Intelligent archive management method and system

Abstract

The invention is suitable for the technical field of archive management, and provides an intelligent archive management method and system, and the method comprises the steps: constructing a prediction model based on an attention mechanism bidirectional space-time neural network through integrating historical archive circulation data, organization structure topological features and real-time environment variable parameters, and generating archive circulation path probability distribution; monitoring file state change in real time, and triggering a path deduction engine to generate a future 72-hour prediction map; and establishing a cross-department collaborative optimization mechanism, and dynamically generating an alternative approval path. The system comprises a multi-dimensional space-time analysis module, a prediction model construction module, a change monitoring and path deduction module and an alternative path generation module. According to the scheme, the problems of insufficient multi-source data fusion, low prediction precision and lagging flow adjustment in the prior art are solved, accurate prediction and intelligent regulation and control of archive circulation are realized, the management efficiency and risk response capability are improved, and the method is suitable for intelligent archive management scenes of various organizations.

Description

Technical Field

[0001] This invention belongs to the field of archives management technology, and in particular relates to an intelligent archives management method and system. Background Technology

[0002] With the deepening development of informatization and digitalization, the field of archival management is transforming from traditional paper-based management to intelligent and automated management. Currently, the industry has gradually introduced data collection and analysis technologies, but it still faces challenges such as the difficulty of integrating multi-source heterogeneous data, insufficient accuracy in predicting archival circulation paths, and rigid cross-departmental collaboration processes. The demand for intelligent management is driving technology towards multi-dimensional data integration, dynamic prediction, and real-time optimization, but existing solutions have significant technical bottlenecks in spatiotemporal feature modeling, real-time risk perception, and adaptive process adjustment.

[0003] In existing technologies, some document management systems achieve simple process recording by integrating historical flow data, or perform limited optimization of approval paths based on rule engines. For example, some solutions use a single data source (such as relying solely on process logs) for statistical analysis, or preset approval paths through static organizational charts, lacking dynamic awareness of real-time environmental variables (such as staff on-duty rate and equipment status); a few solutions introduce machine learning models, but most use unidirectional neural networks, without combining attention mechanisms with bidirectional spatiotemporal features for coupled analysis, making it difficult to capture the temporal dependencies and spatial topological relationships in document flow. Summary of the Invention

[0004] The purpose of this invention is to provide an intelligent file management method and system, which aims to solve the technical problems existing in the prior art as identified in the background art.

[0005] This invention is implemented as follows: an intelligent archive management method, the method comprising:

[0006] Integrate historical archive transfer data, organizational structure topology features, and real-time environmental variable parameters;

[0007] A prediction model is constructed based on all integrated content, and a bidirectional spatiotemporal neural network with attention mechanism is used for dynamic coupling analysis. The input of the prediction model includes the current archive attribute features, organizational structure topology features and real-time environmental variable parameters, and the output is the probability distribution of the archive flow path.

[0008] Real-time monitoring of archive metadata change events and business process status codes; identification of archive status change nodes; triggering the path inference engine at archive status change nodes; and transforming the probability distribution of archive circulation paths into a predicted path map for the next 72 hours.

[0009] Establish a cross-departmental collaborative optimization mechanism. When a risk of delay is detected in the approval path, activate the process replanning algorithm based on the organizational structure topology characteristics to generate an alternative path solution.

[0010] As a further aspect of the present invention, the integration of historical archive transfer data, organizational structure topology features, and real-time environmental variable parameters specifically includes:

[0011] Collect multi-source heterogeneous data, including file attributes, process logs, LDAP organizational directories, network traffic characteristics, and sensor time-series data;

[0012] Construct a spatiotemporal feature engineering pipeline, extract archive lifecycle stage markers from historical archive circulation data, generate departmental collaboration density maps from organizational structure topology features, and extract process node throughput features from real-time environmental variable parameters;

[0013] A hierarchical spatiotemporal encoder is established to transform discrete process log events into continuous spatiotemporal vector representations. The encoder includes a temporal convolutional layer and a spatial graph attention layer.

[0014] Establish a dynamic knowledge graph, define a four-dimensional relationship model between archive entities and departments, personnel, and equipment, and construct a semantic network with time constraints.

[0015] As a further aspect of the present invention, the construction of a prediction model based on all integrated content, and the dynamic coupling analysis using a bidirectional spatiotemporal neural network with an attention mechanism, specifically includes:

[0016] Data is collected in real time from historical archive circulation data, organizational structure topology features and real-time environmental variable parameters. Heterogeneous data is accessed using data interfaces, and a unified spatiotemporal synchronization strategy is used to align the data sources in time and space dimensions. Through feature standardization and normalization, an archive management dataset is obtained.

[0017] Convert discrete information of each category in the archive management dataset into a continuous spatiotemporal vector representation;

[0018] A bidirectional spatiotemporal neural network model is constructed, and spatiotemporal vectors fused from the archive management dataset are used for analysis. The future state trend is captured through forward propagation, and historical dependencies and prior information are incorporated through backpropagation to form a prediction model. The prediction model is trained using the archive management dataset corresponding to the historical archive circulation data.

[0019] When a change in the status of an archive triggers a prediction requirement, the feature vector of the latest preprocessed archive management dataset is input into the prediction model to generate a probability distribution of the archive transfer path in real time for the next 72 hours.

[0020] As a further aspect of the present invention, the real-time monitoring of archive metadata change events and business process status codes, identification of archive status change nodes, and triggering of a path inference engine at the archive status change node to transform the probability distribution of archive transfer paths into a predicted path map for the next 72 hours, specifically includes:

[0021] By linking with the data interface, we can continuously monitor the changes in archival metadata and business process status codes, compare and analyze the collected real-time data, and identify status change nodes in the archival management dataset.

[0022] When a node indicating a change in file status is detected, the path deduction engine is triggered. Combining the probability distribution of file transfer paths, it automatically receives the latest prediction results and uses the output of the prediction model as input for deduction.

[0023] The probability distribution of the archive transfer path is transformed into an intuitive 72-hour predicted path map, and the credibility of each path node is evaluated.

[0024] As a further aspect of the present invention, the establishment of a cross-departmental collaborative optimization mechanism, when a risk of delay is detected in the approval path, activates a process replanning algorithm based on organizational structure topology features to generate an alternative path solution, specifically including:

[0025] Real-time monitoring of cross-departmental approval paths; dynamic comparison and evaluation of key links and node statuses in the approval process by linking with organizational structure topology features in the file management dataset; identification of delay risks in the approval process.

[0026] When an approval node with a risk of delay is identified, the process replanning algorithm is triggered. The algorithm uses a hierarchical strategy to determine the scope of the delay risk and builds a temporary task queue to provide feedback on the impending approval delay risk.

[0027] By utilizing the latest organizational structure topology features and departmental collaboration density map information, combined with the results of delay risk identification, a process replanning algorithm is used to dynamically solve the approval path and calculate several alternative path schemes in real time.

[0028] The credibility of several alternative paths is assessed in real time, and comprehensive optimization is performed based on the current approval process status and departmental resource distribution to select the optimal alternative path.

[0029] Another object of the present invention is to provide an intelligent file management system, the system comprising:

[0030] The multi-dimensional spatiotemporal analysis module is used to integrate historical archive transfer data, organizational structure topology features, and real-time environmental variable parameters;

[0031] The prediction model building module is used to build a prediction model based on all integrated content. It uses a bidirectional spatiotemporal neural network with an attention mechanism for dynamic coupling analysis. The input of the prediction model includes the current archive attribute features, organizational structure topology features and real-time environmental variable parameters, and the output is the probability distribution of the archive flow path.

[0032] The change monitoring and path prediction module is used to monitor file metadata change events and business process status codes in real time, identify file status change nodes, and trigger the path prediction engine at file status change nodes to convert the probability distribution of file circulation paths into a predicted path map for the next 72 hours.

[0033] The alternative path generation module is used to establish a cross-departmental collaborative optimization mechanism. When a risk of delay is detected in the approval path, the process replanning algorithm is activated based on the organizational structure topology features to generate an alternative path solution.

[0034] As a further embodiment of the present invention, the multi-dimensional spatiotemporal analysis module includes:

[0035] The multi-source heterogeneous data acquisition unit is used to collect multi-source heterogeneous data, including file attributes, process logs, LDAP organizational directories, network traffic characteristics, and sensor time-series data.

[0036] The spatiotemporal feature engineering pipeline construction unit is used to construct spatiotemporal feature engineering pipelines, extract archive life cycle stage markers from historical archive circulation data, generate departmental collaboration density maps from organizational structure topology features, and extract process node throughput features from real-time environmental variable parameters.

[0037] A hierarchical spatiotemporal encoder establishment unit is used to establish a hierarchical spatiotemporal encoder, which transforms discrete process log events into continuous spatiotemporal vector representations. The encoder includes a temporal convolutional layer and a spatial graph attention layer.

[0038] The dynamic knowledge graph construction unit is used to build a dynamic knowledge graph, define the four-dimensional relationship pattern between archive entities and departments, personnel, and equipment, and construct a semantic network with time constraints.

[0039] As a further embodiment of the present invention, the prediction model construction module includes:

[0040] The heterogeneous data access and spatiotemporal synchronization unit is used to collect data in real time from historical archive circulation data, organizational structure topology features and real-time environmental variable parameters. It uses data interfaces to access heterogeneous data and uses a unified spatiotemporal synchronization strategy to align each data source in time and space dimensions. Through feature standardization and normalization processing, it obtains the archive management dataset.

[0041] The dataset feature standardization and processing unit is used to convert discrete information of various categories in the archive management dataset into continuous spatiotemporal vector representations;

[0042] The model building unit is used to construct a bidirectional spatiotemporal neural network model. It uses the spatiotemporal vectors fused from the archive management dataset for analysis, captures future state trends through forward propagation, and incorporates historical dependencies and prior information through backpropagation to form a prediction model. The prediction model is then trained using the archive management dataset corresponding to the historical archive circulation data.

[0043] The status prediction and path probability analysis unit is used to input the feature vector of the latest preprocessed archive management dataset into the prediction model when the archive status change triggers the prediction requirement, and generate the probability distribution of archive flow path in real time for the next 72 hours.

[0044] As a further embodiment of the present invention, the change monitoring and path deduction module includes:

[0045] The change monitoring and status analysis unit is used to continuously monitor file metadata change events and business process status codes by linking with the data interface, and to compare and analyze the collected real-time data to identify status change nodes in the file management dataset.

[0046] The path deduction unit is used to trigger the path deduction engine when a node of file status change is identified. It combines the probability distribution of file flow path, automatically receives the latest prediction results, and uses the output of the prediction model as input for deduction.

[0047] The prediction path map generation unit is used to convert the probability distribution of archive flow paths into an intuitive 72-hour prediction path map, while simultaneously performing a credibility assessment on each path node.

[0048] As a further embodiment of the present invention, the alternative path generation module includes:

[0049] The real-time monitoring unit for approval paths is used to monitor cross-departmental approval paths in real time. By linking with the organizational structure topology features in the file management dataset, it dynamically compares and evaluates the status of key links and nodes in the approval process, and identifies the risk of delays in the approval process.

[0050] The delay risk identification unit is used to trigger the process replanning algorithm when an approval node with delay risk is identified. It uses a hierarchical strategy to determine the scope of the delay risk and builds a temporary task queue to provide feedback on the impending approval delay risk.

[0051] The dynamic approval path replanning unit is used to dynamically solve the approval path by utilizing the latest organizational structure topology features and departmental collaboration density map information, combined with the results of delay risk identification, and using a process replanning algorithm to calculate several alternative path schemes in real time.

[0052] The scheme credibility assessment unit is used to conduct real-time credibility assessments of several alternative path schemes, and to comprehensively optimize them by combining the current approval process status and departmental resource distribution to select the optimal alternative path.

[0053] The beneficial effects of this invention are:

[0054] This invention integrates historical archive circulation data, organizational structure topology features, and real-time environmental variable parameters through a multi-dimensional spatiotemporal analysis module to construct a hierarchical spatiotemporal encoder and dynamic knowledge graph, thereby realizing a full-dimensional representation of archive management data and laying the foundation for accurate prediction. A bidirectional spatiotemporal neural network with an attention mechanism is used to construct a prediction model. By dynamically coupling spatiotemporal features through forward and backward propagation, a probability distribution of circulation paths in the next 72 hours is generated, improving the prediction accuracy.

[0055] Real-time monitoring of changes in archive metadata and business process status triggers the path deduction engine to generate a visual prediction map, enabling millisecond-level risk response; the cross-departmental collaboration optimization mechanism activates the process replanning algorithm based on organizational structure topology features, dynamically generating alternative path solutions to reduce the risk of approval delays.

[0056] This solution has significantly improved the efficiency of document circulation, shortened the approval cycle, and upgraded the system from passive recording to proactive prediction and intelligent control, thereby enhancing the adaptability and automation level of document management under complex organizational structures. Attached Figure Description

[0057] Figure 1 A flowchart illustrating an intelligent file management method provided in an embodiment of the present invention;

[0058] Figure 2 A flowchart for integrating historical archive transfer data, organizational structure topology features, and real-time environmental variable parameters provided in an embodiment of the present invention;

[0059] Figure 3 A flowchart of dynamic coupling analysis using a bidirectional spatiotemporal neural network with an attention mechanism provided in an embodiment of the present invention;

[0060] Figure 4 A flowchart for identifying archive status change nodes provided in an embodiment of the present invention;

[0061] Figure 5 A flowchart for establishing a cross-departmental collaboration optimization mechanism provided in an embodiment of the present invention;

[0062] Figure 6 A structural block diagram of an intelligent archive management system provided in an embodiment of the present invention;

[0063] Figure 7 This is a structural block diagram of the multi-dimensional spatiotemporal analysis module provided in an embodiment of the present invention;

[0064] Figure 8 This is a structural block diagram of the prediction model construction module provided in an embodiment of the present invention;

[0065] Figure 9 This is a structural block diagram of the change monitoring and path deduction module provided in an embodiment of the present invention;

[0066] Figure 10 This is a structural block diagram of the alternative path generation module provided in an embodiment of the present invention. Detailed Implementation

[0067] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0068] Figure 1 A flowchart of an intelligent file management method provided in an embodiment of the present invention is shown below. Figure 1 As shown, the method includes:

[0069] S100 integrates historical archive transfer data, organizational structure topology features, and real-time environmental variable parameters;

[0070] This step first collects data from multiple heterogeneous data sources, such as file attributes, process logs, LDAP organizational directories, network traffic characteristics, and sensor time-series data. For example, when a company implements this, it simultaneously accesses the approval process logs of personnel files, departmental structure data in the OA system, and temperature and humidity data of the storage environment collected by sensors in the computer room to ensure coverage of the entire lifecycle of file circulation and related environmental information.

[0071] By using a spatiotemporal feature engineering pipeline to perform in-depth data processing, lifecycle stage markers such as file creation, approval, and archiving are extracted from historical flow data. Density maps reflecting the frequency of departmental collaboration are generated based on organizational structure. At the same time, throughput characteristics of each process node are extracted from real-time environmental variables. For example, in the scenario of financial reimbursement file management, by analyzing the approval time of each department at different times, time series characteristics of node processing efficiency are formed.

[0072] A hierarchical spatiotemporal encoder is used to transform discrete process log events into continuous spatiotemporal vector representations. The temporal convolutional layer in the encoder can capture periodic time patterns such as monthly approval peaks, while the spatial graph attention layer can identify close collaborations such as between the finance department and the purchasing department in the reimbursement process, enabling unstructured data to have machine-understandable semantic representations.

[0073] Construct a dynamic knowledge graph to define a four-dimensional relationship model between archive entities and departments, personnel, and equipment. For example, a contract archive will be associated with elements such as the drafter, legal review department, storage server, and approval deadline, forming a semantic network with time constraints. When a department head is on a business trip, the system can perceive the changes in the processing capacity of that node in real time through the knowledge graph.

[0074] This step, through the fusion of multi-source data and the construction of deep features, upgrades record management from fragmented information to systematic knowledge. The collection of multi-source heterogeneous data breaks down data silos, enabling the system to comprehensively understand record management scenarios from three dimensions: historical flow patterns, organizational structure relationships, and real-time environmental changes. For example, in hospital medical record management, by integrating medical record flow records, departmental structure, and medical staff scheduling data, the risk of approval delays for special cases can be accurately predicted.

[0075] The spatiotemporal characteristic engineering pipeline makes implicit knowledge of time and space dimensions explicit. The departmental collaboration density map can intuitively reflect the bottleneck nodes of cross-departmental processes. The throughput characteristics of process nodes can be used for real-time load balancing. After a manufacturing company applied it, it shortened the time spent on cross-departmental approval of production files.

[0076] The hierarchical spatiotemporal encoder solves the problem of adapting discrete data to continuous models. The temporal convolutional layer captures temporal patterns, enabling the system to predict processing pressures such as a surge in financial files at the end of a quarter. The spatial graph attention layer models the organizational structure, ensuring that process path predictions conform to the actual enterprise architecture. After institutional reform, a government unit's system automatically adapts to the new departmental relationships through the encoder, maintaining the accuracy of path prediction.

[0077] The time-constraint mechanism of dynamic knowledge graphs enables the system to have environmental awareness. When equipment fails or personnel change, the knowledge graph can update the relationship links in real time. For example, after a bank's file management system detects a scanning equipment failure, it automatically switches the file entry process to the backup equipment node, ensuring the efficiency of the process.

[0078] This comprehensive data integration and feature construction lays a solid foundation for the accuracy of subsequent prediction models, the real-time nature of path deduction, and the effectiveness of cross-departmental collaborative optimization, enabling archive management to upgrade from passive recording to proactive prediction and intelligent control.

[0079] like Figure 2As shown, the integration of historical archive transfer data, organizational structure topology features, and real-time environmental variable parameters specifically includes:

[0080] S110 collects multi-source heterogeneous data, including file attributes, process logs, LDAP organizational directories, network traffic characteristics, and sensor time-series data;

[0081] S120 constructs a spatiotemporal feature engineering pipeline, extracts archive lifecycle stage markers from historical archive circulation data, generates departmental collaboration density maps from organizational structure topology features, and extracts process node throughput features from real-time environmental variable parameters.

[0082] S130, Establish a hierarchical spatiotemporal encoder to transform discrete process log events into continuous spatiotemporal vector representations. The encoder includes a temporal convolutional layer and a spatial graph attention layer.

[0083] S140: Establish a dynamic knowledge graph, define a four-dimensional relationship model between archive entities and departments, personnel, and equipment, and construct a semantic network with time constraints.

[0084] S200, a prediction model is constructed based on all integrated content, and a bidirectional spatiotemporal neural network with attention mechanism is used for dynamic coupling analysis. The input of the prediction model includes the current archive attribute features, organizational structure topology features and real-time environmental variable parameters, and the output is the probability distribution of the archive flow path.

[0085] This step essentially achieves accurate prediction of the document circulation path through multi-dimensional data fusion and intelligent algorithm modeling. Specifically, the system first collects data in real time from historical document circulation data, organizational structure topology features, and real-time environmental variable parameters. For example, when a large manufacturing enterprise implemented this system, it simultaneously accessed the production document approval records of the past three years, departmental structure data from the ERP system, and real-time business traffic data collected by workshop sensors. Heterogeneous data was accessed through a unified data interface, and a spatiotemporal synchronization strategy was used to align the timestamps of different systems (such as unifying the approval time of the OA system and the circulation time of the document management system to the millisecond level). The system then used Z-Score standardization to process features of different dimensions, such as approval time and number of people in departments, to form a standardized document management dataset.

[0086] Next, the discrete information in the dataset (such as department names and approval status) is transformed into continuous spatiotemporal vectors through an embedding layer. For example, "Finance Department" is encoded as a multidimensional vector containing collaboration frequency and historical processing efficiency. Then, a bidirectional spatiotemporal neural network model is constructed. The forward propagation path of this model captures the future state trend of file circulation (such as predicting the possible time of the next 3 approval nodes based on the current progress), while the backward propagation path incorporates historical dependencies (such as referring to the delay patterns of similar files at similar stages). At the same time, an attention mechanism automatically focuses on key nodes (such as assigning higher weight to approval positions with a retention rate of more than 30% in the past six months).

[0087] The model is trained using historical archive transfer data. For example, a hospital uses the transfer records of 20,000 medical records from the past as a sample and optimizes the network parameters through backpropagation, so that the model can learn implicit patterns such as "the probability of emergency medical records being delayed when approved by the medical affairs department between 18:00 and 8:00 the next day increases by 40%".

[0088] When a change in the status of a file (such as from "drafting" to "submitting for approval") triggers a forecasting requirement, the system inputs the latest file attribute characteristics (such as security classification and number of pages), organizational structure topology characteristics (such as the current staff attendance rate in the approval department), and real-time environmental variables (such as the number of tasks currently pending in the department) into the model to generate the probability distribution of each flow path in real time within the next 72 hours. For example, it predicts that a procurement contract file has a 65% probability of flowing through the path of "procurement department → finance department → vice president in charge" and a 25% probability of changing to the path of "procurement department → vice president assistant → vice president in charge" due to a temporary meeting in the finance department.

[0089] This step, through dynamic coupling analysis and intelligent modeling, upgrades the prediction of file circulation from experience-driven to data-driven. The spatiotemporal synchronization and standardized processing of multi-source data solves the problem of heterogeneous data fusion. For example, after integrating personnel file circulation data with organizational structure data from the government OA system, a government unit improved the model's accuracy in predicting cross-departmental approval paths.

[0090] The forward and backward propagation mechanism of bidirectional spatiotemporal neural networks enables the model to capture the current development trend of the process and make corrections based on historical experience. For example, in the management of archives of scientific research projects in universities, the model learns the seasonal pattern of "delay in the archiving process of scientific research results due to professional title evaluation in March every year" through backpropagation and makes probabilistic adjustments to the circulation path of relevant archives in advance.

[0091] The introduction of the attention mechanism effectively filters out noisy information and focuses on key influencing factors. For example, in the approval of bank loan files, the model automatically identifies the high-weight event "the credit department manager is on a business trip" through the attention mechanism, increases the retention probability of this node from the usual 15% to 40%, and adjusts the probability distribution of subsequent paths accordingly.

[0092] The real-time generated probability distribution of the flow path provides data support for dynamic control. After applying this model, an e-commerce company was able to predict the flow bottleneck of order files 72 hours in advance and automatically trigger alternative approval paths, which improved the order processing efficiency by 35% during the promotion period.

[0093] This predictive mechanism, which integrates spatiotemporal characteristics, historical experience, and real-time environment, can not only accurately predict potential risks in the transfer of archives, but also provide a scientific basis for subsequent path deduction and cross-departmental collaboration optimization, enabling the archive management system to have the intelligent capability of "early perception and proactive planning".

[0094] like Figure 3 As shown, the prediction model built based on all integrated content, and the dynamic coupling analysis using a bidirectional spatiotemporal neural network with an attention mechanism, specifically includes:

[0095] S210 collects data in real time from historical archive circulation data, organizational structure topology features and real-time environmental variable parameters, uses data interfaces to access heterogeneous data, and uses a unified spatiotemporal synchronization strategy to align each data source in time and space dimensions. Through feature standardization and normalization processing, an archive management dataset is obtained.

[0096] S220 converts discrete information of each category in the archive management dataset into a continuous spatiotemporal vector representation;

[0097] S230, construct a bidirectional spatiotemporal neural network model, use the spatiotemporal vector after centralized fusion of the archive management data for analysis, capture future state trends through forward propagation, and use backpropagation to incorporate historical dependencies and prior information to form a prediction model, and use the archive management dataset corresponding to the historical archive circulation data to train the prediction model.

[0098] S240: When a change in the status of an archive triggers a prediction requirement, the feature vector of the latest preprocessed archive management dataset is input into the prediction model to generate the probability distribution of the archive transfer path in real time for the next 72 hours.

[0099] S300 monitors file metadata change events and business process status codes in real time, identifies file status change nodes, and triggers the path inference engine at file status change nodes to transform the probability distribution of file circulation paths into a predicted path map for the next 72 hours.

[0100] This step, through data interface linkage with the OA system and document management database, monitors changes in document metadata (such as changes in security classification or responsible person) and business process status codes (such as "pending," "processing," and "completed" statuses of approval nodes) at millisecond-level frequency. For example, in credit document management, a financial institution can capture metadata change events such as "loan contract version update" and "approver temporarily absent" in real time, while simultaneously connecting with the core business system to obtain process status code sequences. When the monitored data is compared with preset rules and a status change node is identified (such as a document changing from "department review" to "cross-departmental review"), the system automatically triggers the path deduction engine, retrieves the probability distribution of the flow path generated by step S200 (such as a certain type of loan document having a 30% probability of being stuck in the compliance department during the cross-departmental review stage), and combines the latest organizational structure topology characteristics (such as the number of tasks currently pending in the compliance department) and real-time environmental variables (such as the compliance department head attending an emergency meeting), using algorithms such as Monte Carlo simulation to predict the potential flow path for the next 72 hours.

[0101] The final generated prediction path map is presented in the form of a visual node link. Each node is labeled with the probability of transfer (e.g., the probability of the path "Compliance Department → Deputy Director in charge" is 75%) and the credibility score (calculated based on the consistency of historical data, e.g., the historical accuracy rate of this path is 88%). For example, in the project file map of a real estate company, the system predicts that the probability of the "Planning Department Approval" node being delayed due to personnel being on business trips is 60%, and highlights the alternative path "Planning Department → Deputy Director's Approval".

[0102] This step establishes a dynamic management closed loop of "real-time perception - intelligent simulation - visual early warning". The real-time monitoring mechanism achieves millisecond-level response to changes in the status of archives, improving the timeliness of risk detection by more than 90% compared to the traditional timed inspection mode. For example, a government archives system can adjust the circulation path in advance to avoid non-classified nodes by capturing "classification upgrade" events in real time.

[0103] The dynamic triggering mechanism of the path projection engine enables the predictive model to be updated in real time according to business changes. For example, in the management of college admission files, when a metadata change of "adjustment of enrollment plan" is detected, the system immediately re-projects the admission notice issuance process, providing a warning of the risk of delay at the mailing node 72 hours in advance. The visualized predictive path map provides managers with an intuitive decision support tool. A manufacturing company, by highlighting the congestion risk of the "supply chain file → finance department" path on the map, preemptively allocated three finance personnel to provide support, improving the processing efficiency of this link by 40%.

[0104] The credibility assessment mechanism enhances the interpretability of the prediction results. Banks can select path schemes with a credibility of ≥80% to implement, thereby controlling the prediction error rate within 15%.

[0105] This ability to transform abstract probability distributions into intuitive and visual graphs not only makes the implicit risks of archive transfer explicit, but also provides ample time for manual intervention or automatic system optimization through a 72-hour advance prediction window, realizing a paradigm shift in archive management from "post-event processing" to "pre-event control".

[0106] like Figure 4 As shown, the real-time monitoring of archive metadata change events and business process status codes, identification of archive status change nodes, and triggering of the path inference engine at the archive status change node, transforming the probability distribution of archive circulation paths into a predicted path map for the next 72 hours, specifically includes:

[0107] S310, through linkage with the data interface, continuously monitors archival metadata change events and business process status codes, and compares and analyzes the collected real-time data to identify status change nodes in the archival management dataset.

[0108] S320: When a file status change node is detected, the path inference engine is triggered. Combining the probability distribution of the file flow path, the latest prediction result is automatically received and the prediction model output is used as the input for inference.

[0109] S330 transforms the probability distribution of archive transfer paths into an intuitive 72-hour predicted path map, while simultaneously conducting credibility assessments for each path node.

[0110] S400 establishes a cross-departmental collaborative optimization mechanism. When a risk of delay is detected in the approval path, a process replanning algorithm is activated based on the organizational structure topology characteristics to generate an alternative path solution.

[0111] This step involves real-time integration with the OA approval system and the document management database to continuously monitor the node status of cross-department approval paths. For example, in the procurement contract approval process of a multinational company, the system obtains data such as the number of pending tasks and staff on-duty rate of each node in the "procurement department → legal department → finance department" path in real time. This data is then linked with the departmental collaboration density map in the organizational structure topology for analysis. When the number of pending contracts in the legal department exceeds the threshold and the core reviewer is on a business trip, the system automatically identifies that there is an 80% risk of delay at that node.

[0112] Once risk identification is triggered, the process replanning algorithm is activated. It assesses the scope of impact through a tiered strategy (e.g., determining whether the delay affects subsequent financial payment milestones) and constructs a temporary task queue to push risk warnings to relevant departments. For example, in the equipment procurement document approval process of a manufacturing company, if it detects a delay in the technical department's review node due to an expert review meeting, it immediately generates a temporary task queue to prompt the procurement department to prepare supplementary materials in advance. The algorithm combines the latest departmental collaboration density map (e.g., a recent 20% increase in collaboration frequency between the finance and audit departments) and organizational structure topology features to dynamically solve for alternative paths. For example, it adjusts the conventional path "Technology Department → Finance Department" to the alternative path "Technology Department → Audit Department → Finance Department," while simultaneously calculating the processing time and resource utilization rate of each solution.

[0113] Ultimately, the optimal solution is selected through a credibility assessment model (which considers parameters such as historical success rate and current departmental resource load). For example, in the approval of drug procurement files in a hospital, the system assessed the credibility of the path "Pharmacy Department → Procurement Department → Vice President in charge" as 92%, which shortened the approval time by 4 hours compared to the original path.

[0114] This step establishes a cross-departmental collaborative intelligent control system encompassing "risk prediction, intelligent restructuring, and dynamic optimization." The real-time monitoring mechanism enables the system to proactively identify potential congestion risks, advancing risk detection time by more than 72 hours compared to traditional manual inspection methods.

[0115] The dynamic solution capability of the process replanning algorithm breaks the limitations of fixed approval paths. For example, during a major promotional period, an e-commerce company automatically switched to a temporary path of "customer service department → expedited processing team → finance department" in the approval of order files through the algorithm, which improved the approval efficiency of high-priority orders by 50%.

[0116] The optimization strategy based on organizational structure topology ensures that the alternative path conforms to the actual collaboration mode of the enterprise. In the cross-provincial project file approval of a certain energy group, the system automatically selects the regional headquarters as the intermediate node according to the historical collaboration density of each branch, which reduces the cross-regional approval time by 35%.

[0117] The credibility assessment and resource load linkage mechanism ensures the feasibility of the solution. In the loan file approval process, the bank uses this mechanism to prioritize the path with the lower current load of the credit department and risk control department, increasing the approval rate from 78% to 91%.

[0118] This intelligent cross-departmental collaboration optimization mechanism not only solves the inefficiency caused by node delays in traditional approval processes, but also realizes the dynamic allocation of archive circulation resources through data-driven path reconstruction. This enables approval processes under complex organizational structures to have adaptability and anti-interference capabilities, truly achieving a core breakthrough in archive management from "process recording" to "intelligent control".

[0119] like Figure 5 As shown, the establishment of a cross-departmental collaborative optimization mechanism, when a risk of delay is detected in the approval path, activates a process replanning algorithm based on organizational structure topology features to generate alternative path solutions, specifically including:

[0120] S410 monitors cross-departmental approval paths in real time. By linking with the organizational structure topology features in the document management dataset, it dynamically compares and evaluates the status of key links and nodes in the approval process, and identifies the risk of delays in the approval process.

[0121] S420: When an approval node with a risk of delay is identified, the process replanning algorithm is triggered. The algorithm uses a hierarchical strategy to determine the scope of the delay risk and builds a temporary task queue to provide feedback on the impending approval delay risk.

[0122] S430 utilizes the latest organizational structure topology features and departmental collaboration density map information, combined with the results of delay risk identification, and employs a process replanning algorithm to dynamically solve the approval path and calculate several alternative path solutions in real time.

[0123] S440 performs real-time credibility assessments on several alternative path options and comprehensively optimizes them based on the current approval process status and departmental resource distribution to select the optimal alternative path.

[0124] Figure 6 A structural block diagram of an intelligent archive management system provided in an embodiment of the present invention is shown below. Figure 6 As shown, the system includes:

[0125] The multi-dimensional spatiotemporal analysis module is used to integrate historical archive transfer data, organizational structure topology features, and real-time environmental variable parameters;

[0126] The prediction model building module is used to build a prediction model based on all integrated content. It uses a bidirectional spatiotemporal neural network with an attention mechanism for dynamic coupling analysis. The input of the prediction model includes the current archive attribute features, organizational structure topology features and real-time environmental variable parameters, and the output is the probability distribution of the archive flow path.

[0127] The change monitoring and path prediction module is used to monitor file metadata change events and business process status codes in real time, identify file status change nodes, and trigger the path prediction engine at file status change nodes to convert the probability distribution of file circulation paths into a predicted path map for the next 72 hours.

[0128] The alternative path generation module is used to establish a cross-departmental collaborative optimization mechanism. When a risk of delay is detected in the approval path, the process replanning algorithm is activated based on the organizational structure topology features to generate an alternative path solution.

[0129] like Figure 7 As shown, the multi-dimensional spatiotemporal analysis module includes:

[0130] The multi-source heterogeneous data acquisition unit is used to collect multi-source heterogeneous data, including file attributes, process logs, LDAP organizational directories, network traffic characteristics, and sensor time-series data.

[0131] The spatiotemporal feature engineering pipeline construction unit is used to construct spatiotemporal feature engineering pipelines, extract archive life cycle stage markers from historical archive circulation data, generate departmental collaboration density maps from organizational structure topology features, and extract process node throughput features from real-time environmental variable parameters.

[0132] A hierarchical spatiotemporal encoder establishment unit is used to establish a hierarchical spatiotemporal encoder, which transforms discrete process log events into continuous spatiotemporal vector representations. The encoder includes a temporal convolutional layer and a spatial graph attention layer.

[0133] The dynamic knowledge graph construction unit is used to build a dynamic knowledge graph, define the four-dimensional relationship pattern between archive entities and departments, personnel, and equipment, and construct a semantic network with time constraints.

[0134] like Figure 8 As shown, the prediction model building module includes:

[0135] The heterogeneous data access and spatiotemporal synchronization unit is used to collect data in real time from historical archive circulation data, organizational structure topology features and real-time environmental variable parameters. It uses data interfaces to access heterogeneous data and uses a unified spatiotemporal synchronization strategy to align each data source in time and space dimensions. Through feature standardization and normalization processing, it obtains the archive management dataset.

[0136] The dataset feature standardization and processing unit is used to convert discrete information of various categories in the archive management dataset into continuous spatiotemporal vector representations;

[0137] The model building unit is used to construct a bidirectional spatiotemporal neural network model. It uses the spatiotemporal vectors fused from the archive management dataset for analysis, captures future state trends through forward propagation, and incorporates historical dependencies and prior information through backpropagation to form a prediction model. The prediction model is then trained using the archive management dataset corresponding to the historical archive circulation data.

[0138] The status prediction and path probability analysis unit is used to input the feature vector of the latest preprocessed archive management dataset into the prediction model when the archive status change triggers the prediction requirement, and generate the probability distribution of archive flow path in real time for the next 72 hours.

[0139] like Figure 9As shown, the change monitoring and path prediction module includes:

[0140] The change monitoring and status analysis unit is used to continuously monitor file metadata change events and business process status codes by linking with the data interface, and to compare and analyze the collected real-time data to identify status change nodes in the file management dataset.

[0141] The path deduction unit is used to trigger the path deduction engine when a node of file status change is identified. It combines the probability distribution of file flow path, automatically receives the latest prediction results, and uses the output of the prediction model as input for deduction.

[0142] The prediction path map generation unit is used to convert the probability distribution of archive flow paths into an intuitive 72-hour prediction path map, while simultaneously performing a credibility assessment on each path node.

[0143] like Figure 10 As shown, the alternative path generation module includes:

[0144] The real-time monitoring unit for approval paths is used to monitor cross-departmental approval paths in real time. By linking with the organizational structure topology features in the file management dataset, it dynamically compares and evaluates the status of key links and nodes in the approval process, and identifies the risk of delays in the approval process.

[0145] The delay risk identification unit is used to trigger the process replanning algorithm when an approval node with delay risk is identified. It uses a hierarchical strategy to determine the scope of the delay risk and builds a temporary task queue to provide feedback on the impending approval delay risk.

[0146] The dynamic approval path replanning unit is used to dynamically solve the approval path by utilizing the latest organizational structure topology features and departmental collaboration density map information, combined with the results of delay risk identification, and using a process replanning algorithm to calculate several alternative path schemes in real time.

[0147] The scheme credibility assessment unit is used to conduct real-time credibility assessments of several alternative path schemes, and to comprehensively optimize them by combining the current approval process status and departmental resource distribution to select the optimal alternative path.

[0148] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0149] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

[0150] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An intelligent archive management method, characterized in that, The method includes: Integrate historical archive transfer data, organizational structure topology features, and real-time environmental variable parameters; A prediction model is constructed based on all integrated content, and a bidirectional spatiotemporal neural network with attention mechanism is used for dynamic coupling analysis. The input of the prediction model includes the current archive attribute features, organizational structure topology features and real-time environmental variable parameters, and the output is the probability distribution of the archive flow path. Real-time monitoring of archive metadata change events and business process status codes; identification of archive status change nodes; triggering the path inference engine at archive status change nodes; and transforming the probability distribution of archive circulation paths into a predicted path map for the next 72 hours. Establish a cross-departmental collaborative optimization mechanism. When a risk of delay is detected in the approval path, activate the process replanning algorithm based on the organizational structure topology characteristics to generate an alternative path solution.

2. The method according to claim 1, characterized in that, The integration of historical archive transfer data, organizational structure topology features, and real-time environmental variable parameters specifically includes: Collect multi-source heterogeneous data, including file attributes, process logs, LDAP organizational directories, network traffic characteristics, and sensor time-series data; Construct a spatiotemporal feature engineering pipeline, extract archive lifecycle stage markers from historical archive circulation data, generate departmental collaboration density maps from organizational structure topology features, and extract process node throughput features from real-time environmental variable parameters; A hierarchical spatiotemporal encoder is established to transform discrete process log events into continuous spatiotemporal vector representations. The encoder includes a temporal convolutional layer and a spatial graph attention layer. Establish a dynamic knowledge graph, define a four-dimensional relationship model between archive entities and departments, personnel, and equipment, and construct a semantic network with time constraints.

3. The method according to claim 2, characterized in that, The prediction model, built upon the integrated content, employs a bidirectional spatiotemporal neural network with an attention mechanism for dynamic coupling analysis, specifically including: Data is collected in real time from historical archive circulation data, organizational structure topology features and real-time environmental variable parameters. Heterogeneous data is accessed using data interfaces, and a unified spatiotemporal synchronization strategy is used to align the data sources in time and space dimensions. Through feature standardization and normalization, an archive management dataset is obtained. Convert discrete information of each category in the archive management dataset into a continuous spatiotemporal vector representation; A bidirectional spatiotemporal neural network model is constructed, and spatiotemporal vectors fused from the archive management dataset are used for analysis. The future state trend is captured through forward propagation, and historical dependencies and prior information are incorporated through backpropagation to form a prediction model. The prediction model is trained using the archive management dataset corresponding to the historical archive circulation data. When a change in the status of an archive triggers a prediction requirement, the feature vector of the latest preprocessed archive management dataset is input into the prediction model to generate a probability distribution of the archive transfer path in real time for the next 72 hours.

4. The method according to claim 3, characterized in that, The real-time monitoring of archive metadata change events and business process status codes identifies archive status change nodes. At these nodes, a path prediction engine is triggered to transform the probability distribution of archive transfer paths into a predicted path map for the next 72 hours. Specifically, this includes: By linking with the data interface, we can continuously monitor the changes in archival metadata and business process status codes, compare and analyze the collected real-time data, and identify status change nodes in the archival management dataset. When a node indicating a change in file status is detected, the path deduction engine is triggered. Combining the probability distribution of file transfer paths, it automatically receives the latest prediction results and uses the output of the prediction model as input for deduction. The probability distribution of the archive transfer path is transformed into an intuitive 72-hour predicted path map, and the credibility of each path node is evaluated.

5. The method according to claim 4, characterized in that, The aforementioned cross-departmental collaborative optimization mechanism, when detecting a risk of delays in the approval process, activates a process replanning algorithm based on organizational structure topology features to generate alternative path solutions, specifically including: Real-time monitoring of cross-departmental approval paths; dynamic comparison and evaluation of key links and node statuses in the approval process by linking with organizational structure topology features in the file management dataset; identification of delay risks in the approval process. When an approval node with a risk of delay is identified, the process replanning algorithm is triggered. The algorithm uses a hierarchical strategy to determine the scope of the delay risk and builds a temporary task queue to provide feedback on the impending approval delay risk. By utilizing the latest organizational structure topology features and departmental collaboration density map information, combined with the results of delay risk identification, a process replanning algorithm is used to dynamically solve the approval path and calculate several alternative path schemes in real time. The credibility of several alternative paths is assessed in real time, and comprehensive optimization is performed based on the current approval process status and departmental resource distribution to select the optimal alternative path.

6. An intelligent archive management system, characterized in that, The system includes: The multi-dimensional spatiotemporal analysis module is used to integrate historical archive transfer data, organizational structure topology features, and real-time environmental variable parameters; The prediction model building module is used to build a prediction model based on all integrated content. It uses a bidirectional spatiotemporal neural network with an attention mechanism for dynamic coupling analysis. The input of the prediction model includes the current archive attribute features, organizational structure topology features and real-time environmental variable parameters, and the output is the probability distribution of the archive flow path. The change monitoring and path prediction module is used to monitor file metadata change events and business process status codes in real time, identify file status change nodes, and trigger the path prediction engine at file status change nodes to convert the probability distribution of file circulation paths into a predicted path map for the next 72 hours. The alternative path generation module is used to establish a cross-departmental collaborative optimization mechanism. When a risk of delay is detected in the approval path, the process replanning algorithm is activated based on the organizational structure topology features to generate an alternative path solution.

7. The system according to claim 6, characterized in that, The multi-dimensional spatiotemporal analysis module includes: The multi-source heterogeneous data acquisition unit is used to collect multi-source heterogeneous data, including file attributes, process logs, LDAP organizational directories, network traffic characteristics, and sensor time-series data. The spatiotemporal feature engineering pipeline construction unit is used to construct spatiotemporal feature engineering pipelines, extract archive life cycle stage markers from historical archive circulation data, generate departmental collaboration density maps from organizational structure topology features, and extract process node throughput features from real-time environmental variable parameters. A hierarchical spatiotemporal encoder establishment unit is used to establish a hierarchical spatiotemporal encoder, which transforms discrete process log events into continuous spatiotemporal vector representations. The encoder includes a temporal convolutional layer and a spatial graph attention layer. The dynamic knowledge graph construction unit is used to build a dynamic knowledge graph, define the four-dimensional relationship pattern between archive entities and departments, personnel, and equipment, and construct a semantic network with time constraints.

8. The system according to claim 7, characterized in that, The prediction model construction module includes: The heterogeneous data access and spatiotemporal synchronization unit is used to collect data in real time from historical archive circulation data, organizational structure topology features and real-time environmental variable parameters. It uses data interfaces to access heterogeneous data and uses a unified spatiotemporal synchronization strategy to align each data source in time and space dimensions. Through feature standardization and normalization processing, it obtains the archive management dataset. The dataset feature standardization and processing unit is used to convert discrete information of various categories in the archive management dataset into continuous spatiotemporal vector representations; The model building unit is used to construct a bidirectional spatiotemporal neural network model. It uses the spatiotemporal vectors fused from the archive management dataset for analysis, captures future state trends through forward propagation, and incorporates historical dependencies and prior information through backpropagation to form a prediction model. The prediction model is then trained using the archive management dataset corresponding to the historical archive circulation data. The status prediction and path probability analysis unit is used to input the feature vector of the latest preprocessed archive management dataset into the prediction model when the archive status change triggers the prediction requirement, and generate the probability distribution of archive flow path in real time for the next 72 hours.

9. The system according to claim 8, characterized in that, The change monitoring and path prediction module includes: The change monitoring and status analysis unit is used to continuously monitor file metadata change events and business process status codes by linking with the data interface, and to compare and analyze the collected real-time data to identify status change nodes in the file management dataset. The path deduction unit is used to trigger the path deduction engine when Dangdang identifies a node where the file status changes. It combines the probability distribution of the file flow path, automatically receives the latest prediction results, and uses the output of the prediction model as input for deduction. The prediction path map generation unit is used to convert the probability distribution of archive flow paths into an intuitive 72-hour prediction path map, while simultaneously performing a credibility assessment on each path node.

10. The system according to claim 9, characterized in that, The alternative path generation module includes: The real-time monitoring unit for approval paths is used to monitor cross-departmental approval paths in real time. By linking with the organizational structure topology features in the file management dataset, it dynamically compares and evaluates the status of key links and nodes in the approval process, and identifies the risk of delays in the approval process. The delay risk identification unit is used to trigger the process replanning algorithm when an approval node with delay risk is identified. It uses a hierarchical strategy to determine the scope of the delay risk and builds a temporary task queue to provide feedback on the impending approval delay risk. The dynamic approval path replanning unit is used to dynamically solve the approval path by utilizing the latest organizational structure topology features and departmental collaboration density map information, combined with the results of delay risk identification, and using a process replanning algorithm to calculate several alternative path schemes in real time. The scheme credibility assessment unit is used to conduct real-time credibility assessments of several alternative path schemes, and to comprehensively optimize them by combining the current approval process status and departmental resource distribution to select the optimal alternative path.

Images