the micro-trajectory of the specific user from the global graph base, obtaining and outputting a second user interaction link graph with a fine-grained single-user dimension.

[0109] This specification introduces a single-function full graph and a single-user link graph response mechanism, endowing the system with dual-dimensional analysis capabilities, enabling precise drilling down from the macro-level overview to the micro-level behavior. Specifically, based on application category extraction and hotspot path pruning, it can automatically eliminate low-frequency interaction noise, presenting the core traffic backbone of specific functions with high fidelity, and improving the timeliness of macro-level bottleneck analysis. Simultaneously, by using the target user event sequence chain as a topology boundary mask to perform path pruning, it achieves precise and lossless extraction of specific individual micro-interaction trajectories from the massive graph base, suitable for deep causal backtracking scenarios such as anomaly tracing. This mechanism reduces the computational overhead of graphics rendering and retrieval while enhancing its engineering application value in multiple scenarios.

[0110] Optional, such as Figure 5 As shown, Figure 5 This is a flowchart illustrating an incremental update process. Specifically, it involves monitoring page operation behavior data of the application page and performing incremental updates on the topology behavior branches of the user interaction link graph based on this page operation behavior data. The following approach can be used as a reference: S402: Divide the trunk snapshot storage area into a differential incremental buffer; The backbone snapshot storage area refers to a physical or logical storage region explicitly partitioned within a background system's cache, in-memory database, or distributed storage cluster, specifically used for persistent snapshots of the global baseline topology. This area features high read throughput and typically maintains structural immutability until a global qualitative change occurs in the static baseline, thus isolating frequent online write loads.

[0111] The differential increment buffer refers to a temporary, high-speed physical or logical storage area in the background system, independent of the main snapshot storage area, used for low-latency, high-speed writing of real-time reporting behavior stream data. As an isolation flushing layer for asynchronous data stacking, it effectively smooths out peak traffic from high concurrency online, ensuring the matrix stability of subsequent differential comparison operations.

[0112] S404: Write the user interaction link graph as a global baseline graph into the backbone snapshot storage area, and write the monitored real-time page operation behavior data as an asynchronous incremental data stream into the difference incremental buffer; The global baseline graph, as a complete causal graph theory network representing the full business flow and page micro-operation structure, resides in the backbone snapshot storage area and is used as a benchmark entity for subsequent differential comparison and incremental correction.

[0113] Asynchronous incremental data streams refer to user page operation log streams that are captured asynchronously in real time from the online client cluster and continuously pushed and reported. This data stream reflects the topology branch changes in the latest running state of the system in a non-blocking, time-sequential incremental manner, and has unstructured or semi-structured streaming characteristics before entering the differential incremental buffer.

[0114] S406: Perform graph structure difference comparison between the asynchronous incremental data stream in the differential incremental buffer and the trunk snapshot storage area to obtain change behavior segments including local topological changes, and perform incremental update processing on the user interaction link graph based on the change behavior segments.

[0115] Graph structure difference comparison refers to the algebraic comparison calculation process that uses graph theory difference algorithms, such as graph diff, to perform algebraic operations such as intersection, union, and difference on the vertex and edge sets of two independent graph networks, thereby accurately identifying the residual values ​​of the two networks in terms of local topological connectivity, node existence state, and edge flow weights.

[0116] A change behavior segment refers to a non-homogeneous local topological behavior subgraph structure that is separated after graph structure difference comparison and is manifested in local memory as node addition or deletion, directed edge flow direction drift or flow weight mutation. It is used as the minimum incremental entity for performing local graph patching hot plugging.

[0117] Indicatively, executing S402-S406, the service platform first invokes the graph database's memory management mechanism to perform segmentation and access permission configuration of physical memory addresses in the server's main memory space or distributed cache nodes. This explicitly isolates and partitions a high-throughput read-only backbone snapshot storage area and a high-concurrency write differential incremental buffer. The control graph computing engine, through memory copying or handle redirection, solidifies the user interaction link graph generated in the previous steps as a global baseline graph and writes it to the backbone snapshot storage area for baseline protection. Simultaneously, the service platform's streaming log listener encapsulates the real-time page operation behavior data monitored online into a non-blocking asynchronous incremental data stream, continuously writing it to the differential incremental buffer for streaming temporary storage using a high-speed append mode. Subsequently, it concurrently reads the... The asynchronous incremental data stream in the differential incremental buffer and the global baseline graph stored in the backbone snapshot storage area are analyzed and calculated. The unique node identifiers and directed edge attributes contained in both are extracted. A graph structure difference comparison algorithm based on hash key-value comparison is called to perform a depth-first network element structure differential comparison on the data in the two areas. This accurately isolates and extracts the change behavior fragments that represent topology drift residuals and include local topology changes. Finally, the change behavior fragments are used as graph structure patches. The memory patch update mechanism of the graph database is used to directly address and cut into the corresponding local topology behavior branches in the backbone snapshot storage area. Under the physical premise that the connectivity of the routing structure of other unchanged branches in the global system is not affected at all, local vertex dynamic insertion, deletion or weight accumulation processing is performed on the target branch of the user interaction link graph.

[0118] This specification describes how a dual-zone isolated storage architecture decouples high-throughput static graph baseline reading from high-concurrency dynamic behavior stream writing, effectively mitigating sudden online traffic spikes. Simultaneously, a graph structure difference comparison algorithm based on low-dimensionality reduction residual extraction is introduced, transforming global large-scale graph reconstruction into patch extraction for locally altered topologies. This allows the graph computing engine to locate only affected specific topology behavior branches and perform in-place memory patch hot-plugging. While fully preserving the connectivity constraints of the existing graph network backbone and unchanged branches, it achieves low-latency, low-computing-cost streaming dynamic fresh-keeping updates of the panoramic user interaction link graph. This avoids the severe waste of server cluster CPU and I / O resources caused by performing a full global recalculation of massive historical logs, improving the timeliness and resource throughput efficiency of the big data operation and maintenance monitoring system.

[0119] Optionally, the incremental update of the user interaction link graph based on the changed behavior fragment can be performed in the following manner: Step E2: Extract the page identifier and element identifier from the changed behavior fragment, and construct an index table based on the page identifier and the element identifier; Page identifiers refer to string codes used to uniquely anchor a specific application view, mini-program window, or web page interface within a global network; element identifiers refer to cascading tags or metadata handles used to uniquely identify a specific foreground visual control component, button control, or physical touchpoint within the physical boundaries of a single page.

[0120] An index table is a lightweight double-key data structure dynamically allocated in memory or a caching system to establish one-to-one or one-to-many mappings between pages and elements and specific graph theory topology elements.

[0121] Step E4: Query the corresponding local change topology in the trunk snapshot storage area based on the index table; Local change topology refers to the set of affected local subgraphs or branch nodes that are spatially identical to the currently dynamically generated change behavior fragments in the global baseline graph residing in the trunk snapshot storage area and urgently require topology preservation and maintenance.

[0122] Step E6: Keep the global trunk topology in the trunk snapshot storage area unchanged, and hot-plug the changed behavior fragment into the local changed topology in the memory graph database for incremental updates.

[0123] The global backbone topology refers to the core skeleton network structure in the global base of the user interaction link diagram, which is used to represent the macroscopic routing flow status between independent application pages and is not affected by the current online microscopic behavior variation.

[0124] An in-memory graph database refers to a directed graph data management engine that resides in the main memory space of a server and is built based on graph theory adjacency lists or sparse matrix structures. It is used to perform high-performance reading, writing, and dynamic traversal of nodes and directed edges.

[0125] Hot-plugging refers to a local topology correction mechanism performed in a graph-based network under dynamic operation. It achieves local dynamic insertion and replacement of local topology branches by directly rewriting the in-degree and out-degree pointers of the adjacency list in main memory, without affecting the physical connectivity of other unchanged branch routing structures globally.

[0126] Incremental update refers to a low-computational-cost process for maintaining the freshness of streaming graph topology. It applies only the latest calculated local algebraic residuals to the target behavior branch, avoiding a global recalculation of the entire historical log.

[0127] Indicatively, steps E2-E6 are executed. The control graph computation engine first calls the object attribute extraction operator to perform a deep scan of the changed behavior fragments produced in the preceding steps. From this, it completely extracts the page identifier and element identifier of each mutated graph node. Using the page identifier as the primary key and the element identifier as the secondary key, it calls the hash constructor in the memory stack to construct an index table with a double-key mapping structure. The cascading index entries contained in the index table are read and sent to the memory management unit of the backbone snapshot storage area as the addressing basis. Through concurrent key-value retrieval under optimistic locking constraints, the physical node handle that perfectly matches the page identifier and element identifier is found in the resident global baseline graph. The system accurately locates the corresponding local change topology in the backbone snapshot storage area. Finally, the control graph calculation engine activates the hot-swappable in-situ update operator in the memory graph database. Under the strict physical premise of maintaining the complete closed loop and constant physical condition of the connectivity constraints, directed edge topology flow direction, and global disk traffic weight of the global backbone topology stored in the backbone snapshot storage area, the system dynamically stitches the root node pointer of the changed behavior segment to the target junction of the queried local change topology by directly rewriting the out-degree pointer address of the corresponding node in the memory adjacency table. This achieves lossless replacement of local behavior trajectories and dynamic accumulation of traffic weights at the memory level. Based on this seamless hot-swappable connection, the incremental update processing of the topology behavior branch is completed.

[0128] In this specification, based on steps E2-E6, the parallel and precise location of mutated nodes in the massive graph base is achieved using a double-key cascaded index table, shortening the addressing latency of differential topology. Relying on the hot-swap mechanism of in-situ pointer rewriting, under the premise of strictly ensuring the absolute constancy of global skeleton and unchanged path connectivity constraints, high-isolation local primitive hot correction is achieved, avoiding global cascade updates and topology data oscillations, and reducing the maintenance overhead and computing power load of the in-memory graph database.

[0129] The following will combine Figure 6 This manual provides a detailed description of the link diagram processing device provided. It should be noted that... Figure 6 The link graph processing device shown is used to execute this specification. Figures 1-5 The methods of the embodiments shown are illustrated only in connection with this specification for ease of explanation. For specific technical details not disclosed, please refer to this specification. Figures 1-5 The example shown.

[0130] Please see Figure 6This diagram illustrates the structure of the link graph processing device described in this specification. The link graph processing device 1 can be implemented as all or part of a device through software, hardware, or a combination of both. According to some embodiments, the link graph processing device 1 includes a data acquisition module 11, a sequence parsing module 12, a graph generation module 13, and an incremental update module 14, specifically used for: The data acquisition module 11 is used to acquire the full SPM data and page user operation sequence data for the application page; Sequence parsing module 12 is used to parse the full SPM data to determine the page jump path sequence and the cross-level jump bridge sequence from the click position to the page, and to determine the page operation topology sequence based on the page user operation sequence data; The graph generation module 13 is used to use the page jump transition sequence as the topology bridge directed edge, and to attach the page operation topology sequence as a local substructure to the interface node corresponding to the cross-level jump bridge sequence to generate an initial user interaction link graph. The incremental update module 14 is used to monitor page operation behavior data of the application page and perform incremental update processing of the topology behavior branches of the user interaction link graph based on the page operation behavior data.

[0131] Optionally, obtaining the full SPM data and page user operation sequence data for the application page includes: Collect raw application page stream logs reported by the client; User access event features, element exposure event features, and element click event features are extracted from the original flow logs of the application page. Based on the user access event features, element exposure event features, and element click event features, feature aggregation is performed to obtain the full SPM data. The user page behavior in the original flow log of the application page is parsed to obtain page user operation sequence data.

[0132] Optionally, the step of parsing the user page behavior in the original flow log of the application page to obtain page user operation sequence data includes: Page tracking events are identified from the original stream logs of the application page. Based on the DOM order in the DOM topology tree structure of the current application page, the page tracking events are spatially topologically sorted to construct a first data sequence. Event tracking points are identified from the original flow logs of the application page. The timestamp difference between the current event tracking point and the adjacent event tracking points is obtained. If the timestamp difference is within the preset time sequence verification window, the current event tracking point is sequentially included in the second data sequence. Based on the first data sequence and the second data sequence, sequence synthesis processing is performed to obtain page user operation sequence data.

[0133] Optionally, the step of using the cross-level jump bridging sequence as a topology bridging directed edge, and attaching the in-page operation topology sequence as a local substructure to the page node corresponding to the inter-page jump bridging sequence to generate an initial user interaction link graph includes: The full SPM data, the inter-page jump path sequence, the cross-level jump bridging sequence, and the intra-page operation topology sequence are encapsulated into a general graph data structure; In the general graph data structure, the operation topology sequence within the page is taken as a long behavioral path. The path is cut according to the click position breakpoint in the long behavioral path to obtain multiple spatial local micro operation sub-sequences. In the general graph data structure, the cross-level jump bridging sequence is used to perform cross-application page topology mounting concatenation processing on each of the micro-operation sub-sequences to generate an initial user interaction link graph.

[0134] Optionally, the step of performing cross-application page topology mounting and concatenation processing on each of the micro-operation sub-sequences using the cross-level jump bridging sequence to generate an initial user interaction link graph includes: Based on the cross-level jump bridging sequence, the target mounting parent page node corresponding to each of the micro-operation sub-sequences in the general graph data structure is determined. Based on the target mounted parent page node, each of the micro-operation sub-sequences is associated as a local operation subtree with the corresponding target mounted parent page node, and the same interactive elements in the micro-operation sub-sequences are deduplicated and cached in the general graph data structure to build the interactive association relationship between sibling operation nodes; The target is attached to the end operation node of the application page to which the parent page node belongs, and then connected to the root node of the next application page through directed edges within the general graph data structure to obtain the full topology network element structure. The entire topology network element structure is transformed into a tree structure to generate and output an initial user interaction link diagram.

[0135] Optionally, after performing tree structure transformation on the full topology network element structure and generating and outputting the initial user interaction link diagram, the method further includes: When a single-function full graph generation instruction is received, the target application category and the number of hotspot path truncations carried by the single-function full graph generation instruction are determined. The full page topology network corresponding to the target application category is extracted from the user interaction link graph. The full page topology network is then pruned according to the number of hotspot path truncations to obtain and output the first user interaction link graph of the single-function full dimension. When a single-user link graph generation instruction is received, the target user identifier and target anchoring event carried by the single-user link graph generation instruction are determined. Based on the target user identifier and target anchoring event, a target user event sequence chain is determined. The target user event sequence chain is used as a topology boundary mask to perform network path pruning processing on the user interaction link graph, and a second user interaction link graph with a fine dimension of single user is obtained and output.

[0136] Optionally, the monitoring of page operation behavior data of the application page, and the incremental update processing of the topology behavior branches of the user interaction link graph based on the page operation behavior data, include: Divide the trunk snapshot storage area into a differential increment buffer; The user interaction link graph is written as a global baseline graph into the backbone snapshot storage area, and the monitored real-time page operation behavior data is written as an asynchronous incremental data stream into the difference incremental buffer. The asynchronous incremental data stream in the differential incremental buffer is compared with the backbone snapshot storage area to obtain change behavior segments including local topological changes. The user interaction link graph is then incrementally updated based on the change behavior segments.

[0137] Optionally, the incremental update process of the user interaction link graph based on the changed behavior fragment includes: Extract the page identifier and element identifier from the changed behavior fragment, and construct an index table based on the page identifier and the element identifier; Based on the index table, query the corresponding local change topology in the trunk snapshot storage area; Keeping the global trunk topology in the trunk snapshot storage area unchanged, the changed behavior fragments are hot-plugged into the locally changed topology in the memory graph database for incremental updates.

[0138] It should be noted that the link graph processing device provided in the above embodiments is only illustrated by the division of the above functional modules when executing the link graph processing method. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. In addition, the link graph processing device and the link graph processing method embodiments provided in the above embodiments belong to the same concept, and the implementation process is detailed in the method embodiments, which will not be repeated here.

[0139] The serial numbers in this specification are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0140] This specification also provides a computer storage medium capable of storing multiple instructions adapted to be loaded and executed by a processor as described above. Figures 1-5 The link graph processing method of the illustrated embodiment can be found in the following documentation for its specific execution process. Figures 1-5 The specific details of the illustrated embodiments will not be elaborated here.

[0141] This specification also provides a computer program product that stores at least one instruction, said at least one instruction being loaded and executed by the processor as described above. Figures 1-5 The link graph processing method of the illustrated embodiment can be found in the following documentation for its specific execution process. Figures 1-5 The specific details of the illustrated embodiments will not be elaborated here.

[0142] Please refer to Figure 7 This is a structural block diagram of an electronic device provided in an embodiment of this specification. The electronic device in this specification may include one or more of the following components: a processor 1010, a memory 1020, an input device 1030, an output device 1040, and a bus 1050. The processor 1010, memory 1020, input device 1030, and output device 1040 may be connected to each other via the bus 1050.

[0143] Processor 1010 may include one or more processing cores. Processor 1010 connects to various parts of the electronic device using various interfaces and lines, and performs various functions and processes data by running or executing instructions, programs, code sets, or instruction sets stored in memory 1020, and by calling data stored in memory 1020. Optionally, processor 1010 may be implemented using at least one hardware form of digital signal processing (DSP), field-programmable gate array (FPGA), or programmable logic array (PLA). Processor 1010 may integrate one or more of a central processing unit (CPU), graphics processing unit (GPU), and modem. The CPU primarily handles the operating system, user interface, and applications; the GPU is responsible for rendering and drawing the displayed content; and the modem handles wireless communication. It is understood that the modem may also not be integrated into processor 1010 and may be implemented separately through a communication chip.

[0144] The memory 1020 may include random access memory (RAM) or read-only memory (ROM). Optionally, the memory 1020 may include non-transitory computer-readable storage medium. The memory 1020 may be used to store instructions, programs, code, code sets, or instruction sets.

[0145] The input device 1030 is used to receive input instructions or data, and includes, but is not limited to, a keyboard, mouse, camera, microphone, or touch device. The output device 1040 is used to output instructions or data, and includes, but is not limited to, a display device and a speaker. In this embodiment, the input device 1030 can be a temperature sensor for acquiring the operating temperature of the electronic device. The output device 1040 can be a speaker for outputting audio signals.

[0146] In addition, those skilled in the art will understand that the structure of the electronic device shown in the above figures does not constitute a limitation on the electronic device. The electronic device may include more or fewer components than shown, or combine certain components, or have different component arrangements. For example, the electronic device may also include radio frequency circuits, input units, sensors, audio circuits, wireless fidelity (WIFI) modules, power supplies, Bluetooth modules, etc., which will not be described in detail here.

[0147] In the embodiments of this specification, the executing entity for each step can be the electronic device described above. Optionally, the executing entity for each step can be the operating system of the electronic device. The operating system can be Android, iOS, or other operating systems; this specification does not limit this.

[0148] exist Figure 7 In the electronic device, the processor 1010 can be used to call a program stored in the memory 1020 and execute it to implement the link graph processing method as described in the various method embodiments of this specification.

[0149] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory, or random access memory, etc.

[0150] It should be noted that the information (including but not limited to user device information, user personal information, etc.), data (including but not limited to data used for analysis, stored data, displayed data, etc.), and signals involved in the embodiments of this specification are all authorized by the user or fully authorized by all parties, and the collection, use, and processing of related data must comply with the relevant laws, regulations, and standards of the relevant countries and regions. For example, the features, data, and information involved in this specification were all obtained under full authorization.

[0151] The above-disclosed embodiments are merely preferred embodiments of this specification and should not be construed as limiting the scope of this specification. Therefore, any equivalent variations made in accordance with the claims of this specification shall still fall within the scope of this specification.

Claims

1. A link graph processing method, characterized in that, The method includes: Obtain full SPM data and user action sequence data for the application page; The full SPM data is analyzed to determine the page jump path sequence and the cross-level jump bridging sequence from the click point to the page, and the page operation topology sequence is determined based on the page user operation sequence data; Using the page jump transition sequence as the topology bridge directed edge, the page operation topology sequence as the local substructure is mounted to the interface node corresponding to the cross-level jump bridge sequence to generate an initial user interaction link graph. Monitor page operation behavior data of the application page, and perform incremental update processing of the topology behavior branches of the user interaction link graph based on the page operation behavior data.

2. The method according to claim 1, characterized in that, The acquisition of full SPM data and user operation sequence data for the application page includes: Collect raw application page stream logs reported by the client; User access event features, element exposure event features, and element click event features are extracted from the original flow logs of the application page. Based on the user access event features, element exposure event features, and element click event features, feature aggregation is performed to obtain the full SPM data. The user page behavior in the original flow log of the application page is parsed to obtain page user operation sequence data.

3. The method according to claim 2, characterized in that, The step of parsing user page behavior in the original flow log of the application page to obtain page user operation sequence data includes: Page tracking events are identified from the original stream logs of the application page. Based on the DOM order in the DOM topology tree structure of the current application page, the page tracking events are spatially topologically sorted to construct a first data sequence. Event tracking points are identified from the original flow logs of the application page. The timestamp difference between the current event tracking point and the adjacent event tracking points is obtained. If the timestamp difference is within the preset time sequence verification window, the current event tracking point is sequentially included in the second data sequence. Based on the first data sequence and the second data sequence, sequence synthesis processing is performed to obtain page user operation sequence data.

4. The graph generation method according to claim 1, characterized in that, The step of using the cross-level jump bridge sequence as the topology bridge directed edge, and attaching the in-page operation topology sequence as a local substructure to the page node corresponding to the inter-page jump bridge sequence to generate an initial user interaction link graph includes: The full SPM data, the inter-page jump path sequence, the cross-level jump bridging sequence, and the intra-page operation topology sequence are encapsulated into a general graph data structure; In the general graph data structure, the operation topology sequence within the page is taken as a long behavioral path. The path is cut according to the click position breakpoint in the long behavioral path to obtain multiple spatial local micro operation sub-sequences. In the general graph data structure, the cross-level jump bridging sequence is used to perform cross-application page topology mounting concatenation processing on each of the micro-operation sub-sequences to generate an initial user interaction link graph.

5. The graph generation method according to claim 4, characterized in that, The step of performing cross-application page topology mounting and concatenation processing on each of the micro-operation sub-sequences using the cross-level jump bridging sequence to generate an initial user interaction link graph includes: Based on the cross-level jump bridging sequence, the target mounting parent page node corresponding to each of the micro-operation sub-sequences in the general graph data structure is determined. Based on the target mounted parent page node, each of the micro-operation sub-sequences is associated as a local operation subtree with the corresponding target mounted parent page node, and the same interactive elements in the micro-operation sub-sequences are deduplicated and cached in the general graph data structure to build the interactive association relationship between sibling operation nodes; The target is attached to the end operation node of the application page to which the parent page node belongs, and then connected to the root node of the next application page through directed edges within the general graph data structure to obtain the full topology network element structure. The entire topology network element structure is transformed into a tree structure to generate and output an initial user interaction link diagram.

6. The graph generation method according to claim 5, characterized in that, After performing tree structure transformation on the full topology network element structure to generate and output the initial user interaction link graph, the process further includes: When a single-function full graph generation instruction is received, the target application category and the number of hotspot path truncations carried by the single-function full graph generation instruction are determined. The full page topology network corresponding to the target application category is extracted from the user interaction link graph. The full page topology network is then pruned according to the number of hotspot path truncations to obtain and output the first user interaction link graph of the single-function full dimension. When a single-user link graph generation instruction is received, the target user identifier and target anchoring event carried by the single-user link graph generation instruction are determined. Based on the target user identifier and target anchoring event, a target user event sequence chain is determined. The target user event sequence chain is used as a topology boundary mask to perform network path pruning processing on the user interaction link graph, and a second user interaction link graph with a fine dimension of single user is obtained and output.

7. The graph generation method according to claim 1, characterized in that, The monitoring application page's page operation behavior data, based on which the user interaction link graph is incrementally updated with topology behavior branches, includes: Divide the trunk snapshot storage area into a differential increment buffer; The user interaction link graph is written as a global baseline graph into the backbone snapshot storage area, and the monitored real-time page operation behavior data is written as an asynchronous incremental data stream into the difference incremental buffer. The asynchronous incremental data stream in the differential incremental buffer is compared with the backbone snapshot storage area to obtain change behavior segments including local topological changes. The user interaction link graph is then incrementally updated based on the change behavior segments.

8. The graph generation method according to claim 7, characterized in that, The incremental update process of the user interaction link graph based on the changed behavior fragment includes: Extract the page identifier and element identifier from the changed behavior fragment, and construct an index table based on the page identifier and the element identifier; Based on the index table, query the corresponding local change topology in the trunk snapshot storage area; Keeping the global trunk topology in the trunk snapshot storage area unchanged, the changed behavior fragments are hot-plugged into the locally changed topology in the memory graph database for incremental updates.

9. A link graph processing device, characterized in that, The device includes: The data acquisition module is used to acquire full SPM data for the application page and page user operation sequence data; The sequence parsing module is used to parse the full SPM data to determine the page jump path sequence and the cross-level jump bridge sequence from the click point to the page, and to determine the page operation topology sequence based on the page user operation sequence data; The graph generation module is used to use the page jump transition sequence as the topology bridge directed edge, and to attach the page operation topology sequence as a local substructure to the interface node corresponding to the cross-level jump bridge sequence to generate an initial user interaction link graph. The incremental update module is used to monitor page operation behavior data of the application page and perform incremental update processing of the topology behavior branches of the user interaction link graph based on the page operation behavior data.

10. A computer storage medium storing a plurality of instructions adapted for loading by a processor and executing the steps of the method as claimed in any one of claims 1 to 8.

11. A computer program product storing at least one instruction, said at least one instruction being loaded by a processor and executing the steps of the method as claimed in any one of claims 1 to 8.

12. An electronic device, comprising: A processor and a memory; wherein the memory stores a computer program adapted to be loaded by the processor and to execute the steps of the method as described in any one of claims 1 to 8.