A multi-source book data intelligent analysis method, system, terminal and storage medium

Abstract

The application discloses a kind of multi-source book data intelligent analysis method, system, terminal and storage medium, the method includes: obtaining multi-source book data, and multi-source book data is preprocessed, and target multi-source book data is obtained;Target multi-source book data is carried out data relationship integration processing, and target knowledge graph is obtained;According to target knowledge graph, user portrait generation processing and book borrowing trend prediction processing are carried out, and user portrait result and book borrowing trend result are obtained;The target knowledge graph, the user portrait result and the book borrowing trend result are visualized and exhibited.The application is preprocessed and data relationship integration processing to multi-source book data, and target knowledge graph is obtained, and user portrait generation processing and book borrowing trend prediction processing are carried out by target knowledge graph, can realize the accurate analysis to user behavior and book borrowing trend, and then realize the optimization allocation to book resources.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to an intelligent analysis method, system, terminal, and computer-readable storage medium for multi-source book data. Background Technology

[0002] Libraries are institutions that collect, organize, and preserve books and materials for people to read and reference. With the development of society, people's demand for reading is increasing, and therefore, libraries have become the source of borrowing data for more and more people.

[0003] However, existing integrated library management systems often face the problem of poor integration of multi-source data from various departments when dealing with collaborative work among multiple departments. This makes it impossible to analyze user behavior and book borrowing trends from multiple sources of book data, resulting in difficulties in allocating book resources and failing to meet user needs.

[0004] Therefore, existing technologies still need to be improved and developed. Summary of the Invention

[0005] The main objective of this invention is to provide an intelligent analysis method, system, terminal, and computer-readable storage medium for multi-source book data. This invention aims to solve the problem that existing library integrated management systems often face difficulties in integrating multi-source data from various departments when handling collaborative work among multiple departments. This results in the inability to analyze user behavior and book borrowing trends from multi-source book data, leading to difficulties in allocating book resources and failing to meet user needs.

[0006] To achieve the above objectives, the present invention provides an intelligent analysis method for multi-source book data, the method comprising the following steps:

[0007] Acquire multi-source book data and preprocess the multi-source book data to obtain target multi-source book data;

[0008] The target multi-source book data is processed by data relationship integration to obtain the target knowledge graph;

[0009] Based on the target knowledge graph, user profile generation and book borrowing trend prediction are performed to obtain user profile results and book borrowing trend results.

[0010] The target knowledge graph, the user profile results, and the book borrowing trend results are visualized.

[0011] Optionally, the intelligent analysis method for multi-source book data, wherein acquiring multi-source book data and preprocessing the multi-source book data to obtain target multi-source book data specifically includes:

[0012] A distributed log collection tool is used to acquire multi-source book data, which includes structured data, unstructured data, real-time data streams, and external data sources.

[0013] The multi-source book data is standardized using metadata standards to obtain standardized multi-source book data.

[0014] The standardized multi-source book data is processed using a data mapping tool to obtain mapped multi-source book data.

[0015] The mapped multi-source book data is cleaned to obtain the target multi-source book data.

[0016] Optionally, in the intelligent analysis method for multi-source book data, the data cleaning process includes data deduplication, missing value filling, formatting, and outlier handling.

[0017] The step of cleaning the mapped multi-source book data to obtain the target multi-source book data specifically includes:

[0018] Obtain the data identifier in the mapped multi-source book data, determine the duplicate data in the mapped multi-source book data based on the data identifier, and perform the data deduplication process on the duplicate data to obtain the first multi-source book data;

[0019] The missing values ​​in the first multi-source book data are obtained, and the missing values ​​are filled using a preset filling method to obtain the second multi-source book data. The preset filling method includes interpolation, mean filling and missing value deletion.

[0020] Obtain the date format corresponding to the second multi-source book data, and perform the formatting process on the second multi-source book data according to the date format to obtain the third multi-source book data;

[0021] The outlier detection algorithm is used to identify outliers in the third multi-source book data to obtain target outlier data. The outlier data is then processed to obtain target multi-source book data.

[0022] Optionally, the intelligent analysis method for multi-source book data, wherein the step of integrating data relationships in the target multi-source book data to obtain a target knowledge graph specifically includes:

[0023] A centralized data model is constructed, and the target multi-source book data is processed by the centralized data model to perform entity relationship structuring to obtain structured multi-source book data;

[0024] The core entities in the structured multi-source book data are identified, and the core entities are subjected to identifier assignment and fuzzy matching to obtain associated multi-source book data.

[0025] Entity relationships are extracted from the associated multi-source book data using natural language processing and semantic techniques to obtain entity relationship extraction results, and a target knowledge graph is constructed based on the entity relationship extraction results.

[0026] Optionally, the intelligent analysis method for multi-source book data, wherein the step of generating user profiles and predicting book borrowing trends based on the target knowledge graph to obtain user profile results and book borrowing trend results specifically includes:

[0027] The target knowledge graph is subjected to feature extraction and cluster analysis to obtain feature extraction results and user behavior patterns. Based on the feature extraction results and user behavior patterns, user profile generation is performed to obtain user profile results.

[0028] Obtain time-series data from the target knowledge graph and aggregate the time-series data to obtain aggregated time-series data;

[0029] A preset deep learning model is determined, and the aggregated time series data is input into the preset deep learning model to output the book borrowing trend results.

[0030] Optionally, the intelligent analysis method for multi-source book data, wherein the step of performing feature extraction and cluster analysis on the target knowledge graph to obtain feature extraction results and user behavior patterns, and then performing user profile generation based on the feature extraction results and user behavior patterns to obtain user profile results, specifically includes:

[0031] The target knowledge graph is subjected to feature extraction processing to obtain feature extraction results, wherein the feature extraction results include basic features, behavioral features and semantic features;

[0032] The feature extraction results are subjected to cluster analysis using a clustering analysis algorithm to obtain user behavior patterns;

[0033] A pre-trained model is determined, and the pre-trained model is trained and fine-tuned based on the target knowledge graph and the user behavior pattern to obtain an end-to-end profile model.

[0034] Obtain user reading data corresponding to the target user, input the user reading data into the end-to-end profiling model, and output the user profiling result.

[0035] Optionally, the intelligent analysis method for multi-source book data, wherein the step of generating user profiles and predicting book borrowing trends based on the target knowledge graph to obtain user profile results and book borrowing trend results, further includes:

[0036] Based on the user profile results, the corresponding recommended book data and predictive decision direction are determined, and the recommended book data and predictive decision direction are pushed to the corresponding target user;

[0037] Determine the current book resource allocation plan, and optimize the current book resource allocation plan based on the book borrowing trend results to obtain the target book resource allocation result.

[0038] Furthermore, to achieve the above objectives, the present invention also provides an intelligent analysis system for multi-source book data, wherein the intelligent analysis system for multi-source book data includes:

[0039] The data preprocessing module is used to acquire multi-source book data and preprocess the multi-source book data to obtain target multi-source book data.

[0040] The data relationship integration module is used to perform data relationship integration processing on the target multi-source book data to obtain the target knowledge graph;

[0041] The profile generation and trend prediction module is used to generate user profiles and predict book borrowing trends based on the target knowledge graph, so as to obtain user profile results and book borrowing trend results.

[0042] The visualization module is used to visualize the target knowledge graph, the user profile results, and the book borrowing trend results.

[0043] This invention involves acquiring multi-source book data and preprocessing it to obtain target multi-source book data. Data relationship integration processing is then performed on the target multi-source book data to obtain a target knowledge graph. User profile generation and book borrowing trend prediction are then performed based on the target knowledge graph to obtain user profile results and book borrowing trend results. Finally, the target knowledge graph, user profile results, and book borrowing trend results are visualized. This invention, by preprocessing and integrating multi-source book data to obtain a target knowledge graph, and then using this target knowledge graph for user profile generation and book borrowing trend prediction, enables precise analysis of user behavior and book borrowing trends. Based on the generated user profile results and book borrowing trend results, optimized allocation of book resources can be achieved, effectively improving the availability of book resources and optimizing the user experience. Attached Figure Description

[0044] Figure 1 This is a flowchart of a preferred embodiment of the intelligent analysis method for multi-source book data of the present invention;

[0045] Figure 2 This is a schematic diagram of the overall structure of a preferred embodiment of the intelligent analysis method for multi-source book data of the present invention;

[0046] Figure 3 This is a schematic diagram of the model architecture of a preferred embodiment of the intelligent analysis method for multi-source book data of the present invention;

[0047] Figure 4 This is a structural diagram of a preferred embodiment of the intelligent analysis system for multi-source book data of the present invention;

[0048] Figure 5 This is a structural diagram of a preferred embodiment of the terminal of the present invention. Detailed Implementation

[0049] To make the objectives, technical solutions, and advantages of this invention clearer and more explicit, 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 of the invention and are not intended to limit the invention.

[0050] Currently, existing technologies such as Integrated Library Systems (ILS), Radio-frequency Identification (RFID), and Digital Resource Management Systems (DRMS) used in traditional libraries suffer from the following main drawbacks and challenges: 1. Data silos: The lack of an effective data interaction mechanism between ILS and DRMS ​​creates data silos. 2. Poor system compatibility: Different systems may employ different technical architectures (e.g., monolithic architecture, microservice architecture) and databases (e.g., MySQL, MongoDB), leading to poor inter-system compatibility and difficulties in data exchange. 3. Lack of data analysis and visualization tools: A lack of powerful data analysis tools (e.g., Apache Spark, Tableau) for in-depth mining and multi-dimensional analysis of integrated data. 4. Insufficient application of visualization tools (e.g., Power BI, D3.js), resulting in the inability to generate intuitive analysis reports or visualized knowledge graphs.

[0051] To address the aforementioned issues, this invention proposes an intelligent analysis method for multi-source book data. This invention employs advanced database technology, supporting multiple databases such as MySQL, PostgreSQL, and MongoDB, ensuring efficient and scalable data storage. The data acquisition module, through flexible interface configuration, ensures real-time data import and accurate recording. The system's statistical analysis capabilities utilize big data technology to conduct in-depth analysis of borrowing, reader behavior, and resource usage. Chart display employs visualization tools such as ECharts and D3.js, providing rich reports and dynamic charts.

[0052] This invention constructs an intelligent library data analysis system that integrates multi-source data and utilizes machine learning, big data analytics, and artificial intelligence technologies to deeply mine multi-dimensional data such as reader behavior, borrowing trends, and resource utilization. It provides accurate analysis results and personalized services, while supporting data-driven decision optimization through real-time monitoring and visualization. This effectively improves library management efficiency, service quality, and resource utilization, meets reader needs, and supports scientific research and teaching development.

[0053] The intelligent analysis method for multi-source book data described in the preferred embodiment of the present invention, such as... Figure 1 As shown, the intelligent analysis method for multi-source book data includes the following steps:

[0054] Step S10: Obtain multi-source book data and preprocess the multi-source book data to obtain target multi-source book data.

[0055] The multi-source book data in this invention includes structured data, unstructured data, real-time data streams, and external data sources, originating from different departments and systems. This invention collects these different formats of data through data integration. The multi-source book data obtained after integration can also be called heterogeneous data sources or multimodal data sources.

[0056] The existing library system data suffers from the following problems: 1. Difficulty in data integration: When handling collaborative work among multiple departments such as resource development, circulation services, subject services, and special collections development, the existing Library Integrated Management System (ILS) often faces problems with data integration, leading to data silos and preventing the formation of complete knowledge graphs or the generation of high-quality analytical reports. Specific technical challenges and problems include: a. Inconsistent data formats across heterogeneous systems: Resource development, circulation services, subject services, and special collections development may use different data standards and formats (such as MARC (Machine-Readable Cataloging), Dublin Core, MODS (Metadata Object Description Schema), etc.), making direct data integration difficult. b. Lack of a unified metadata standard for data exchange between the Library Integrated Management System (ILS) and the Digital Resource Management System (DRMS), resulting in difficulties in data mapping and conversion. c. Incompatible or limited API interfaces: API interfaces of different systems (such as RESTful API, SOAP) may use different protocols or data formats (such as JSON, XML), lacking a unified interface specification. Some systems have limited API functionality, unable to support complex data queries or batch operations, resulting in low data synchronization efficiency. d. Insufficient data synchronization and real-time performance: Data synchronization between the book integrated management system and the digital resource management system usually relies on scheduled batch updates rather than real-time synchronization, leading to data delays and inconsistencies. The lack of a real-time data stream processing mechanism based on message queues (such as Kafka, RabbitMQ) makes it impossible to achieve real-time data updates across systems. 2. Insufficient data analysis accuracy: Existing analysis methods often lack depth and detail, making it difficult to accurately analyze multi-dimensional data such as reader behavior and borrowing trends, limiting decision support functions. Specific technical challenges and problems include: a. Insufficient data collection and cleaning: Data collection tools (such as Flume, Logstash) fail to comprehensively cover reader behavior data (such as borrowing records, search history, dwell time, etc.), resulting in incomplete data sources. Insufficient application of data cleaning techniques (such as Python-based Pandas and OpenRefine) fails to effectively handle noise, missing values, and outliers in the data, affecting the accuracy of subsequent analysis. b) Limited multi-dimensional data analysis capabilities: A lack of ability to perform correlation analysis on multi-dimensional data (such as time, space, user attributes, and resource type) makes it difficult to reveal complex borrowing trends and reader behavior patterns. Existing analysis tools (such as Excel and SPSS) are inadequate in handling large-scale, high-dimensional data and cannot support complex analytical tasks.c. Insufficient application of machine learning models: There is a lack of in-depth application of machine learning algorithms (such as clustering analysis, classification algorithms, and regression analysis), making it impossible to uncover potential patterns from massive amounts of data. The application of deep learning technologies (such as LSTM, Long Short-Term Memory networks, Transformer, and self-attention neural networks) in time series analysis (such as borrowing trend prediction) is insufficient, resulting in low prediction accuracy. 3. Poor visualization effects: Although existing visualization tools can provide basic statistical displays, their interactivity and dynamism are insufficient, making it difficult to meet users' personalized needs. Specific technical challenges and problems include: a. Limited functionality of visualization tools: Existing visualization tools (such as Tableau and Power BI) mainly support static charts (such as bar charts, pie charts, and line charts), lacking support for complex data (such as network diagrams, heat maps, and geospatial data). They lack the ability to visualize dynamic data (such as real-time borrowing data and user behavior stream data), failing to reflect data changes in real time. b. Insufficient interactivity: Existing visualization tools have limited interactive functions, preventing users from exploring data in depth through drag-and-drop, zoom, and filtering operations. The lack of web-based interactive visualization frameworks (such as D3.js and ECharts) makes it impossible to provide rich user interaction experiences (such as dynamic filtering and linked analysis).

[0057] This invention integrates and standardizes data from various departments (such as the Resource Construction Department, Lending Service Department, Subject Service Department, and Special Collections Construction Department) and systems (such as the Book Integration Management System and Digital Resource Management System) through a unified platform, eliminating data silos and enabling information sharing and interaction. Specific technical implementations include: 1. Unified Data Standards and Metadata Management: Adopting internationally recognized metadata standards (such as MARC, Dublin Core, and MODS) to standardize data from various departments, ensuring data format consistency. Using metadata management tools (such as OpenRefine and CatMDEdit) to clean, transform, and map metadata, resolving data heterogeneity issues. 2. Data Warehouse and ETL Technology: Constructing a unified data warehouse (such as Hadoop, Snowflake, and Amazon Redshift) to centrally store data from various departments and systems. Using ETL (Extract, Transform, Load) tools (such as Apache NiFi, Talend, and Informatica) to extract, transform, and load data, ensuring efficient data integration and synchronization. 3. API Interface and Microservice Architecture: Provide a unified RESTful API or GraphQL interface to support data interaction between departments and systems. Employ a microservice architecture (such as Spring Cloud or Kubernetes) to achieve modularity and loose coupling of the system, facilitating data sharing and functional expansion.

[0058] The system processing flow of this invention is as follows: 1. Automatic data cleaning from multiple data sources: Data source integration includes web crawling, importing text and table files, extracting from multiple database interfaces, and integrating with third-party systems. 2. Data hierarchical processing and relationship integration: Data is hierarchically classified according to attributes, and data relationships are generated according to hierarchical inclusion relationships and left-right correlation relationships. 3. Intelligent generation of digital model profiles: Profiles are automatically generated for different dimensions such as books, readers, and classrooms. The profiles clearly show the reading and management directions for each category. Further intelligent analysis of the profile content can be performed to provide suitable decision-making directions for different profiles. 4. Progressive digital cockpit presentation: The presentation progresses from coarse to fine management granularity. Each granularity has manual scheduling and processing capabilities, facilitating task assignment and book management for administrators. The main page mainly features analysis reports and data analysis statistics for each dimension, while progressive pages display drill-down information in the form of relationship graphs (where drill-down information refers to expanding the data from the current data to the next level for more detailed analysis).

[0059] Specifically, a distributed log collection tool is used to acquire multi-source book data, which includes structured data, unstructured data, real-time data streams, and external data sources; metadata standards are used to standardize the multi-source book data to obtain standardized multi-source book data; and a data mapping tool is used to map the standardized multi-source book data to obtain mapped multi-source book data.

[0060] The first step is to deploy the environment (database deployment), which specifically includes: 1. Selecting a database suitable for the project requirements, such as MySQL, Radius, PostgreSQL, MongoDB, etc. 2. Implementing database installation and configuration: configuring optimal parameters to ensure high performance and security. 3. Implementing data backup and driver processing: enhancing the system's fault tolerance through top-down maintenance or clustered driving.

[0061] like Figure 2 As shown, the development environment is set up as follows: 1. Install development tools: IDE (Integrated Development Environment, a program specifically for software development) such as VS Code and WebStorm, and configure relevant plugins. 2. Implement version control and collaboration: Use Git and resource management platforms such as GitHub or GitLab to configure a complete development and testing environment to ensure thorough testing and automated deployment of the application. By adopting machine learning and big data analytics technologies, the accuracy and depth of analysis of borrowing and reader behavior can be significantly improved, providing more valuable support for decision-making. Specific technical implementations include: a. Data collection and preprocessing: Use distributed log collection tools (such as Flume and Logstash) to comprehensively collect reader behavior data (such as borrowing records, search history, page click flow, etc.). b. Data source access functions include: Data source identification and classification: Structured data (accessing relational databases (such as MySQL and PostgreSQL) through JDBC / ODBC interfaces to read structured data such as borrowing records and reader information) including borrowing records (MySQL / Oracle), reader information (LDAP / identity system), and resource metadata (MARC format). Semi-structured data: log files (access logs, retrieval logs, JSON / CSV format), open API data (such as third-party database interfaces); unstructured data: unstructured data accessed using APIs or web crawling techniques (such as reader comments, document abstracts, web page content, and audio / video resources); real-time data streams: real-time data accessed through message queues (such as Kafka, RabbitMQ) (such as borrowing operations, retrieval logs); external data sources: external data accessed through open APIs (such as third-party databases, research platforms) to expand the dimensions of analysis.

[0062] like Figure 2 As shown, the choice of data acquisition technologies includes batch data acquisition, real-time data stream acquisition, and API integration. 1. Batch Data Acquisition: ETL tools (Apache NiFi, Talend): periodically extract borrowing records from relational databases (such as MySQL). File Import: process CSV / Excel files using scripts (Python Pandas) or tools (Sqoop). 2. Real-time Data Stream Acquisition: Message Queues (Apache Kafka, RabbitMQ): capture real-time events (such as borrowing operations, retrieval requests). Stream Processing Frameworks (Apache Flink, Spark Streaming): parse log stream data in real time (such as Nginx access logs). 3. API Integration: RESTful API / GraphQL: connect to external systems (such as e-book platforms, research databases) to obtain metadata. Web Scraping Technologies (Scrapy, Selenium): crawl publicly available data (such as academic paper citation information).

[0063] The choice of data access technologies includes ETL tools, API interfaces, and file import. ETL tools: Using ETL tools (such as Apache NiFi and Talend) to extract, transform, and load data, supporting the integration of multi-source data. API interfaces: Providing RESTful APIs or GraphQL interfaces to support data interaction between external systems and the library system. File import: Supporting batch import of file formats such as CSV, Excel, and JSON, facilitating the access of historical data.

[0064] Data standardization and mapping includes metadata management and data mapping. Metadata management: Standardizing multi-source data using metadata standards (such as MARC, Dublin Core). Data mapping: Mapping fields from different data sources to the target model using data mapping tools (such as OpenRefine).

[0065] Real-time data access includes stream processing frameworks and webhooks and message queues. Stream processing frameworks use stream processing technologies (such as Apache Flink and Spark Streaming) to access and process data streams in real time. Webhooks and message queues capture system events (such as borrowing and returning items) in real time through webhooks or message queues (such as Kafka).

[0066] The process involves: acquiring data identifiers from the mapped multi-source book data; identifying duplicate data in the mapped multi-source book data based on the data identifiers; performing deduplication on the duplicate data to obtain first multi-source book data; acquiring missing values ​​in the first multi-source book data; performing missing value filling on the missing values ​​using a preset filling method to obtain second multi-source book data, wherein the preset filling method includes interpolation, mean filling, and missing value deletion; acquiring the date format corresponding to the second multi-source book data; performing formatting on the second multi-source book data according to the date format to obtain third multi-source book data; and identifying outliers in the third multi-source book data using an outlier detection algorithm to obtain target outlier data, and performing outlier processing on the target outlier data to obtain target multi-source book data.

[0067] This invention employs data cleaning tools (such as Pandas and OpenRefine) to deduplicate, impute missing values, and handle outliers to ensure data quality. The data cleaning functions are as follows: 1. The data cleaning process includes: Data probing: Using data probing tools (such as Pandas Profiling) to analyze data distribution, missing values, outliers, etc. Data deduplication: Removing duplicate records using unique identifiers (such as ISBN, reader ID). Missing value handling: Handling incomplete data using methods such as interpolation, mean imputation, or deletion of missing values. Outlier handling: Identifying and handling outliers using statistical methods (such as Z-score, IQR) or machine learning algorithms (such as Isolation Forest). Format standardization: Standardizing the format of fields such as date, time, and text (such as YYYY-MM-DD, UTF-8 encoding). Examples of deduplication and completion in data cleaning are as follows: Using Pandas or OpenRefine to process duplicate borrowing records and fill in missing fields (such as reader department information); Format standardization: Standardizing different date formats (such as "2023-10-01" and "01 / 10 / 2023") to ISO 8601. Outlier handling: Detecting abnormal borrowing durations (such as borrowing records older than one year) using Z-score or IQR methods. 2. Data cleaning techniques include: Rule engine: Automatically cleaning data based on predefined rules (such as borrowing date cannot be later than return date). Machine learning: Using clustering and classification algorithms to identify noise and anomalies in the data. Natural language processing (NLP): Processing text data (such as reader comments and document abstracts) through word segmentation, stop word removal, and stemming. The application and effects of model algorithm cleaning techniques in the library intelligent analysis system are as follows: 1. Purpose: Cleaning and preprocessing raw data (such as borrowing records and reader behavior data) to remove noise, missing values, and outliers. Standardized data formats ensure data quality and provide reliable input for subsequent analysis. 2. Results: Improves the accuracy and reliability of data analysis, optimizes the training effect of machine learning models, enhances prediction accuracy (such as borrowing trends and reader preferences), reduces the risk of erroneous decisions, and provides higher-quality data support for library management and service optimization.

[0068] In summary, this invention discloses a cross-modal model cleaning algorithm, the core applications of which include:

[0069] 1. Multi-source data fusion and cleaning: Used to uniformly process heterogeneous data such as text (e.g., user comments), images (e.g., book pictures), and time-series data (e.g., logs), eliminating noise and conflicts (e.g., discrepancies between book descriptions and images); also used to establish semantic associations through cross-modal alignment (e.g., through the CLIP model), ensuring data consistency (e.g., matching user comment sentiment with image content).

[0070] 2. Intelligent Quality Repair: Used for automatic error correction, such as correcting typos in text using BERT / Transformer (improving text accuracy) and repairing blurry images using GAN (improving PSNR). Also used for anomaly detection: identifying fake comments or stolen images through multimodal joint modeling (e.g., text + image). c. Feature Enhancement: Generating unified feature vectors through cross-modal representation learning (e.g., ViLBERT) to improve the performance of downstream tasks.

[0071] The actual effects include:

[0072] 1. Improved data quality: Data noise was significantly reduced after cleaning, and the completeness rate of key fields was significantly improved. 2. Improved accuracy in identifying multimodal conflicts (such as text-image mismatch). 3. Optimized model performance: Cross-modal pre-training effectively improved the average F1 score for NLP / CV tasks; in small sample scenarios, data augmentation was superior to single-modal methods. 4. Improved image-text matching accuracy after cleaning; improved accuracy in predicting user profiles by integrating text, images, and behaviors.

[0073] Step S20: Perform data relationship integration processing on the target multi-source book data to obtain the target knowledge graph.

[0074] Specifically, a centralized data model is constructed, and the target multi-source book data is processed through entity relationship structuring using the centralized data model to obtain structured multi-source book data; the core entities in the structured multi-source book data are identified, and the core entities are processed through identifier allocation and fuzzy matching to obtain associated multi-source book data; entity relationships are extracted from the associated multi-source book data using natural language processing and semantic techniques to obtain entity relationship extraction results, and a target knowledge graph is constructed based on the entity relationship extraction results.

[0075] Data relationship integration function: 1. Data standardization and unified modeling includes: a. Metadata management: Adopting international standards (such as MARC, Dublin Core) to unify the metadata definition of different data sources and solve the problem of inconsistent field naming and format. b. Unified data model: Designing a centralized data model (such as star schema, graph model) to structure the entity relationships such as readers, resources, borrowing records, and subject classification.

[0076] 2. Entity resolution and association include: a) Unique identifiers: Assigning unique IDs (such as reader ID, ISBN) to core entities such as readers and resources to achieve entity matching across data sources. b) Fuzzy matching technology: Using string similarity algorithms (such as Levenshtein distance) or machine learning models (such as Siamese networks) to associate similar but not completely identical records (such as author names with different spellings).

[0077] 3. Graph databases and knowledge graphs include: a. Graph databases (such as Neo4j, TigerGraph): Store complex relationships between entities (such as "Reader A borrows resource B → related subject C"), supporting multi-hop queries and semantic reasoning. b. Knowledge graph construction: Extract entity relationships through natural language processing (NLP) and semantic technologies (RDF, OWL) to form a unified knowledge network across departments.

[0078] 4. Relationship integration includes: real-time relationship updates, stream processing frameworks (such as Apache Kafka, Flink): real-time capture of borrowing, returning, and other events, and dynamic updates of entity relationships (such as automatically expanding nodes after a new borrowing record is added). Incremental ETL: only synchronizing newly added or changed data, reducing the overhead of full integration.

[0079] This invention employs multi-database federation technologies based on distributed queries, including MySQL, Radis, No4J, and ClickHouse. Its core applications include:

[0080] 1. Heterogeneous Data Integration: Seamless joint queries are achieved between MySQL (transactional data), Redis (cached / real-time data), Neo4j (relational networks), and ClickHouse (analytic data) using federated query technologies (such as Presto and ClickHouse table engines), breaking down data silos. Simultaneously, cross-database JOIN, subqueries, and aggregation operations are supported, such as real-time correlation with user behavior (ClickHouse).

[0081] 2. Performance optimizations include: a) Accelerated hot data: Redis caches frequently accessed data, reducing query latency from milliseconds to microseconds. b) Speeded up complex analysis: ClickHouse columnar storage enables second-level aggregation of hundreds of millions of data points, faster than traditional MySQL analysis. c) Efficient relational queries: Neo4j handles multi-hop relational queries (such as fraud detection) faster than SQL databases.

[0082] 3. Scenario-based support includes: a) Real-time business: Redis + MySQL ensures high-concurrency transactions. b) Intelligent analysis: ClickHouse + Neo4j supports real-time user profiling and path analysis (such as recommendation systems). c) Consistency guarantee: Data consistency is ensured through distributed transactions (XA protocol) and CDC synchronization (such as Debezium).

[0083] The actual effects include:

[0084] 1. Performance Improvement: The response time for complex queries has been optimized from minutes to seconds, and the system maintains high stability in high-concurrency scenarios.

[0085] 2. Cost optimization: Using the database on demand improves resource utilization, reduces redundant processes, and effectively lowers operation and maintenance costs.

[0086] 3. Business value includes: a) Precise recommendation: Neo4j relationship network + ClickHouse behavioral analysis improves recommendation conversion rates. b) Real-time risk control: Multi-database joint queries identify complex fraud patterns. c) Decision efficiency: Dynamic federated queries support instant cross-source analysis, improving decision-making speed.

[0087] This invention also discloses a graph relation integration technology based on relational reasoning, the core applications of which include:

[0088] 1. Multi-source knowledge fusion: Integrating structured (MySQL), semi-structured (JSON logs), and unstructured (PDF papers) data to construct a unified knowledge graph, solving the data silo problem. Cross-domain terminology mapping is achieved through ontology alignment (such as BERT-OWL), for example, associating medical terms with biological terms.

[0089] 2. Deep Relationship Discovery: Based on path reasoning (such as PathCon), it mines hidden associations (such as drug side effects and gene pathways) and supports complex reasoning; it also includes time-series graph analysis (such as Temporal KG) to track relationship evolution.

[0090] 3. Intelligent Decision Support: Dynamic conflict resolution (such as KG-BERT) resolves multi-source knowledge contradictions, improving decision consistency. Real-time inference engines (such as Drools+Neo4j) support millisecond-level business response.

[0091] The actual effects include:

[0092] 1. Knowledge completeness: The coverage of entities in the knowledge graph is improved, the density of relationships is increased, and the efficiency of discovering cross-disciplinary knowledge associations is improved.

[0093] 2. System efficiency: The response time for federated graph queries has been reduced, and the automated knowledge update cycle has been shortened from days to minutes.

[0094] Step S30: Perform user profile generation and book borrowing trend prediction processing based on the target knowledge graph to obtain user profile results and book borrowing trend results.

[0095] The machine learning algorithms and large-scale models applied in this invention include: 1. Using clustering algorithms (such as K-Means and DBSCAN) to classify readers and identify the behavioral patterns of different groups. 2. Applying classification algorithms (such as decision trees, random forests, and XGBoost) to predict readers' borrowing preferences and churn risk. 3. Using regression analysis (such as linear regression and LSTM) to predict borrowing trends and resource demand.

[0096] The deep learning technologies employed in this invention include: 1. Using deep learning models (such as LSTM and GRU) to accurately predict time-series data (such as borrowing volume and resource utilization rate). 2. Applying natural language processing technologies (such as BERT and Transformer) to analyze reader reviews and search keywords to uncover potential needs.

[0097] This invention combines a large model with a target knowledge graph, which can accurately construct user profiles and book borrowing trends based on user reading information and library borrowing information, thereby providing book recommendations and decision-making suggestions to users and making reasonable allocations of library resources.

[0098] Specifically, feature extraction is performed on the target knowledge graph to obtain feature extraction results, which include basic features, behavioral features, and semantic features. A clustering analysis algorithm is then used to perform clustering analysis on the feature extraction results to obtain user behavior patterns. A pre-trained model is determined, and the pre-trained model is trained and fine-tuned based on the target knowledge graph and the user behavior patterns to obtain an end-to-end profile model. User reading data corresponding to the target user is acquired and input into the end-to-end profile model to output user profile results.

[0099] like Figure 2 As shown, the profile generation function in this invention includes: 1. Feature engineering, including basic features: extracting static information (such as reader identity, department, and grade) and statistical features (such as average monthly borrowing volume and high-frequency access time). User behavior and interest features: using LSTM networks to model users' recent book borrowing records and extracting short-term user interest features; using multi-head attention mechanisms to analyze users' long-term book borrowing records and constructing long-term user interest features. Semantic features: using NLP techniques (such as BERT) to analyze reader reviews and search terms, extracting topic distribution and sentiment tendencies. 2. Profile modeling techniques, including cluster analysis: dividing reader groups (such as "research readers" or "leisure reading readers") using algorithms such as K-Means and DBSCAN. Tag system construction: generating tags (such as "high-frequency users" or "interdisciplinary interest users") based on rule engines (such as Drools) and machine learning. Deep learning model: using neural networks (such as Transformer) to build an end-to-end profile model and associate multi-dimensional behavioral features. 3. Real-time updates and dynamic adjustments, including incremental learning: Reader profiles are dynamically updated using online machine learning (such as FTRL and online random forests) to adapt to behavioral changes. Stream processing framework: Apache Flink is used to process behavioral data in real time and update profile labels (e.g., real-time tagging of "recently active users").

[0100] The reader behavior analysis includes: Analysis methods: K-Means clustering to segment reader groups, and Apriori association rule mining to discover borrowing patterns. Tools: Spark MLlib, Weka. Resource utilization analysis includes: Analysis methods: Statistics on resource borrowing rates and dwell time to identify underutilized resources. Tools: SQL, Tableau. Subject-specific service support includes: Analysis methods: Constructing a subject-specific citation network to analyze research hotspots and trends (as shown in the Neo4j database). Tools: Gephi, Cytoscape.

[0101] Obtain time-series data from the target knowledge graph, and aggregate the time-series data to obtain aggregated time-series data; determine a preset deep learning model, and input the aggregated time-series data into the preset deep learning model to output book borrowing trend results.

[0102] Furthermore, based on the user profile results, corresponding recommended book data and predictive decision directions are determined, and the recommended book data and predictive decision directions are pushed to the corresponding target users; the current book resource allocation plan is determined, and the current book resource allocation plan is optimized based on the book borrowing trend results to obtain the target book resource allocation result.

[0103] In its implementation, this invention utilizes a multi-head attention mechanism and a Long Short-Term Memory (LSTM) network to construct a deep learning architecture for analyzing and predicting users' book borrowing behavior. The LSTM network is used to model the user's recent book borrowing records, extracting short-term interest features. The multi-head attention mechanism is used to analyze the user's long-term book borrowing records, constructing long-term interest features. Finally, the user's short- and long-term interest features are combined to predict and analyze their next book borrowing attempt, and the predicted borrowing results are recommended to the user, thereby improving the user's book borrowing experience.

[0104] like Figure 3As shown, the specific technical implementation process of this invention, which uses a multi-head attention mechanism and a long short-term memory network (LSTM) to construct a deep learning architecture for analyzing and predicting users' book borrowing behavior, includes: 1. Treating each user's most recent n book borrowing records as short-term borrowing distances, and using word embedding to model the name of each book as a feature vector; 2. Using a multi-layer long short-term memory network (LSTM) to model the book feature vectors into a user's short-term borrowing interest feature vector; 3. Treating each user's remaining book borrowing records as long-term borrowing records, and using word embedding to model the name of each book as a feature vector; 4. Using a multi-layer multi-head attention mechanism network (MUti-atten) to model the feature sequence into a user's long-term borrowing interest feature vector, and concatenating the user's long-term and short-term borrowing interest feature vectors; 5. Using a fully connected layer network (FC) to compress the features, expanding the compressed features and combining them with the feature vector of each book; 6. Predicting the probability that the user will borrow the book.

[0105] By constructing enterprise and individual profiles, we can accurately match enterprise needs with student abilities. For example, by analyzing enterprise recruitment data and student practice data, we can provide universities with customized talent training programs and recommend outstanding talents that meet the needs of enterprises. This will form an employment ecosystem of "school-enterprise collaboration" and "precise matching", enhance students' employment competitiveness, and meet the talent needs of enterprises.

[0106] The scenario definition and requirements analysis process includes: 1. Defining analysis objectives: such as optimizing resource procurement, improving reader satisfaction, increasing resource utilization, and supporting subject-specific services. 2. Identifying key indicators: such as borrowing volume, search frequency, reader activity, and resource utilization. 3. Scenario modeling and subject-specific service support.

[0107] Resource procurement optimization includes: Analysis methods: predicting future demand based on historical borrowing data and subject hotspots (e.g., time series analysis, regression models). Tools: Python (Pandas, Scikit-learn), R language.

[0108] The graph relationship integration technology described in this invention has the following uses: Purpose: Primarily used to integrate and correlate various resources and data to construct a comprehensive knowledge graph. This technology analyzes the relationships between different data sources (such as books, reader behavior, borrowing records, etc.) to reveal potential connections between data, thereby providing more intelligent services. Effects: Data integration eliminates data silos, enabling seamless connection and integration of data from different departments and systems. Intelligent Recommendation: Through graph analysis, the system can provide personalized book recommendations based on readers' historical behavior and preferences. Decision Support: Helps library managers understand resource usage, reader needs, etc., achieving more precise resource allocation and service optimization.

[0109] By integrating multi-source data resources from the library, the system provides teachers and students with precise teaching support tools and data analysis services. Teachers can utilize the system's borrowing data, resource usage data, and reader behavior data to optimize course design and teaching content. For example, by analyzing popular borrowed books and subject trends, they can adjust teaching focus and recommended reading materials. Students, on the other hand, can access high-quality resources related to their courses through the system's personalized recommendation function, improving learning efficiency. Simultaneously, the system supports real-time data monitoring and dynamic adjustments, helping teachers understand students' learning progress and resource usage in a timely manner, thereby optimizing teaching strategies. Furthermore, the system's knowledge graph and semantic analysis functions can assist teachers in conducting interdisciplinary teaching and research, further enhancing teaching quality and academic influence. Through data-driven teaching optimization and precise resource matching, the system significantly improves teaching effectiveness and efficiency, creating a more efficient learning and research environment for teachers and students.

[0110] The library's intelligent data analysis system statistically analyzes visitor traffic information from various dimensions, uncovers reader needs and usage habits, and correlates this data with library space resource information to optimize resource allocation. This provides data support for improving service quality and enhancing the reader experience. Data sources for library visitor traffic analysis include access control systems, borrowing systems, self-service terminals, Wi-Fi, and intelligent monitoring information. This data records readers' arrival times, dwell times, and activity patterns in real time. Data from different sources can be cross-checked to enhance data utilization value; for example, comparing access control data with borrowing records can identify peak times and areas of high reader activity.

[0111] The focus of library visitor traffic analysis is to identify reader behavior, discovering common visit times, most popular reading areas, and frequently borrowed book types. The analysis results can show the difference in peak visit times between weekdays and weekends, or indicate the density of people in certain areas (such as study rooms or multimedia rooms) during specific times. Libraries can then take corresponding measures, such as extending opening hours, adding seats, and optimizing resource allocation. Visitor traffic analysis helps libraries to precisely allocate resources, such as adjusting the timing and quantity of books and periodicals available for shelving. For high-traffic areas, additional bookshelves, seats, and power outlets can be added, or self-service borrowing and returning machines can be installed to reduce waiting times. Analyzing the number of seminar room reservations and usage identifies peak and off-peak periods. Analyzing seminar room reservation data helps understand usage patterns and demand at different times, allowing for different management strategies. During off-peak periods, flexible reservation policies or special events can be considered to attract more users to the seminar rooms. Finally, visitor flow analysis can also be enhanced with predictive capabilities; by modeling and analyzing historical data, the number of visitors in a future time period can be predicted, allowing for advance planning and coordination to avoid management breakdowns due to excessive visitor flow or resource waste due to insufficient visitor flow; for example, if the library anticipates a surge in visitors during major exams, it can accordingly increase opening hours and increase the number of seats in study rooms.

[0112] Borrowing data is one of the most important types of data for libraries. Systematic analysis of borrowing data can reveal readers' reading preferences and the utilization of library resources. Borrowing data analysis helps libraries optimize resource allocation and supports improvements in collection updates and spatial layout. Borrowing data includes the number of times books are borrowed, the borrowing time, and the borrower's identity (e.g., students, teachers, members of the public). This data is stored chronologically or by category, forming a complete borrowing dataset. By analyzing the borrowing volume of students in different grades, the differences in book demand across academic years can be revealed, allowing for more targeted reading resources to be provided to students of different grades; for example, lower-grade students may need basic textbooks, while upper-grade students tend to prefer professional literature or research materials. Borrowing statistics from different colleges help libraries identify which disciplines and fields have high demand for books, thus enabling the rational allocation of book resources. Furthermore, the ratio of books in the library's collection to those actually borrowed reflects the proportion of books in the collection, assesses the utilization of the collection, empowers book acquisition strategies, eliminates books with low borrowing rates, and optimizes the collection structure.

[0113] By analyzing borrowing frequency, libraries can identify which books are most popular with readers. These books may be particularly popular during certain periods, reflecting the shared interests of readers at that time. Secondly, borrower behavior analysis allows libraries to better understand the needs of different groups, such as students, teachers, and others, by analyzing their borrowing habits. This enables libraries to provide more targeted thematic services. Furthermore, analyzing borrowing frequency and cycles, such as whether certain types of books are frequently borrowed during specific seasons or semesters, helps libraries dynamically adjust resources in future planning. Utilization rate analysis of the collection is a crucial aspect of borrowing data analysis. By calculating the borrowing rate of books in the collection, libraries can identify books with high and low borrowing frequencies, allowing for more informed decisions when updating the collection. Finally, borrowing data analysis can be combined with other data (such as visitor traffic and access to electronic resources) to yield more valuable results. For example, analyzing the correlation between borrowing volume and visitor traffic within a specific time period can optimize library opening hours and service arrangements.

[0114] This invention discloses a user long-term and short-term behavioral interest feature analysis technology based on LSTM+multi-head attention mechanism. Its applications include: modeling and analyzing users' historical book borrowing records, constructing readers' book borrowing interest preference features, thereby constructing user interest profiles and providing a basis for personalized book recommendations.

[0115] The benefits include: 1. Accuracy: To prevent errors in feature analysis caused by changes in user borrowing interests, this invention divides user borrowing history into short-term and long-term borrowing records, and models the short-term and long-term records respectively using LSTM and multi-head attention mechanisms, reducing errors caused by changes in user interests. 2. Lightweight: Multi-head attention mechanisms are effective for sequential time series modeling but have high computational complexity and resource requirements; the LSTM framework has low computational complexity but is difficult to model long-term sequences due to gradient explosion. This invention divides user borrowing records into long-term and short-term records, and uses the LSTM framework to model short-term records, reducing computational complexity while maintaining effectiveness.

[0116] Step S40: Visualize the target knowledge graph, the user profile results, and the book borrowing trend results.

[0117] By introducing advanced visualization tools (such as ECharts and D3.js), the interactivity and dynamic effects of charts can be significantly improved, meeting the personalized needs of different users. Specific technical implementations include: 1. Interactive Chart Design: Using JavaScript libraries such as ECharts and D3.js to create interactive charts (such as bar charts, line charts, and pie charts), allowing users to explore data through clicks, drags, and zooms. Dynamic filtering and linked analysis functions are implemented, allowing users to select specific data dimensions (such as time range or reader groups) and update the chart in real time. 2. Dynamic Data Display: Utilizing ECharts' timeline function to display time-series data (such as borrowing trends and resource utilization rates), and presenting the data change process through animation effects. Using D3.js's transition and interpolation functions to achieve smooth data transitions and dynamic updates. 3. Multi-Dimensional Data Visualization: Using parallel coordinates and radar charts to display multi-dimensional data (such as reader attributes, resource types, and borrowing frequency), helping users discover complex relationships. 4. Report Model Statistical Functions include: a. Data Aggregation and Modeling, including data warehouse technology: using Hadoop, Snowflake, etc. to build a data warehouse, integrating multi-source data such as borrowing records, resource usage, and reader behavior. OLAP Analysis: achieving multi-dimensional data aggregation (e.g., grouping statistics by time, subject, and reader type) through OLAP engines such as ClickHouse and Apache Druid. Statistical Modeling: building statistical models based on SQL or Python (e.g., borrowing volume prediction, resource utilization analysis, user activity calculation). b. Report Template Design, including: BI Tool Integration: designing visual report templates using tools such as Tableau and Power BI, supporting charts (bar charts, heatmaps), tables, and dashboards. Dynamic Parameter Configuration: allowing users to customize filtering conditions (e.g., time period, subject classification) to generate dynamic reports. c. Automated Generation and Updates, including: Scheduled Task Scheduling: triggering ETL processes and report generation on a scheduled basis through tools such as Airflow and Cron to ensure data real-time performance. Incremental calculation: Only updates newly added or changed data, reducing the overhead of full calculation (e.g., daily borrowing volume increment statistics). d. Multi-dimensional statistical support includes: Cross-analysis: Supports multi-dimensional combined statistics (e.g., "trend of computer-related book borrowing by students in a certain college over the past six months"). Comparative analysis: Provides year-on-year and month-on-month comparison functions (e.g., this month's borrowing volume vs. the same period last year).

[0118] The relationship graph display function includes: 1. Data modeling and storage, graph databases (such as Neo4j, TigerGraph): storing entities (such as readers, books, subjects, authors) and relationships (such as "borrowing", "related subjects", "co-authors"). Supporting efficient multi-hop queries (such as "find all books borrowed by a reader and their related subjects"). 2. Knowledge graph construction: using semantic technologies (RDF, OWL) to define entity relationships, and combining natural language processing (NLP) to extract associations from unstructured data (such as documents, reviews).

[0119] Visualization technologies include: Front-end frameworks (such as D3.js, ECharts, Vis.js): dynamically rendering nodes (entities) and edges (relationships), supporting interactive operations such as dragging, zooming, and highlighting. Displaying data hierarchically by entity type (reader, book) or relationship weight (borrowing frequency). 3D visualizations (such as Three.js, WebGL): displaying complex three-dimensional relationship networks (such as interdisciplinary networks, resource reference chains).

[0120] Dynamic querying and real-time updates include: Cypher query language (Neo4j): dynamically generating subgraphs by writing query statements (e.g., MATCH(reader)-[borrowing]->(book)). Real-time data streams (e.g., Kafka, Flink): capturing events such as borrowing and comments, and updating graph relationships in real time (e.g., automatically expanding nodes after adding a new borrowing record).

[0121] Interactive features include: Node filtering and focusing: Users can filter nodes by attributes (such as subject, borrowing time) or focus on key entities (such as popular books). Path exploration: Supports finding the shortest path between two entities (such as "the associated path from reader A to subject D"). Semantic search: Generates dynamic graphs based on natural language queries (such as "display books and borrowers related to artificial intelligence").

[0122] Relationship visualization and interaction include visualization tools (such as D3.js, ECharts, Vis.js): dynamically rendering nodes (entities) and edges (relationships), supporting interactive operations such as dragging, zooming, and highlighting. Semantic search: generating dynamic graphs based on natural language queries (such as "display books and borrowers related to artificial intelligence").

[0123] The output results include: 1. Visual charts: Displaying analysis results using bar charts, line charts, pie charts, heatmaps, etc. (e.g., borrowing trends, resource utilization). 2. Interactive dashboards: Providing dynamic filtering and linked analysis functions, allowing users to customize data viewing (e.g., filtering by time or subject). 3. Reports and documents: Generating reports in PDF, Word, or HTML formats, including analysis conclusions and recommendations (e.g., monthly borrowing analysis reports). 4. API interfaces: Providing RESTful APIs or GraphQL interfaces to support other systems in accessing the analysis results (e.g., recommendation systems).

[0124] Dynamic report analysis technology includes: Purpose: Primarily used to generate real-time, interactive statistical reports, helping users dynamically adjust analysis dimensions and filtering conditions to deeply uncover trends and patterns behind the data. Effects: Real-time updates: Users can instantly view the latest borrowing data, reader behavior, and resource usage, ensuring the timeliness of information. Flexibility: Users can customize report content and display methods as needed, flexibly filtering and comparing data from different dimensions. Precise decision support: Through dynamic analysis, administrators can quickly identify problems and take corresponding measures to improve library resource allocation and operational efficiency.

[0125] Enhancing visualization capabilities is crucial for improving user experience and decision-making efficiency. Introducing advanced visualization tools (such as ECharts, D3.js, and Tableau) improves chart interactivity and dynamic effects, meeting the personalized needs of diverse users. Interactive dashboards are designed to support dynamic filtering, linked analysis, and real-time updates. Combining 3D visualization (such as Three.js) and knowledge graphs to display complex relationship networks transforms complex data into intuitive and easy-to-understand charts, lowering the data comprehension barrier. Furthermore, supporting multi-terminal access and personalized customization significantly improves user experience and decision-making efficiency.

[0126] This invention discloses a dynamic report analysis technology based on interactive analysis:

[0127] Core uses include:

[0128] 1. Real-time Data Insights: Supports low-latency querying of massive amounts of data (e.g., ClickHouse + Presto federated analysis), dynamically generating real-time dashboards such as reading trends and user behavior. Streaming processing (e.g., Flink) achieves second-level data refresh, suitable for scenarios such as promotional event monitoring and IoT device alerts.

[0129] 2. Self-service Exploratory Analysis: Drag-and-drop interaction (such as Tableau / Power BI) allows business users to create reports independently, reducing IT dependence (and improving development efficiency). Natural Language Query (NLQ) supports "voice question - chart generation".

[0130] 3. Intelligent Decision Support: Automatic anomaly detection (such as Prophet time series prediction) flags data fluctuations and pushes early warning notifications. Enhances the analysis and interpretation of data trends, achieving an interpretability score of 80%.

[0131] Actual results: Efficiency has been improved, and chart generation time has been reduced from hours to minutes.

[0132] Furthermore, the scalability of this technology is a key advantage in intelligent data analysis across different structural scenarios. By introducing large models (such as BERT and Gork) and advanced intelligent analysis techniques, the depth, breadth, and flexibility of data analysis can be significantly enhanced, meeting the needs of multiple scenarios and fields.

[0133] In diverse application scenarios, intelligent data analysis technology demonstrates remarkable adaptability and scalability. By integrating cutting-edge artificial intelligence technologies, including deep semantic understanding models in natural language processing (such as DeepSeek and the GPT series), knowledge-enhanced language models (such as BERT), and large models with complex reasoning capabilities (such as Gork), data analysis systems have achieved a qualitative leap. The fusion of these advanced technologies allows data analysis to move beyond traditional structured processing and delve deeper into the semantic relationships and potential patterns behind the data.

[0134] The introduction of knowledge graph technology further enhances the ability to analyze cross-domain data, while the continuous learning mechanism ensures that the system can adapt to new business scenarios and data characteristics in a timely manner. This invention significantly improves capabilities in three dimensions: 1. In terms of analytical depth, it achieves a leap from descriptive analysis to predictive and prescriptive analysis; 2. In terms of application breadth, it covers comprehensive processing from structured data to unstructured multimodal data; 3. In terms of responsiveness, it supports multiple modes from pre-set analysis to on-the-spot exploration, enabling libraries to respond quickly to changes and maintain a competitive edge in digital transformation.

[0135] Furthermore, such as Figure 4 As shown, based on the above-mentioned intelligent analysis method for multi-source book data, the present invention also provides an intelligent analysis system for multi-source book data, wherein the intelligent analysis system for multi-source book data includes:

[0136] The data preprocessing module 51 is used to acquire multi-source book data and preprocess the multi-source book data to obtain target multi-source book data.

[0137] Data relationship integration module 52 is used to perform data relationship integration processing on the target multi-source book data to obtain the target knowledge graph;

[0138] The profile generation and trend prediction module 53 is used to perform user profile generation processing and book borrowing trend prediction processing based on the target knowledge graph to obtain user profile results and book borrowing trend results.

[0139] The visualization module 54 is used to visualize the target knowledge graph, the user profile results, and the book borrowing trend results.

[0140] Furthermore, such as Figure 5 As shown, based on the above-mentioned intelligent analysis method and system for multi-source book data, the present invention also provides a terminal, which includes a processor 10, a memory 20 and a display 30. Figure 5 Only some of the terminal components are shown; however, it should be understood that it is not required to implement all of the components shown, and more or fewer components may be implemented instead.

[0141] In some embodiments, the memory 20 may be an internal storage unit of the terminal, such as a hard disk or memory. In other embodiments, the memory 20 may be an external storage device of the terminal, such as a plug-in hard disk, smart media card (SMC), secure digital card (SD), flash card, etc. Further, the memory 20 may include both internal and external storage devices. The memory 20 is used to store application software and various types of data installed on the terminal, such as the program code installed on the terminal. The memory 20 can also be used to temporarily store data that has been output or will be output. In one embodiment, the memory 20 stores an intelligent analysis program 40 for multi-source book data, which can be executed by the processor 10 to implement the intelligent analysis method for multi-source book data in this application.

[0142] In some embodiments, the processor 10 may be a central processing unit (CPU), a microprocessor, or other data processing chip, used to run program code stored in the memory 20 or process data, such as executing the intelligent analysis method of the multi-source book data.

[0143] In some embodiments, the display 30 may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, or an OLED (Organic Light-Emitting Diode) touchscreen. The display 30 is used to display information on the terminal and to display a visual user interface.

[0144] In one embodiment, the steps of the intelligent analysis method for multi-source book data are implemented when the processor 10 executes the intelligent analysis program 40 for multi-source book data in the memory 20.

[0145] In summary, this invention provides an intelligent analysis method, system, and terminal for multi-source book data. The method includes: acquiring multi-source book data and preprocessing the multi-source book data to obtain target multi-source book data; performing data relationship integration processing on the target multi-source book data to obtain a target knowledge graph; performing user profile generation processing and book borrowing trend prediction processing based on the target knowledge graph to obtain user profile results and book borrowing trend results; and visually displaying the target knowledge graph, the user profile results, and the book borrowing trend results. This invention, by preprocessing and integrating data relationships of multi-source book data to obtain a target knowledge graph, and then using the target knowledge graph for user profile generation and book borrowing trend prediction, enables accurate analysis of user behavior and book borrowing trends. Furthermore, based on the generated user profile results and book borrowing trend results, it enables optimized allocation of book resources, effectively improving the availability of book resources and optimizing the user experience.

[0146] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal that includes that element.

[0147] Of course, those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware (such as a processor, controller, etc.). The program can be stored in a computer-readable storage medium, and when executed, it can include the processes described in the above method embodiments. The computer-readable storage medium can be a memory, magnetic disk, optical disk, etc.

[0148] It should be understood that the application of the present invention is not limited to the examples above. Those skilled in the art can make improvements or modifications based on the above description, and all such improvements and modifications should fall within the protection scope of the appended claims.

Claims

1. An intelligent analysis method for multi-source book data, characterized in that, The intelligent analysis method for multi-source book data includes: Acquire multi-source book data and preprocess the multi-source book data to obtain target multi-source book data; The target multi-source book data is processed by data relationship integration to obtain the target knowledge graph; The step of integrating data relationships in the target multi-source book data to obtain the target knowledge graph specifically includes: A centralized data model is constructed, and the target multi-source book data is processed by the centralized data model to perform entity relationship structuring to obtain structured multi-source book data; The core entities in the structured multi-source book data are identified, and the core entities are subjected to identifier assignment and fuzzy matching to obtain associated multi-source book data. The centralized data model includes star schema and graph schema. A unique ID is assigned to the core entity to enable entity matching across data sources; The fuzzy matching process involves using string similarity algorithms or machine learning models to associate similar but not completely identical records. Entity relations are extracted from the associated multi-source book data using natural language processing and semantic techniques to obtain entity relation extraction results, and a target knowledge graph is constructed based on the entity relation extraction results. A stream processing framework is used to capture borrowing and return events in real time and dynamically update entity relationships. Based on the target knowledge graph, user profile generation and book borrowing trend prediction are performed to obtain user profile results and book borrowing trend results. The process of generating user profiles and predicting book borrowing trends based on the target knowledge graph, to obtain user profile results and book borrowing trend results, specifically includes: The target knowledge graph is subjected to feature extraction and cluster analysis to obtain feature extraction results and user behavior patterns. Based on the feature extraction results and user behavior patterns, user profile generation is performed to obtain user profile results. Obtain time-series data from the target knowledge graph and aggregate the time-series data to obtain aggregated time-series data; A preset deep learning model is determined, and the aggregated time series data is input into the preset deep learning model to output the book borrowing trend results; The process involves performing feature extraction and clustering analysis on the target knowledge graph to obtain feature extraction results and user behavior patterns. Based on these results, a user profile generation process is then performed to obtain the user profile results. Specifically, this includes: The target knowledge graph is subjected to feature extraction processing to obtain feature extraction results, wherein the feature extraction results include basic features, behavioral features and semantic features; The basic features include static information and statistical features; The behavioral characteristics include short-term interest characteristics and long-term user interest characteristics; The LSTM network is used to model the user's recent book borrowing records and extract the user's short-term interest features. By analyzing users' long-term book borrowing records using multi-head attention mechanisms, we can construct users' long-term interest characteristics. NLP techniques were used to analyze reader comments and search terms, extract topic distribution and sentiment trends, and obtain semantic features. The feature extraction results are subjected to cluster analysis using a clustering analysis algorithm to obtain user behavior patterns; A pre-trained model is determined, and the pre-trained model is trained and fine-tuned based on the target knowledge graph and the user behavior pattern to obtain an end-to-end profile model. Obtain user reading data corresponding to the target user, input the user reading data into the end-to-end profiling model, and output user profiling results; A deep learning architecture is constructed using a multi-head attention mechanism and a long short-term memory network, and the user's book borrowing behavior is analyzed and predicted based on the deep learning architecture. The method employs a multi-head attention mechanism and a long short-term memory network to construct a deep learning architecture, and analyzes and predicts users' book borrowing behavior based on this deep learning architecture. Specifically, this includes: Using each user’s n book borrowing records within a preset time period as the short-term borrowing distance, the word embedding method is used to model the name of each book as a book feature vector; The book feature vector is modeled into a user's short-term borrowing interest feature vector using a multi-layer long short-term memory network. Each user's remaining book borrowing records are taken as the user's long-term borrowing records, and the name of each book is modeled as a feature sequence using word embedding method; The feature sequence is modeled as a long-term borrowing interest feature vector by using a multi-layer multi-head attention mechanism network. The short-term borrowing interest feature vector and the long-term borrowing interest feature vector are concatenated and a fully connected layer network is used to compress the features. The compressed features are then expanded and combined with the feature vector of each book to obtain a deep learning architecture. The probability of a user borrowing a book is predicted using the deep learning architecture, thus obtaining the prediction result of the user's book borrowing behavior. The target knowledge graph, the user profile results, and the book borrowing trend results are visualized.

2. The intelligent analysis method for multi-source book data according to claim 1, characterized in that, The process of acquiring multi-source book data and preprocessing the multi-source book data to obtain target multi-source book data specifically includes: A distributed log collection tool is used to acquire multi-source book data, which includes structured data, unstructured data, real-time data streams, and external data sources. The multi-source book data is standardized using metadata standards to obtain standardized multi-source book data. The standardized multi-source book data is processed using a data mapping tool to obtain mapped multi-source book data. The mapped multi-source book data is cleaned to obtain the target multi-source book data.

3. The intelligent analysis method for multi-source book data according to claim 2, characterized in that, The data cleaning process includes data deduplication, missing value filling, formatting, and outlier handling. The step of cleaning the mapped multi-source book data to obtain the target multi-source book data specifically includes: Obtain the data identifier in the mapped multi-source book data, determine the duplicate data in the mapped multi-source book data based on the data identifier, and perform the data deduplication process on the duplicate data to obtain the first multi-source book data; The missing values ​​in the first multi-source book data are obtained, and the missing values ​​are filled using a preset filling method to obtain the second multi-source book data. The preset filling method includes interpolation, mean filling and missing value deletion. Obtain the date format corresponding to the second multi-source book data, and perform the formatting process on the second multi-source book data according to the date format to obtain the third multi-source book data; The outlier detection algorithm is used to identify outliers in the third multi-source book data to obtain target outlier data. The outlier data is then processed to obtain target multi-source book data.

4. The intelligent analysis method for multi-source book data according to claim 1, characterized in that, The process of generating user profiles and predicting book borrowing trends based on the target knowledge graph, to obtain user profile results and book borrowing trend results, further includes: Based on the user profile results, the corresponding recommended book data and predictive decision direction are determined, and the recommended book data and predictive decision direction are pushed to the corresponding target user; Determine the current book resource allocation plan, and optimize the current book resource allocation plan based on the book borrowing trend results to obtain the target book resource allocation result.

5. An intelligent analysis system for multi-source book data, characterized in that, The intelligent analysis system for multi-source book data is used to implement the intelligent analysis method for multi-source book data according to any one of claims 1-4, and the intelligent analysis system for multi-source book data includes: The data preprocessing module is used to acquire multi-source book data and preprocess the multi-source book data to obtain target multi-source book data. The data relationship integration module is used to perform data relationship integration processing on the target multi-source book data to obtain the target knowledge graph; The profile generation and trend prediction module is used to generate user profiles and predict book borrowing trends based on the target knowledge graph, so as to obtain user profile results and book borrowing trend results. The visualization module is used to visualize the target knowledge graph, the user profile results, and the book borrowing trend results.

6. A terminal, characterized in that, The terminal includes: a memory, a processor, and an intelligent analysis program for multi-source book data stored in the memory and executable on the processor. When the intelligent analysis program for multi-source book data is executed by the processor, it implements the steps of the intelligent analysis method for multi-source book data as described in any one of claims 1-4.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores an intelligent analysis program for multi-source book data, which, when executed by a processor, implements the steps of the intelligent analysis method for multi-source book data as described in any one of claims 1-4.

Images