A university library comprehensive intelligent management and control system based on digital twinning

Abstract

The application discloses a kind of based on digital twinning's comprehensive intelligentization management and control system of university library, comprising: information modeling module, for constructing state entity and behavior entity, constitute state twin system;Conflict analysis module, for constructing conflict behavior atlas, marking conflict hotspot books;Similarity calculation module, for according to the similarity between book feature vector and reader behavior vector, generate candidate book set;Adjustment engine module, for using improved DICE algorithm to perform adjustment recommendation and feedback control, generate recommended book list;State update module, for monitoring the weight change of conflict edge and the change state of conflict density;Build state twin feedback update mechanism, generate book resource conflict change trend data, adjustment response analysis data and book resource state visualization interface.The application improves the adjustment efficiency of book borrowing conflict and the balance of resource utilization.

Description

Technical Field

[0001] This invention relates to the field of library resource management and intelligent information processing technology, and in particular to a comprehensive intelligent management and control system for university libraries based on digital twins. Background Technology

[0002] With the continuous advancement of smart campus and information technology construction in universities, university libraries urgently need to introduce more efficient intelligent technologies to support resource allocation, borrowing services, and behavior management. Existing library management systems mainly rely on static database-driven book retrieval services and simple rule-based borrowing process management. While they possess basic book information management and reservation functions, they are significantly inadequate in scenarios involving high concurrency and dynamically changing user behavior.

[0003] First, existing systems typically employ a fixed information architecture, lacking the ability to dynamically model reader behavior, borrowing conflicts, and changes in book status, thus failing to track the state evolution during the book borrowing process. Second, when faced with scenarios where multiple users simultaneously reserve popular books, traditional systems respond solely based on chronological order or fixed rules, failing to fully consider user preferences, historical behavior, and conflict mitigation paths, resulting in inefficient resource allocation and decreased reader satisfaction. Furthermore, existing recommendation mechanisms are mostly based on book content similarity or borrowing popularity rankings, making it difficult to effectively integrate book conflict status with individual reader behavioral characteristics, resulting in recommendations lacking specificity and failing to proactively alleviate resource congestion.

[0004] Therefore, how to provide a comprehensive intelligent management and control system for university libraries based on digital twins is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] One objective of this invention is to propose a comprehensive intelligent management and control system for university libraries based on digital twins. This invention fully utilizes state entity modeling, conflict behavior graph construction, book and reader feature similarity calculation, adjustment recommendation control and feedback update mechanism, and describes in detail the system method for realizing book resource conflict state identification, hot book replacement recommendation and resource state evolution tracking in virtual library space. It has the advantages of high resource allocation efficiency, strong conflict mitigation ability and high level of intelligent book management.

[0006] According to an embodiment of the present invention, a comprehensive intelligent management and control system for university libraries based on digital twins includes:

[0007] The information modeling module is used to collect information on book resources, reader behavior, reservation requests, and borrowing records to construct state entities and behavior entities; the state entities and behavior entities are then mapped to dynamic nodes in the virtual library space to form a state twin system.

[0008] The conflict analysis module is used to construct a conflict behavior graph, mapping each book to a graph node; if multiple readers initiate borrowing requests for the same book within a preset time period and some requests are not fulfilled, then conflict edges are constructed in the conflict behavior graph; the conflict density per unit time is calculated for the graph nodes; if the conflict density is greater than a first preset threshold, then the corresponding book is marked as a conflict hotspot book.

[0009] The similarity calculation module is used to extract keywords, category tags, topic types, and borrowing behavior tags of the conflict hot books to generate book feature vectors; combine the readers' historical borrowing behavior records to construct reader behavior vectors; and generate a candidate book set based on the similarity between the book feature vectors and the reader behavior vectors.

[0010] The adjustment engine module is used to perform adjustment recommendation and feedback control using an improved DICE algorithm to generate a recommended book list; the improved DICE algorithm introduces a conflict perception enhancement channel, a twin state gating structure, and a conflict resolution objective function;

[0011] The status update module is used to monitor the changes in the weight and density of conflict edges. If the cumulative decrease in the weight of a conflict edge within a set time window is greater than the second preset threshold and the conflict density is less than the third preset threshold, the conflict hotspot book mark of the corresponding book is canceled. A status twin feedback update mechanism is constructed to generate data on the trend of book resource conflict changes, adjustment response analysis data, and a visualization interface of book resource status.

[0012] Optionally, modules can be integrated using the following methods:

[0013] S1. Collect book resource information, reader behavior information, reservation request information, and borrowing record information to construct state entities and behavior entities; the state entities include reservation queue, borrowing status, and remaining reservation count; the behavior entities include borrowing preference tags, historical borrowing records, and reservation failure rate; map the state entities and behavior entities to dynamic nodes in the virtual library space to form a state twin system.

[0014] S2. Construct a conflict behavior graph based on the state twin system, and map each book as a graph node; if multiple readers initiate borrowing requests for the same book within a preset time period and some requests are not fulfilled, then construct conflict edges in the conflict behavior graph. The weight of the conflict edges is calculated based on the conflict frequency, reservation failure rate and time overlap, and the time range of the conflict behavior and the behavioral characteristics of the participating readers are recorded.

[0015] S3. Calculate the conflict density per unit time for the map nodes; if the conflict density is greater than a first preset threshold, mark the corresponding book as a conflict hotspot book;

[0016] S4. Extract keywords, category tags, theme types, and borrowing behavior tags from the conflict hotspot books to generate book feature vectors; construct reader behavior vectors by combining readers' historical borrowing behavior records, and generate a candidate book set based on the similarity between the book feature vectors and the reader behavior vectors.

[0017] S5. An improved DICE algorithm is used to perform adjustment recommendation and feedback control to generate a recommended book list; the improved DICE algorithm introduces a conflict perception enhancement channel, a twin state gating structure, and a conflict resolution objective function;

[0018] S6. Monitor the changes in the weight and density of conflict edges; if the cumulative decrease in the weight of a conflict edge within a set time window is greater than the second preset threshold and the conflict density is less than the third preset threshold, then cancel the conflict hotspot book mark for the corresponding book.

[0019] S7. Construct a state twin feedback update mechanism to synchronously update the borrowing request information, reservation result information, conflict behavior graph structure change information and reader response behavior information to the state twin system, perform joint update operations on state entities, behavior entities and conflict behavior graphs, and generate book resource conflict change trend data, adjustment response analysis data and book resource status visualization interface.

[0020] Optionally, S1 specifically includes:

[0021] Collect book resource information, including book identification information, book name information, catalog number information, storage location code information, classification code information, catalog quantity information, and available quantity information; collect reader behavior information, including reader identification information, access time information, borrowing operation identification information, reservation operation identification information, and return operation identification information; collect reservation request information, including reservation initiation time information, reserved book identification information, reservation request status information, and request response result information; collect borrowing record information, including borrowing time information, return time information, overdue status information, and default status information.

[0022] Construct a status entity, encode the reservation queue into a queue sequence data according to time order, mark the borrowing status as available and unavailable records, and record the remaining reservation count as a reservation count parameter; construct a behavior entity, label the borrowing preference tags according to the book category code, construct the historical borrowing records into a behavior trajectory sequence according to time order, and generate a failure rate parameter based on the ratio of the number of failed requests to the total number of requests.

[0023] The state entity and the behavior entity are respectively bound to the book identification information and the reader identification information, and mapped to the node data in the virtual library space. The node data includes node number information, node type identification information, node time attribute information and node state attribute information. Based on the node number information and time sequence information, node association data is generated to form a node structure data set of the state twin system.

[0024] Optionally, S2 specifically includes:

[0025] Write book identification information into the node identification field to generate graph node data records; sort the borrowing request information within a preset time period by time to form request time series data; perform grouping operation on the borrowing request records corresponding to the same book identification information to generate a set of request records for the same book;

[0026] Perform time overlap interval comparison on the same book request record set, and filter the time interval intersection record set; perform request status matching operation on the time interval intersection record set, and extract reservation request status information and request response result information; if there are unmet status records in the request response result information, generate conflict behavior marker data;

[0027] Conflict edge structure data is constructed based on conflict behavior labeling data. The conflict edge structure data includes starting node identification information, target node identification information, and edge number information. Counting processing is performed on conflict frequency information, proportional calculation processing is performed on reservation failure rate information, and interval superposition processing is performed on time overlap information. Weighted synthesis processing is then performed to generate conflict edge weight data. The conflict edge weight data, along with conflict behavior occurrence time range information and participating reader identification information, are written into the conflict behavior graph data set.

[0028] Optionally, S3 specifically includes:

[0029] Read the node data records in the conflict behavior graph, extract the node identification information and the associated conflict edge data; filter the conflict edge data within a preset statistical time interval to form a set of conflict edge data within a time window; perform a count accumulation process on the set of conflict edge data within the time window to generate a statistical result of the number of conflict occurrences.

[0030] The conflict density is generated by comparing the number of conflict occurrences with the time interval. The conflict density of each graph node is then compared node by node to form a sorted list of node conflict densities.

[0031] The node conflict density data in the conflict density sorting list is compared with a first preset threshold for threshold determination; if the node conflict density data is greater than the first preset threshold, the node identification information is written into the conflict hotspot marking table to form a set of conflict hotspot book identification records.

[0032] Optionally, S4 specifically includes:

[0033] Read the set of conflict hotspot book identifier records, and extract the basic book information fields for each book identifier record;

[0034] The book name information in the basic book information field is processed by word segmentation to extract keyword data; the classification code information is processed by label mapping to extract classification label data; the topic type information is processed by clustering and normalization to extract topic type labels; and the book-related borrowing behavior records are processed by label mining to extract borrowing behavior label data.

[0035] Keyword data, category tag data, topic type tag data, and borrowing behavior tag data are concatenated to construct a set of book feature descriptions; the set of book feature descriptions is then vectorized to generate book feature vectors.

[0036] Read the reader's historical borrowing records corresponding to the book identification records, and extract the book identification information and borrowing time sequence information of the borrowed books; perform feature extraction operations on the basic information of the borrowed books to obtain the keywords, category tags, theme types and borrowing behavior tags of the historically borrowed books;

[0037] The tag information of historical borrowed books is processed by feature encoding, and a feature sequence of reader borrowing behavior trajectory is constructed in chronological order; vector fusion processing is performed on the behavior trajectory feature sequence to generate reader behavior vectors;

[0038] The similarity between the book feature vector and the reader behavior vector is calculated to generate a similarity score. The similarity scores are then sorted, and the identification information of books with similarity scores greater than a preset similarity threshold is extracted to form a candidate book set.

[0039] Optionally, S5 specifically includes:

[0040] Read the candidate book set and the conflict hotspot book identifier record set, and construct a recommendation input vector for each candidate book and the corresponding conflict hotspot book. The recommendation input vector includes a book feature vector, a reader behavior vector, and a book conflict label vector. The book conflict label vector is processed by three-dimensional encoding based on the conflict density value, conflict edge weight, and conflict historical frequency.

[0041] Attention enhancement processing is applied to the book conflict label vector based on the recommendation input vector to generate a conflict-aware representation; the conflict-aware representation is then concatenated with the book feature vector and the reader behavior vector to construct a fusion representation tensor.

[0042] Gating and filtering operations are performed on the historical borrowing time sequence features in the reader behavior vector to obtain the state memory vector; the state memory vector is concatenated with the fusion representation tensor to form a set of candidate recommendation representations.

[0043] Based on the candidate recommendation representation set, a scoring and sorting operation is performed. Books with scores greater than the preset recommendation output threshold in the sorting results are used as the recommendation output set. The recommendation output set is compared with the original conflict hotspot book identification records. The conflict replacement success rate, reader acceptance feedback rate and recommendation response delay parameter are statistically analyzed, and the feedback control results are recorded.

[0044] By combining the feedback control results with the recommended output set, a list of recommended books is generated; the list of recommended books includes book identification information, a recommendation reason field, a reader acceptance status field, and a conflict mitigation index.

[0045] Optionally, the improved DICE algorithm introduces a conflict-aware enhancement channel, a twin-state gating structure, and a conflict resolution objective function, specifically as follows:

[0046] A conflict-aware enhancement channel is constructed by performing a convolutional attention fusion operation on the conflict label vector of the input book to generate a conflict-aware representation. The conflict label vector includes a conflict density value encoding, a conflict edge strength vector, and a conflict history cycle index. A normalization mapping operation is performed on the conflict-aware representation as the input path for conflict enhancement.

[0047] A twin state gating structure is constructed to receive the fusion vector of readers' historical borrowing behavior sequences and book features, and send them to the gating unit in batches according to time slices. The gating unit is equipped with a memory control gate, a conflict response gate, and a preference feedback gate. The control gate controls the activation level of borrowing behavior based on the borrowing time decay weight. The response gate activates the conflict feedback path based on historical conflict records. The feedback gate uses the recommended acceptance behavior as the basis for state update.

[0048] A conflict resolution objective function is constructed, and the output recommended book scores, conflict replacement success rate, reader acceptance status and historical conflict mitigation data are jointly evaluated. The conflict resolution objective function includes recommendation accuracy component, conflict coverage component and feedback gain component.

[0049] The gradient descent method is used to fit the conflict resolution objective function in multiple rounds, optimize the recommendation ranking structure and recommendation score threshold parameters, output the final recommended book list, and record the recommendation adjustment loss results and the book conflict mitigation trend value.

[0050] Optionally, S6 specifically includes:

[0051] Read the conflict behavior graph dataset, extract the starting node identifier, target node identifier, edge number, and conflict edge weight data from the conflict edge structure record, and establish a conflict edge monitoring set; set the monitoring time window parameters, perform time series segmentation on the conflict edge weight data, and construct a weight sequence vector;

[0052] Perform maximum and minimum value difference operation on the weight sequence vector of each conflict edge to generate a cumulative decrease parameter; perform a second preset threshold comparison operation on the cumulative decrease parameter to filter out the conflict edge identification information with a decrease greater than the second preset threshold and generate a weight change judgment set.

[0053] Read the conflict density data of the corresponding graph nodes, perform a third preset threshold comparison operation on each graph node, filter out the book node identification information with conflict density less than the third preset threshold, and generate a density change judgment set.

[0054] The intersection of the weight change judgment set and the density change judgment set is performed to extract the book node identification information that simultaneously satisfies the decrease magnitude greater than the second preset threshold and the conflict density less than the third preset threshold, thus forming the conflict hotspot cancellation set.

[0055] Remove the book node identifier information from the conflict hotspot cancellation set from the conflict hotspot marking table, update the conflict hotspot book identifier record set, and record the graph structure change status data.

[0056] Optionally, S7 specifically includes:

[0057] Collect borrowing request information, reservation result information, conflict behavior graph structure change information, and reader response behavior information; update the borrowing request information and reservation result information to the status entity and behavior entity, and adjust the reservation queue, borrowing status, remaining reservation count, and reservation failure rate parameters; update the conflict behavior graph structure change information to the graph node data and conflict edge data, and record the addition and removal status of conflict edges; write the reader response behavior information into the behavior entity, and update the borrowing behavior trajectory and borrowing preference tags.

[0058] Perform joint update operations on node data, edge data, and behavioral data to generate data on the changing trends of book resource conflicts, adjustment response analysis data, and a visualization interface of book resource status.

[0059] The beneficial effects of this invention are:

[0060] This invention addresses the issues of concentrated book reservations, frequent borrowing conflicts, and shortages of popular books in university libraries by constructing a state twin system of state entities and behavioral entities. It combines conflict behavior graph construction, conflict density determination, and conflict edge weight calculation mechanisms. An improved DICE algorithm is employed to introduce a conflict perception enhancement channel, a twin state gating structure, and a conflict resolution objective function. A candidate book set is generated based on the similarity between book feature vectors and reader behavior vectors, and a recommended book list is output through the fusion of representation tensors and state memory vectors. In the conflict hotspot state monitoring stage, a conflict edge monitoring set is established, and the set of conflict hotspot book identifiers is dynamically updated based on the dual criteria of weight decrease and conflict density determination. Finally, a state twin feedback update mechanism synchronizes the updates of borrowing request information, reservation result information, conflict behavior graph structure changes, and reader response behavior information, outputting data on book resource conflict change trends, adjustment response analysis data, and a visualized interface for book resource status. This invention effectively alleviates the backlog of conflict behaviors for popular books, achieves closed-loop control of book resource recommendation adjustment and conflict state evolution management, and improves book resource utilization efficiency and reader service response quality. Attached Figure Description

[0061] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0062] Figure 1 This is a schematic diagram of a comprehensive intelligent management and control system for university libraries based on digital twins, as proposed in this invention.

[0063] Figure 2 This is a flowchart of a comprehensive intelligent management and control system for university libraries based on digital twins, as proposed in this invention. Detailed Implementation

[0064] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0065] refer to Figure 1 A comprehensive intelligent management and control system for university libraries based on digital twins includes:

[0066] The information modeling module is used to collect information on book resources, reader behavior, reservation requests, and borrowing records to construct state entities and behavior entities; the state entities and behavior entities are then mapped to dynamic nodes in the virtual library space to form a state twin system.

[0067] The conflict analysis module is used to construct a conflict behavior graph, mapping each book to a graph node; if multiple readers initiate borrowing requests for the same book within a preset time period and some requests are not fulfilled, then conflict edges are constructed in the conflict behavior graph; the conflict density per unit time is calculated for the graph nodes; if the conflict density is greater than a first preset threshold, then the corresponding book is marked as a conflict hotspot book.

[0068] The similarity calculation module is used to extract keywords, category tags, topic types, and borrowing behavior tags of the conflict hot books to generate book feature vectors; combine the readers' historical borrowing behavior records to construct reader behavior vectors; and generate a candidate book set based on the similarity between the book feature vectors and the reader behavior vectors.

[0069] The adjustment engine module is used to perform adjustment recommendation and feedback control using an improved DICE algorithm to generate a recommended book list; the improved DICE algorithm introduces a conflict perception enhancement channel, a twin state gating structure, and a conflict resolution objective function;

[0070] The status update module is used to monitor the changes in the weight and density of conflict edges. If the cumulative decrease in the weight of a conflict edge within a set time window is greater than the second preset threshold and the conflict density is less than the third preset threshold, the conflict hotspot book mark of the corresponding book is canceled. A status twin feedback update mechanism is constructed to generate data on the trend of book resource conflict changes, adjustment response analysis data, and a visualization interface of book resource status.

[0071] In this implementation, the information modeling module collects information on book resources, reader behavior, reservation requests, and borrowing records to construct state entities and behavioral entities, mapping them to dynamic nodes in the virtual library space to form a state twin system, thereby enhancing the structured modeling capabilities of book resources and user behavior. The conflict analysis module constructs a conflict behavior graph and calculates the conflict density per unit time, effectively identifying high-frequency conflict book resources, marking conflict hotspot books in a timely manner, and enhancing the sensitivity of resource allocation. The similarity calculation module extracts feature vectors of books and readers and calculates similarity to generate a candidate book set, achieving accurate matching of resource substitution recommendations and optimizing the reader's borrowing experience. The adjustment engine module introduces an improved DICE algorithm, integrating a conflict perception enhancement channel, a twin state gating structure, and a conflict resolution objective function to improve the adaptability of recommendation output to conflict states and the accuracy of adjustment feedback. The state update module monitors the changes in conflict edge weights and conflict density, dynamically updates the conflict hotspot book markings, and constructs a state twin feedback update mechanism to generate data on book resource conflict change trends and visual analysis results, further enhancing the timeliness and intelligence of library resource regulation.

[0072] refer to Figure 2In this embodiment, the modules are interconnected through the following method:

[0073] S1. Collect book resource information, reader behavior information, reservation request information, and borrowing record information to construct state entities and behavior entities; the state entities include reservation queue, borrowing status, and remaining reservation count; the behavior entities include borrowing preference tags, historical borrowing records, and reservation failure rate; map the state entities and behavior entities to dynamic nodes in the virtual library space to form a state twin system.

[0074] S2. Construct a conflict behavior graph based on the state twin system, and map each book as a graph node; if multiple readers initiate borrowing requests for the same book within a preset time period and some requests are not fulfilled, then construct conflict edges in the conflict behavior graph. The weight of the conflict edges is calculated based on the conflict frequency, reservation failure rate and time overlap, and the time range of the conflict behavior and the behavioral characteristics of the participating readers are recorded.

[0075] S3. Calculate the conflict density per unit time for the map nodes; if the conflict density is greater than a first preset threshold, mark the corresponding book as a conflict hotspot book;

[0076] S4. Extract keywords, category tags, theme types, and borrowing behavior tags from the conflict hotspot books to generate book feature vectors; construct reader behavior vectors by combining readers' historical borrowing behavior records, and generate a candidate book set based on the similarity between the book feature vectors and the reader behavior vectors.

[0077] S5. An improved DICE algorithm is used to perform adjustment recommendation and feedback control to generate a recommended book list; the improved DICE algorithm introduces a conflict perception enhancement channel, a twin state gating structure, and a conflict resolution objective function;

[0078] S6. Monitor the changes in the weight and density of conflict edges; if the cumulative decrease in the weight of a conflict edge within a set time window is greater than the second preset threshold and the conflict density is less than the third preset threshold, then cancel the conflict hotspot book mark for the corresponding book.

[0079] S7. Construct a state twin feedback update mechanism to synchronously update the borrowing request information, reservation result information, conflict behavior graph structure change information and reader response behavior information to the state twin system, perform joint update operations on state entities, behavior entities and conflict behavior graphs, and generate book resource conflict change trend data, adjustment response analysis data and book resource status visualization interface.

[0080] In this embodiment, S1 specifically refers to:

[0081] Collect book resource information, including book identification information, book name information, catalog number information, storage location code information, classification code information, catalog quantity information, and available quantity information; collect reader behavior information, including reader identification information, access time information, borrowing operation identification information, reservation operation identification information, and return operation identification information; collect reservation request information, including reservation initiation time information, reserved book identification information, reservation request status information, and request response result information; collect borrowing record information, including borrowing time information, return time information, overdue status information, and default status information.

[0082] Construct a status entity, encode the reservation queue into a queue sequence data according to time order, mark the borrowing status as available and unavailable records, and record the remaining reservation count as a reservation count parameter; construct a behavior entity, label the borrowing preference tags according to the book category code, construct the historical borrowing records into a behavior trajectory sequence according to time order, and generate a failure rate parameter based on the ratio of the number of failed requests to the total number of requests.

[0083] The state entity and the behavior entity are respectively bound to the book identification information and the reader identification information, and mapped to the node data in the virtual library space. The node data includes node number information, node type identification information, node time attribute information and node state attribute information. Based on the node number information and time sequence information, node association data is generated to form a node structure data set of the state twin system.

[0084] In this embodiment, S2 specifically refers to:

[0085] Write book identification information into the node identification field to generate graph node data records; sort the borrowing request information within a preset time period by time to form request time series data; perform grouping operation on the borrowing request records corresponding to the same book identification information to generate a set of request records for the same book;

[0086] Perform time overlap interval comparison on the same book request record set, and filter the time interval intersection record set; perform request status matching operation on the time interval intersection record set, and extract reservation request status information and request response result information; if there are unmet status records in the request response result information, generate conflict behavior marker data;

[0087] Conflict edge structure data is constructed based on conflict behavior labeling data. The conflict edge structure data includes starting node identification information, target node identification information, and edge number information. Counting processing is performed on conflict frequency information, proportional calculation processing is performed on reservation failure rate information, and interval superposition processing is performed on time overlap information. Weighted synthesis processing is then performed to generate conflict edge weight data. The conflict edge weight data, along with conflict behavior occurrence time range information and participating reader identification information, are written into the conflict behavior graph data set.

[0088] In this embodiment, S3 specifically refers to:

[0089] This implementation reads the graph node data records from the conflict behavior graph, extracts node identification information and node-associated conflict edge data; sorts the associated conflict edge data of each graph node in ascending order by the timestamp field, constructing a time-series structured conflict edge record set; sets a preset statistical time interval, dividing the conflict edge record set into multiple time windows according to the timestamp field; performs edge count accumulation processing on the conflict edge set data within each time window, generating a statistical result of the number of conflicts occurring within the corresponding time window; performs a ratio operation between the statistical result of the number of conflicts occurring and the time interval length, with the conflict density calculated by dividing the number of conflict edges by the time interval length, generating a conflict density parameter corresponding to each graph node; performs an inter-node comparison operation on the conflict density parameters of the graph nodes, sorting them according to the size of the conflict density value, constructing a node conflict density sorting list; performs a threshold comparison operation between the conflict density data in the node conflict density sorting list and a first preset threshold, if the conflict density data is greater than the first preset threshold, writes the corresponding node identification information into a conflict hotspot marking table, forming a conflict hotspot book identification record set; this method can accurately identify book resources with frequent conflicts within a unit of time, improving the real-time performance and accuracy of book resource hotspot status judgment.

[0090] In this embodiment, S4 specifically refers to:

[0091] The system reads a set of conflict hotspot book identifier records and extracts basic book information fields for each record, including book title, classification code, subject type, and associated borrowing behavior records. It performs word segmentation on the book title information, extracts keyword data based on a defined dictionary resource, and records the keyword frequency order and position index in the title. It performs label mapping on the classification code information, converting the classification code into a unified format of classification label data and recording the classification hierarchy information. It performs unsupervised clustering and normalization on the subject type information, generating subject type label data and extracting the corresponding cluster center index value. Finally, it performs label mining on the associated borrowing behavior records, extracting reader behavior pattern fields, behavior time density fields, and operation type labels to generate borrowing behavior label data.

[0092] Keyword data, category tag data, topic type tag data, and borrowing behavior tag data are concatenated in a set order to construct a set of book feature descriptions; the set of book feature descriptions is vectorized, and a sparse coding and embedded feature filling strategy is used to generate book feature vectors, and zero-filling or nearest neighbor estimation is performed on missing fields to form a complete vector data structure.

[0093] Read the reader's historical borrowing records corresponding to the book identification records, extract the book identification information and borrowing time sequence information of the borrowed books, and arrange them in ascending order of time to construct a borrowing trajectory sequence; for each book record in the borrowing trajectory sequence, call the book basic information fields to extract keywords, category tags, theme types and borrowing behavior tags respectively; perform independent encoding on each field, and combine them in chronological order to form a historical borrowed book tag vector sequence;

[0094] Sequence modeling and feature fusion processing are performed on the historical borrowed book tag vector sequence. A temporal window weighting mechanism is used to extract behavioral stage features, and a time density function is combined to generate reader behavior vectors. The behavior vectors contain behavioral tendency distribution, reading frequency trend and topic preference principal components.

[0095] For each book feature vector and reader behavior vector, a similarity calculation is performed. The similarity calculation method is to divide the dot product of the two vectors by the product of their respective magnitudes. The generated similarity scores are sorted in descending order to construct a similarity ranking list. According to the screening rules of similarity scores greater than a preset similarity threshold in the ranking results, the corresponding book identification information is extracted to form a candidate book set. This method can realize intelligent recommendation of replacement books with high similarity of conflicting and popular books, thereby improving the accuracy of recommendations and reader satisfaction.

[0096] In this embodiment, S5 specifically refers to:

[0097] Read the candidate book set and the conflict hotspot book identifier record set, and construct a recommendation input vector for each candidate book and the corresponding conflict hotspot book. The recommendation input vector includes a book feature vector, a reader behavior vector, and a book conflict label vector. The book conflict label vector is processed by three-dimensional encoding based on the conflict density value, conflict edge weight, and conflict historical frequency.

[0098] Attention enhancement processing is applied to the book conflict label vector based on the recommendation input vector to generate a conflict-aware representation; the conflict-aware representation is then concatenated with the book feature vector and the reader behavior vector to construct a fusion representation tensor.

[0099] Gating and filtering operations are performed on the historical borrowing time sequence features in the reader behavior vector to obtain the state memory vector; the state memory vector is concatenated with the fusion representation tensor to form a set of candidate recommendation representations.

[0100] Based on the candidate recommendation representation set, a scoring and sorting operation is performed. Books with scores greater than the preset recommendation output threshold in the sorting results are used as the recommendation output set. The recommendation output set is compared with the original conflict hotspot book identification records. The conflict replacement success rate, reader acceptance feedback rate and recommendation response delay parameter are statistically analyzed, and the feedback control results are recorded.

[0101] By combining the feedback control results with the recommended output set, a list of recommended books is generated; the list of recommended books includes book identification information, a recommendation reason field, a reader acceptance status field, and a conflict mitigation index.

[0102] In this embodiment, the improved DICE algorithm introduces a conflict-aware enhancement channel, a twin-state gating structure, and a conflict resolution objective function, specifically as follows:

[0103] A conflict perception enhancement channel is constructed, and a convolutional attention fusion operation is performed on the conflict label vector of the input book. The convolution operation uses a multi-scale kernel to extract local variation features of conflict intensity, and the attention mechanism uses a dot-multiplication weighting method to obtain the correlation between conflict frequency and reader response. The conflict label vector includes a conflict density value encoding, a conflict edge strength vector, and a conflict history period index. The conflict density value encoding uses normalized integer encoding to represent the conflict frequency per unit time, the conflict edge strength vector is generated by superimposing the number of conflict occurrences and the time overlap ratio, and the conflict history period index represents the periodic characteristics and fluctuation amplitude of conflict behavior. The fusion result constitutes a conflict perception representation, which is input to a normalized mapping module. A conflict enhancement input vector with a standardized value range is generated using the max-min normalization method, which serves as the conflict representation input for subsequent adjustment paths. This processing method can introduce the dynamic perception capability of conflict features into the adjustment calculation, improving the recommendation engine's ability to identify book conflict characteristics and the accuracy of adjustment.

[0104] A twin-state gating structure is constructed, receiving a fusion vector of the reader's historical borrowing behavior sequence and book features. The behavior sequence and book vector are concatenated according to a unified dimensional mapping rule and fed into a batch structure divided by time slices, sequentially inputting into the gating unit. The gating unit includes a memory control gate, a conflict response gate, and a preference feedback gate. The memory control gate generates a decay weight based on the reciprocal of the borrowing time from the current time, and the decay function adopts an exponential weight reduction strategy to control the activation level of historical behavior. The conflict response gate determines whether to open the conflict feedback path based on whether the similarity value between the reader's historical conflict book list and the currently recommended book is greater than a set threshold, and records the conflict impact factor when it is active. The preference feedback gate receives the recommendation acceptance behavior identifier value and the feedback behavior vector, and updates the gating state using a gating selection mechanism. The fusion vector output by the gating structure includes a preference tendency reinforcement path, a conflict resolution path, and a behavior memory update path, which are used to jointly influence the final recommendation score. Through the above structure, the model's responsiveness to changes in historical behavior and its ability to perceive conflict states are improved.

[0105] A conflict resolution objective function is constructed, which jointly evaluates the output recommended book score, conflict substitution success rate, reader acceptance status, and historical conflict mitigation data. The recommended book score is the ranking score output by the recommendation engine. The conflict substitution success rate is defined as the ratio of the number of successfully borrowed recommended substitute books to the total number of recommendations. The reader acceptance status records whether the borrowing behavior has been completed. Historical conflict mitigation data is the decrease in conflict density within a specified time window. The conflict resolution objective function consists of a recommendation accuracy component, a conflict coverage component, and a feedback gain component. The recommendation accuracy component represents the degree of overlap between recommended books and actual reader borrowing behavior. The conflict coverage component represents the coverage ratio of the recommendation results in the set of conflict hotspot books. The feedback gain component represents the trend value of the decrease in conflict density after the recommendation is accepted. The three components are combined in a weighted linear form to form a joint loss function, and the optimization objective is to minimize the loss function value.

[0106] This method employs gradient descent to fit the conflict resolution objective function through multiple rounds, adjusting the recommendation ranking structure and recommendation score threshold parameters. The recommendation ranking structure includes the internal score ranking logic and priority arrangement of the candidate book set, while the recommendation score threshold parameter represents the minimum recommended output value requirement. Each training round records the adjustment loss results, calculates the difference in conflict density before and after, and constructs a conflict mitigation trend value. Finally, a recommended book list is output, along with the recommendation adjustment loss value and conflict mitigation trend value, serving as one of the input sources for the state update module. This method improves the substitution recommendation effect of conflict-ridden books and the adjustment and optimization capability of the feedback response, achieving the dual goals of book conflict resolution and improved recommendation accuracy.

[0107] In this embodiment, S6 specifically refers to:

[0108] Read the conflict behavior graph dataset and extract the starting node identifier, target node identifier, edge number, and conflict edge weight data from the conflict edge structure record. Write these fields into the conflict edge monitoring set to construct an edge-level monitoring view of the conflict behavior graph. Set the monitoring time window parameter to divide the time interval into continuous equal-length segments. Perform segmentation on the historical weight data of each conflict edge according to time slices to generate a conflict edge weight sequence vector. The dimension of the weight sequence vector is consistent with the number of time slices, representing the conflict edge intensity in each time slice.

[0109] For each conflict edge, the difference between the maximum and minimum values ​​of the weight sequence vector is calculated using the formula: the cumulative decrease is equal to the maximum value minus the minimum value, generating the corresponding cumulative decrease parameter. The cumulative decrease parameter is then compared with a second preset threshold to extract the conflict edge identification information where the decrease is greater than the second preset threshold. This information is written into the weight change judgment set to identify conflict edges that have experienced a significant decrease in intensity within the time window, which is used to dynamically determine the mitigation status of the conflict edges.

[0110] Read the conflict density data corresponding to the graph nodes in the conflict behavior graph. The conflict density is represented by the number of conflict edges associated with the book node per unit time. Compare the conflict density value of each graph node with the third preset threshold one by one, filter out the book node identification information with conflict density less than the third preset threshold, and write it into the density change judgment set to mark the book nodes whose conflict pressure has been relieved.

[0111] The weight change judgment set and the density change judgment set are intersected to extract the book node identification information that simultaneously meets the conditions that the decrease in the conflict edge weight is greater than the second preset threshold and the conflict density of the graph node is less than the third preset threshold, thus forming a conflict hotspot cancellation set; the book node identification information in the conflict hotspot cancellation set is removed from the conflict hotspot marking table, the marking table operation interface is called to perform the deletion operation, and at the same time the node identification information and timestamp are written into the update record table;

[0112] Construct data on changes in the graph structure, and generate corresponding record fields for each hotspot marker cancellation operation, including the cancelled node number, cancellation reason code, time window number, and operation identifier number.

[0113] In this embodiment, S7 specifically refers to:

[0114] Collect borrowing request information, extract request time information, requested book identifier information, request response status and reader identifier information, and construct a borrowing request record set; collect reservation result information, extract reservation initiation time, response time, result status and response feedback content, and construct a reservation feedback record set; collect conflict behavior graph structure change information, extract newly added conflict edge identifier information, removed conflict edge identifier information and graph node label status change information, and construct a structure change record set; collect reader response behavior information, extract recommendation acceptance identifier, actual borrowing behavior identifier, feedback time information and alternative selection identifier, and construct a response behavior record set.

[0115] Write the set of borrowing request records into the data structure of the status entity, and update the fields of reservation queue, borrowing status and remaining reservation count; synchronously write the set of reservation feedback records into the data structures of the status entity and behavior entity, and update the reservation status label, response latency statistics and failure rate parameters;

[0116] Perform node mapping matching operation on the set of structural change records, update the node state attributes and edge structure state attributes in the conflict behavior graph, and record the newly added edge data and removed edge data in the graph; perform reader identifier merging processing on the set of response behavior records, write the recommendation acceptance identifier and alternative selection identifier into the behavior entity structure, and update the behavior trajectory sequence and behavior preference label;

[0117] The system synchronously performs data reconstruction operations on the node set and edge set to generate updated state twin architecture data; performs statistical analysis based on the updated node data, conflict label data, and behavioral feedback data to generate book resource conflict change trend data; performs adjustment response ratio analysis based on the recommendation output set and feedback control records to generate adjustment response analysis data; reads the basic information of books and state entity fields to generate a book resource status visualization display structure and updates the book resource status visualization interface.

[0118] Example 1:

[0119] To verify the feasibility of this invention in practice, it was applied to a comprehensive university library with a collection of over 1.2 million books, an average daily borrowing volume of approximately 4,200 books, and a user base of about 21,000 full-time undergraduate and graduate students, as well as faculty and staff. The library has long suffered from problems such as concentrated conflicts in borrowing popular books, a high reservation failure rate, inefficient adjustment and recommendation processes, and delayed feedback on book resource status, which seriously affect the user borrowing experience and the efficiency of book resource circulation.

[0120] In the deployment process of this invention, the information modeling module first collects book resource information, reader borrowing behavior information, reader reservation request information, and historical borrowing record information, and constructs state entities and behavior entities. State entities include information such as book code, book category tag, borrowing status, and reservation status; behavior entities include reader identification, borrowing time sequence, reservation request time, and historical behavior trajectory. These entities are mapped to dynamic nodes within the virtual library space, forming a state twin system that is updated in real time.

[0121] Based on the conflict analysis module, a conflict behavior graph is constructed, mapping book resources to nodes and establishing conflict edges for borrowing requests for the same book initiated by multiple readers within a specific time period. The system calculates the conflict density of each node per unit time in real time. If the density exceeds a first preset threshold (defined as more than 5 requests per hour with a success rate of less than 50%), the book is marked as a conflict hotspot. Taking "Introduction to Artificial Intelligence" and "Introduction to Quantum Mechanics" as examples, the daily average conflict densities were as high as 7.1 and 6.4 respectively in the early stages of the system's launch, and both were included in the conflict hotspot marking table.

[0122] The similarity calculation module extracts keywords, category tags, topic types, and borrowing behavior tags from conflict-prone books to generate book feature vectors. It then constructs reader behavior vectors based on readers' historical borrowing records and generates a candidate book set based on vector similarity calculations. The adjustment engine module employs an improved DICE algorithm to perform recommendation and feedback adjustment, introducing a conflict perception enhancement channel, a twin state gating structure, and a conflict resolution objective function to improve recommendation accuracy and conflict replacement success rate.

[0123] Within three months of system operation, the system monitored the conflict edge weights and conflict density using the status update module, and cumulatively removed 38 conflict hotspot book tags, achieving twin feedback updates of book status. The system also output the book resource conflict trend and adjustment response status through the status visualization interface.

[0124] To compare the effects before and after the system was put into operation, the adjustment effects of the library's original adjustment system and the solution of this invention were collected in the scenario of conflict hotspot books, and the following data were compiled into Table 1.

[0125] Table 1. Comparison of Conflict Hotspot Book Adjustment Indicators in University Libraries

[0126] Indicator Item Traditional system This invention system Increase Average daily number of conflicting books 42.6 17.3 ↓59.4% Average book reservation failure rate 48.1% 21.7% ↓26.4% Conflict Alternative Book Recommendation Acceptance Rate 34.2% 71.5% ↑108.7% Average time to resolve conflicts over popular books 4.8 days 1.6 days ↓66.7%

[0127] Analysis of the data in Table 1 shows that the system of this invention has significant advantages in conflict book adjustment. The average daily number of conflict-ridden books decreased from 42.6 in the original system to 17.3, a reduction of 59.4%, significantly alleviating resource congestion; the failure rate of conflict book reservations decreased from 48.1% to 21.7%, reducing users' perception of borrowing failures; the acceptance rate of alternative book recommendations increased to 71.5%, more than double that of the original system, indicating that the improved DICE algorithm demonstrates strong behavioral understanding and preference adaptation capabilities in multi-factor recommendations; the average time to alleviate conflict-ridden books was shortened from 4.8 days to 1.6 days, accelerating book circulation efficiency, demonstrating that the conflict behavior map and state twin mechanism constructed by this system can dynamically reflect the actual conflict situation and achieve rapid adjustment.

[0128] Further data was collected on recommendation responses and feedback indicators on the status of book resources, resulting in Table 2, which compares the accuracy of recommendations with the balance of resource usage.

[0129] Table 2 Comparison of Recommendation Response and Resource Status Feedback

[0130] Indicator Item Traditional system This invention system Increase Recommended book click-through rate 41.6% 68.9% ↑65.6% Recommended book actual borrowing conversion rate 28.4% 63.2% ↑122.5% Number of times obscure books are recommended 391 times 1052 times ↑169.2% Book status feedback update delay time Average 8.2 hours Average 2.1 hours ↓74.4%

[0131] Table 2 reflects the advantages of the system of this invention in terms of recommendation accuracy and balanced use of book resources. The click-through rate and actual borrowing conversion rate of recommended books have both significantly improved, reaching 68.9% and 63.2% respectively, indicating that the similarity calculation module and the improved DICE algorithm are more accurate in understanding user needs and matching resources. The number of times less popular books are recommended has increased from 391 times in the original system to 1052 times, indicating that the system effectively breaks the dilemma of concentrated use of popular resources and achieves reasonable allocation and utilization of resources. The book status feedback update delay has been shortened from an average of 8.2 hours to 2.1 hours, indicating that the status update module and the status twin system have significantly enhanced their ability to perceive and respond to resource status, further improving the real-time adjustment and precise control level of the system.

[0132] This embodiment proposes an innovative solution to problems such as book borrowing conflicts, recommendation accuracy, and resource feedback lag. It has advantages such as high accuracy of state twin mapping, fast response to conflict adjustment, strong adaptability of recommendation behavior, and more balanced use of resources. It has good practical value and prospects for promotion in real-world environments.

[0133] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A comprehensive intelligent management and control system for university libraries based on digital twins, characterized in that: include: The information modeling module is used to collect information on book resources, reader behavior, reservation requests, and borrowing records, and to construct state entities and behavior entities. The state entities and behavior entities are mapped to dynamic nodes in the virtual library space, forming a state twin system; The conflict analysis module is used to construct a conflict behavior graph, mapping each book to a graph node; if multiple readers initiate borrowing requests for the same book within a preset time period and some requests are not fulfilled, then conflict edges are constructed in the conflict behavior graph. Calculate the collision density per unit time for the aforementioned map nodes; If the conflict density is greater than the first preset threshold, the corresponding book will be marked as a conflict hotspot book; The similarity calculation module is used to extract keywords, category tags, topic types, and borrowing behavior tags of the conflict hot books to generate book feature vectors; combine the readers' historical borrowing behavior records to construct reader behavior vectors; and generate a candidate book set based on the similarity between the book feature vectors and the reader behavior vectors. The adjustment engine module is used to perform adjustment recommendation and feedback control using an improved DICE algorithm to generate a recommended book list; the improved DICE algorithm introduces a conflict perception enhancement channel, a twin state gating structure, and a conflict resolution objective function; The status update module is used to monitor the changes in the weight and density of conflict edges. If the cumulative decrease in the weight of a conflict edge within a set time window is greater than the second preset threshold and the conflict density is less than the third preset threshold, the conflict hotspot book mark of the corresponding book is canceled. A status twin feedback update mechanism is constructed to generate data on the trend of book resource conflict changes, adjustment response analysis data, and a visualization interface of book resource status.

2. The comprehensive intelligent management and control system for university libraries based on digital twins according to claim 1, characterized in that, Inter-module communication is achieved through the following methods: S1. Collect book resource information, reader behavior information, reservation request information, and borrowing record information to construct state entities and behavior entities; the state entities include reservation queue, borrowing status, and remaining reservation count; the behavior entities include borrowing preference tags, historical borrowing records, and reservation failure rate; map the state entities and behavior entities to dynamic nodes in the virtual library space to form a state twin system. S2. Construct a conflict behavior graph based on the state twin system, and map each book as a graph node; if multiple readers initiate borrowing requests for the same book within a preset time period and some requests are not fulfilled, then construct conflict edges in the conflict behavior graph. The weight of the conflict edges is calculated based on the conflict frequency, reservation failure rate and time overlap, and the time range of the conflict behavior and the behavioral characteristics of the participating readers are recorded. S3. Calculate the collision density per unit time for the map nodes; If the conflict density is greater than the first preset threshold, the corresponding book will be marked as a conflict hotspot book; S4. Extract keywords, category tags, theme types, and borrowing behavior tags from the conflict hotspot books to generate book feature vectors; construct reader behavior vectors by combining readers' historical borrowing behavior records, and generate a candidate book set based on the similarity between the book feature vectors and the reader behavior vectors. S5. An improved DICE algorithm is used to perform adjustment recommendation and feedback control to generate a recommended book list; the improved DICE algorithm introduces a conflict perception enhancement channel, a twin state gating structure, and a conflict resolution objective function; S6. Monitor the changes in the weight and density of conflict edges; if the cumulative decrease in the weight of a conflict edge within a set time window is greater than the second preset threshold and the conflict density is less than the third preset threshold, then cancel the conflict hotspot book mark for the corresponding book. S7. Construct a state twin feedback update mechanism to synchronously update the borrowing request information, reservation result information, conflict behavior graph structure change information and reader response behavior information to the state twin system, perform joint update operations on state entities, behavior entities and conflict behavior graphs, and generate book resource conflict change trend data, adjustment response analysis data and book resource status visualization interface.

3. The comprehensive intelligent management and control system for university libraries based on digital twins according to claim 2, characterized in that, Specifically, S1 is: Collect book resource information, including book identification information, book name information, collection number information, storage location code information, classification code information, collection quantity information, and available quantity information; collect reader behavior information, including reader identification information, access time information, borrowing operation identification information, reservation operation identification information, and return operation identification information; collect reservation request information, including reservation initiation time information, reserved book identification information, reservation request status information, and request response result information. Collect borrowing record information, and obtain information on borrowing time, return time, overdue status, and default status. Construct a status entity, encode the reservation queue into a queue sequence data according to time order, mark the borrowing status as available and unavailable records, and record the remaining reservation count as a reservation count parameter; construct a behavior entity, label the borrowing preference tags according to the book category code, construct the historical borrowing records into a behavior trajectory sequence according to time order, and generate a failure rate parameter based on the ratio of the number of failed requests to the total number of requests. The state entity and the behavior entity are respectively bound to the book identification information and the reader identification information, and mapped to the node data in the virtual library space. The node data includes node number information, node type identification information, node time attribute information and node state attribute information. Based on the node number information and time sequence information, node association data is generated to form a node structure data set of the state twin system.

4. The comprehensive intelligent management and control system for university libraries based on digital twins according to claim 2, characterized in that, Specifically, S2 is: Write the book identification information into the node identification field to generate graph node data records; sort the borrowing request information within a preset time period by time to form request time series data. Perform a grouping operation on borrowing request records corresponding to the same book identification information to generate a set of request records for the same book; Perform time overlap interval comparison on the same book request record set, and filter the time interval intersection record set; perform request status matching operation on the time interval intersection record set, and extract reservation request status information and request response result information; If there are unmet status records in the request response result information, conflict behavior marker data is generated; Conflict edge structure data is constructed based on conflict behavior labeling data. The conflict edge structure data includes starting node identification information, target node identification information, and edge number information. The system performs counting processing on conflict frequency information, proportional calculation processing on appointment failure rate information, and interval overlay processing on time overlap information, and then performs weighted synthesis processing to generate weight data for conflict edges. The weight data of conflict edges, along with the time range information of conflict behavior and the identification information of participating readers, are written into the conflict behavior graph data set.

5. A comprehensive intelligent management and control system for university libraries based on digital twins as described in claim 2, characterized in that, Specifically, S3 is: Read the node data records in the conflict behavior graph, extract the node identification information and the associated conflict edge data; filter the conflict edge data within a preset statistical time interval to form a set of conflict edge data within a time window; perform a count accumulation process on the set of conflict edge data within the time window to generate a statistical result of the number of conflict occurrences. The conflict density is generated by comparing the number of conflict occurrences with the time interval. Perform a node-by-node comparison operation on the conflict density corresponding to the graph node to form a sorted list of node conflict density. The conflict density data of nodes in the conflict density sorting list is compared with a first preset threshold to determine the threshold. If the node conflict density data is greater than the first preset threshold, the node identification information is written into the conflict hotspot marking table to form a set of conflict hotspot book identification records.

6. A comprehensive intelligent management and control system for university libraries based on digital twins according to claim 2, characterized in that, Specifically, S4 is: Read the set of conflict hotspot book identifier records, and extract the basic book information fields for each book identifier record; The book name information in the basic book information field is processed by word segmentation to extract keyword data; the classification code information is mapped by label to extract classification label data; and the topic type information is clustered and normalized to extract topic type labels. Tag mining is performed on book-related borrowing behavior records to extract borrowing behavior tag data; Keyword data, category tag data, topic type tag data, and borrowing behavior tag data are concatenated to construct a set of book feature descriptions; the set of book feature descriptions is then vectorized to generate book feature vectors. Read the reader's historical borrowing records corresponding to the book identification records, and extract the book identification information and borrowing time sequence information of the borrowed books; perform feature extraction operations on the basic information of the borrowed books to obtain the keywords, category tags, theme types and borrowing behavior tags of the historically borrowed books; The tag information of historical borrowed books is processed by feature encoding, and a feature sequence of reader borrowing behavior trajectory is constructed in chronological order; vector fusion processing is performed on the behavior trajectory feature sequence to generate reader behavior vectors; The similarity between the book feature vector and the reader behavior vector is calculated to generate a similarity score. The similarity scores are then sorted, and the identification information of books with similarity scores greater than a preset similarity threshold is extracted to form a candidate book set.

7. A comprehensive intelligent management and control system for university libraries based on digital twins as described in claim 2, characterized in that, Specifically, S5 is: Read the candidate book set and the conflict hotspot book identifier record set, and construct a recommendation input vector for each candidate book and the corresponding conflict hotspot book. The recommendation input vector includes a book feature vector, a reader behavior vector, and a book conflict label vector. The book conflict label vector is processed by three-dimensional encoding based on the conflict density value, conflict edge weight, and conflict historical frequency. Attention enhancement processing is applied to the book conflict label vector based on the recommendation input vector to generate a conflict-aware representation; the conflict-aware representation is then concatenated with the book feature vector and the reader behavior vector to construct a fusion representation tensor. Gating and filtering operations are performed on the historical borrowing time sequence features in the reader behavior vector to obtain the state memory vector; the state memory vector is concatenated with the fusion representation tensor to form a set of candidate recommendation representations. Based on the candidate recommendation representation set, a scoring and sorting operation is performed. Books with scores greater than the preset recommendation output threshold in the sorting results are used as the recommendation output set. The recommendation output set is compared with the original conflict hotspot book identification records. The conflict replacement success rate, reader acceptance feedback rate and recommendation response delay parameter are statistically analyzed, and the feedback control results are recorded. By combining the feedback control results with the recommended output set, a list of recommended books is generated. The recommended book list includes book identification information, a recommendation reason field, a reader acceptance status field, and a conflict mitigation index.

8. A comprehensive intelligent management and control system for university libraries based on digital twins according to claim 2, characterized in that, The improved DICE algorithm introduces a conflict-aware enhancement channel, a twin-state gating structure, and a conflict resolution objective function, specifically: A conflict-aware enhancement channel is constructed by performing a convolutional attention fusion operation on the conflict label vector of the input book to generate a conflict-aware representation. The conflict label vector includes a conflict density value encoding, a conflict edge strength vector, and a conflict history cycle index. A normalization mapping operation is performed on the conflict-aware representation as the input path for conflict enhancement. A twin state gating structure is constructed to receive the fusion vector of readers' historical borrowing behavior sequences and book features, and send them to the gating unit in batches according to time slices. The gating unit is equipped with a memory control gate, a conflict response gate, and a preference feedback gate. The control gate controls the activation level of borrowing behavior based on the borrowing time decay weight. The response gate activates the conflict feedback path based on historical conflict records. The feedback gate uses the recommended acceptance behavior as the basis for state update. A conflict resolution objective function is constructed, and the output recommended book scores, conflict replacement success rate, reader acceptance status and historical conflict mitigation data are jointly evaluated. The conflict resolution objective function includes recommendation accuracy component, conflict coverage component and feedback gain component. The gradient descent method is used to fit the conflict resolution objective function in multiple rounds, optimize the recommendation ranking structure and recommendation score threshold parameters, output the final recommended book list, and record the recommendation adjustment loss results and the book conflict mitigation trend value.

9. A comprehensive intelligent management and control system for university libraries based on digital twins according to claim 2, characterized in that, Specifically, S6 is: Read the conflict behavior graph dataset, extract the starting node identifier, target node identifier, edge number, and conflict edge weight data from the conflict edge structure record, and establish a conflict edge monitoring set; set the monitoring time window parameters, perform time series segmentation on the conflict edge weight data, and construct a weight sequence vector; Perform a maximum-minimum difference operation on the weight sequence vector of each conflicting edge to generate a cumulative decrease parameter; Perform a second preset threshold comparison operation on the cumulative decrease parameter, filter out conflict edge identification information where the decrease is greater than the second preset threshold, and generate a weight change judgment set; Read the conflict density data of the corresponding graph nodes, perform a third preset threshold comparison operation on each graph node, filter out the book node identification information with conflict density less than the third preset threshold, and generate a density change judgment set. The intersection of the weight change judgment set and the density change judgment set is performed to extract the book node identification information that simultaneously satisfies the decrease magnitude greater than the second preset threshold and the conflict density less than the third preset threshold, thus forming the conflict hotspot cancellation set. Remove the book node identifier information from the conflict hotspot cancellation set from the conflict hotspot marking table, update the conflict hotspot book identifier record set, and record the graph structure change status data.

10. A comprehensive intelligent management and control system for university libraries based on digital twins according to claim 2, characterized in that, Specifically, S7 is: Collect borrowing request information, reservation result information, conflict behavior graph structure change information, and reader response behavior information; update the borrowing request information and reservation result information to the status entity and behavior entity, and adjust the reservation queue, borrowing status, remaining reservation count, and reservation failure rate parameters. Update the conflict behavior graph structure change information to the graph node data and conflict edge data, and record the addition and removal status of conflict edges; write the reader response behavior information into the behavior entity, and update the borrowing behavior trajectory and borrowing preference tags. Perform joint update operations on node data, edge data, and behavioral data to generate data on the changing trends of book resource conflicts, adjustment response analysis data, and a visualization interface of book resource status.

Images