Book personalized recommendation method and system based on big data analysis

Abstract

The invention belongs to the technical field of data processing, and discloses a book personalized recommendation method based on big data analysis. The recommendation method comprises the following specific steps: S1: data integration and cleaning; S2: data storage and index optimization; S3: personalized recommendation algorithm design; S4: keyword interaction mechanism; according to the invention, through deep cooperation of data storage optimization and recommendation algorithm design, double breakthrough of performance and precision is realized; a B + tree joint index of MySQL enables a collaborative filtering algorithm to obtain a user borrowing sequence at a millisecond level and support time decay weighted calculation, and a MongoDB composite index compresses response time of community theme extension recommendation, so that recommendation generation time consumption is reduced through cooperation of the B + tree joint index and the MongoDB composite index, and a concurrence scene of thousands of people can be easily coped with.

Description

Technical Field

[0001] The present invention belongs to the field of data processing technology, and specifically relates to a personalized book recommendation method and system based on big data analysis. Background Art

[0002] Personalized book recommendation methods based on big data analysis are a precise recommendation model that relies on massive, multi-source, heterogeneous data processing technologies and intelligent algorithms. Its core approach is to construct a dynamically updated personal interest profile through in-depth mining and analysis of readers' multidimensional behavioral data, thereby generating a customized book list that is highly tailored to their individual needs. Traditional book recommendation methods rely primarily on simple statistical information or rule-based algorithmic design. While these methods can provide a certain degree of recommendation relevance by analyzing static indicators such as borrowing frequency and popular book lists, they have significant limitations. First, they struggle to capture the dynamics of readers' interests and cannot timely reflect the shifting trajectory of readers' preferences. Second, these methods lack in-depth mining of complex borrowing behavior patterns. As a result, the accuracy and conversion rate of recommendations are often unsatisfactory when faced with massive library resources and diverse reader needs. In today's big data environment, readers' reading behavior is fragmented, contextual, and personalized. Traditional methods are unable to identify potential interest connections across domains and are unable to respond to ever-changing reading needs, significantly reducing the intelligence of recommendations and service experience. This limitation not only affects the effective use of library resources, but also, to a certain extent, restricts the depth and breadth of readers' knowledge acquisition, highlighting the necessity and urgency of introducing big data analysis and intelligent algorithms. Summary of the Invention

[0003] The purpose of the present invention is to provide a method and system for personalized book recommendation based on big data analysis to solve the problems raised in the above background technology.

[0004] To achieve the above objectives, the present invention provides the following technical solution: a personalized book recommendation method based on big data analysis, the specific steps of the recommendation method are as follows: S1: Data integration and cleaning: The system integrates multi-source data from the academic affairs department, library, and community, connects data through interfaces in real time, cleans and standardizes it, and stores it in the data warehouse. For example, it extracts information about course names and borrowed books. S2: Data storage and index optimization: The system uses MySQL to store structured data and MongoDB to process unstructured data. It also creates indexes for key fields such as student ID and book ID to improve query efficiency. S3: Personalized recommendation algorithm design: The system integrates collaborative filtering and content recommendation, analyzes the curriculum and textbook associations, analyzes borrowing behavior for weighted recommendations, and expands the book list based on community themes to achieve personalized book push; S4: Keyword interaction mechanism: Students dynamically manage keywords, the system adjusts recommendation weights in real time, and regularly analyzes trends to optimize the model; S5: Student book recommendation portal and review mechanism: Students submit their recommendations through the book recommendation interface, and administrators will review and decide whether to adopt them based on a comprehensive consideration of the library's collection, policies, and funding. S6: Recommendation result display and feedback: The system displays the recommended book details, and students give feedback by likes / comments / disinterest. Algorithm optimization improves accuracy and satisfaction. S7: System performance optimization and maintenance: Regularly optimize database queries, indexes, and models, and monitor performance to ensure high-concurrency and stable operation.

[0005] Preferably, the specific steps of data integration and cleaning in S1 are as follows: Step 1: Multi-source data interface docking and real-time collection The system configured the academic affairs API to periodically pull course schedule data, extracting core fields such as course number and name; deployed library middleware, used JDBC to capture borrowing records in real time, and calculated 30-day borrowing activity; developed a club webhook to monitor registration events and parse JSON to obtain club ID and skill tag information; Step 2: Heterogeneous data cleaning and standardization Course schedule data cleaning: Analyze the syllabus using regular expressions to extract the core elements of teaching objectives and assessment methods; build a standardized dictionary of course names and merge courses with different names for advanced mathematics AI; verify the validity of reference book ISBNs and connect to the National Version Data Center API to automatically complete publisher information; Borrowing record processing: Unify borrowing time as a standard timestamp and calculate the average borrowing period of books; associate OpenLibrary metadata through ISBN and add book subject classification tags; establish an anomaly detection model to trigger manual review when the daily borrowing volume exceeds 20 books; Step 3: Create a star schema data warehouse Build a star-shaped data warehouse: take student behavior as the core fact table, associate the course schedule ID, borrowing ID and activity ID, and provide four-dimensional tables for time, space, book subject, and course knowledge field; connect borrowing and club data through student ID, and realize the linkage between course bibliography and library collection through ISBN mapping, and generate a cross-domain feature matrix of excellent mathematics students + high activity.

[0006] Preferably, the specific steps of data storage and index optimization in S2 are as follows: Step 1: MySQL structured data storage and index optimization MySQL is used to store structured data. A student behavior fact table is designed to integrate student ID and course ID foreign key fields, and the InnoDB engine is used to ensure transaction consistency. To meet high-frequency query requirements, a B+ tree joint index of student ID and borrowing time is created in the borrowing record table. Slow query log monitoring is also deployed, and execution plans are regularly analyzed. Index completion or query rewriting is performed for SQL statements that take more than 1 second to scan the entire table. Step 2: MongoDB unstructured data storage and index acceleration MongoDB is used to store unstructured data, and nested documents are used to flexibly carry metadata about community activities. The BSON format feature is used to support free field expansion. Composite indexes are created in the borrowing record collection to optimize the efficiency of multi-dimensional retrieval by exact matching by student ID, ISBN range query, and reverse chronological order. Full-text indexes are established for the syllabus text fields, and keyword searches are implemented in combination with the text operator to quickly locate course documents containing machine learning terms, significantly improving the retrieval performance of unstructured data.

[0007] Preferably, the specific steps of designing the personalized recommendation algorithm in S3 are as follows: Step 1: Recommend textbooks based on the curriculum Extract student class schedules from the data warehouse, parse the syllabus, and extract course keywords using NLP. Then, link the course dimension table to match reference book topics. Retrieve relevant textbooks through the library API, and generate a supporting book list based on course priority, which is then pushed to the student end. Step 2: Collaborative filtering recommendation based on borrowing behavior A user-book borrowing matrix is constructed and cosine similarity between students is calculated to identify similar user groups for personalized book recommendations for the top 20 books. Time-decay weighting is then applied to the borrowing records of similar users to filter out books they have borrowed. Finally, a "recommended book list for fellow enthusiasts" is generated based on the circulation popularity scores in the book dimension table and displayed in descending order by weighted score, achieving precise recommendations. Step 3: Cross-domain fusion recommendation based on community themes The community to which the student belongs is identified through the community participation record, and the high-frequency borrowing topics in the community's history are extracted. Then, the associated book tag system is expanded to recommend to adjacent fields. Finally, the activity coefficient is calculated by combining the borrowing frequency in the student portrait and the community contribution.

[0008] Preferably, the keyword interaction mechanism in S4 refers to the system supporting students to dynamically manage personalized keywords through an interactive interface, triggering real-time adjustment of recommendation algorithm weights, and accurately matching demand preferences; at the same time, the system periodically analyzes the frequency and trend of keyword usage, and uses machine learning to dynamically optimize recommendation model parameters.

[0009] Preferably, the S5 student book recommendation entry and review mechanism refers to students submitting book recommendation requests through the system's book recommendation interface, including book title, author and recommendation reason information; when library administrators review these requests in the background, they will decide whether to adopt students' suggestions based on the library's collection situation, procurement policy and funding factors.

[0010] Preferably, the specific steps of displaying and providing feedback on the recommendation results in S6 are as follows: Step 1: Visualization of recommendation results After logging into the system, students enter the personalized recommendation area, where a list of books is displayed in the form of cards. Each card contains the title, author, introduction, collection status, and user ratings. Books are sorted by recommendation priority, support filtering by subject classification / popularity, and provide a quick entry for "trial reading chapters" and an extended recommendation module for "similar books"; Interactive buttons are set under each book: Like, Comment, Not Interested, and a lightweight feedback form will pop up after clicking; Step 2: Feedback-driven model iteration The system captures user feedback behavior in real time, injects likes data into the collaborative filtering model as positive samples, and marks as uninterested triggers weight decay of negative samples; Daily incremental training of the recommendation model, adjusting the weights of relevant dimensions in the user feature vector, and optimizing content filtering rules; When the model is fully updated every week, the semantic analysis of the review text is combined to dynamically adjust the book difficulty label matching threshold to form a feedback loop.

[0011] Preferably, the system performance optimization and maintenance in S7 refers to the development team regularly optimizing the database and implementing full-link performance monitoring, combined with load balancing and caching strategies to ensure that the system runs stably and efficiently in high-concurrency scenarios.

[0012] Preferably, the system includes a multi-source data integration and preprocessing module, a distributed data storage and index optimization module, a hybrid recommendation engine module, a user interaction and feedback closed-loop module, an intelligent model tuning and monitoring module, and a user portrait and behavior analysis module; The multi-source data integration and preprocessing module is used to connect to multi-source data in real time, build a star-shaped warehouse after cleaning and standardization, and implement triple quality control verification; the distributed data storage and index optimization module uses MySQL+InnoDB to store structured data, and B+ tree index accelerates query; MongoDB uses BSON to store unstructured data, and compound indexes and full-text indexes support multi-dimensional retrieval; the hybrid recommendation engine module is used to integrate collaborative filtering, content recommendation, cross-domain recommendation, and dynamically adjust weights to achieve accurate push; the user interaction and feedback closed-loop module is used to submit recommendations at the book recommendation entrance, and the administrator reviews and decides whether to adopt them based on the collection, policy, and funding. The results are displayed on the card, and feedback drives model optimization; the intelligent model tuning and monitoring module is used to deploy slow query analysis, load balancing and caching, optimize indexes and models, tune thresholds, and ensure high concurrency and stability; the user portrait and behavior analysis module constructs user portraits based on borrowing, class schedules, and community data, and calculates activity to support recommendation optimization.

[0013] The beneficial effects of the present invention are as follows: 1. This invention achieves a double breakthrough in performance and accuracy through the deep collaboration of data storage optimization and recommendation algorithm design; MySQL's B+ tree joint index enables the collaborative filtering algorithm to obtain user borrowing sequences in milliseconds, supporting time-decay weighted calculations, while MongoDB's composite index compresses the response time of community theme extension recommendations. The combination of the two reduces the time consumption of recommendation generation and easily copes with the concurrent scenario of thousands of people; in terms of unstructured data, MongoDB's full-text index increases the retrieval efficiency of syllabus keywords by 3 times, directly driving the improvement of the accuracy of textbook-related recommendations; the community activity metadata stored in nested documents provides a real-time data source for activity calculation, making cross-domain recommendation click-through rate More importantly, after the performance bottleneck discovered by slow query log analysis was resolved through index optimization, the SQL execution efficiency of the textbook recommendation module increased by 6 times. The "too difficult" label fed back by the recommendation model in turn drove the ISBN mapping table to add a book difficulty coefficient field, forming a closed-loop iterative mechanism of "storage optimization-recommendation improvement-data supplementation". This hybrid architecture of InnoDB and MongoDB not only ensures the transaction consistency of collaborative filtering matrix calculations, but also supports dynamic updates of the NLP keyword dictionary through the flexible scalability of the BSON format, enabling the system to simultaneously meet the dual needs of high-frequency transaction queries and analytical computing, building a data-driven intelligent recommendation ecosystem.

[0014] 2. The present invention builds a data-driven intelligent recommendation closed loop through a five-layer technical architecture: at the data base layer, it connects to the three major interfaces of academic timetable, library borrowing, and club activities in real time, constructs a star-shaped data warehouse through standardized cleaning, and generates a cross-domain user portrait of "mathematical exceller + high activity"; at the storage and computing layer, MySQL joint indexing improves the computing efficiency of collaborative filtering matrix, and MongoDB full-text indexing drives the NLP parsing speed to increase. The hybrid architecture supports responses in scenarios with thousands of concurrent users, while meeting the needs of high-frequency trading queries and analytical computing; at the core algorithm layer, it integrates three engines: timetable keyword association, time-decayed weighted collaborative filtering, and club theme expansion, and combines dynamic keyword weighting to achieve real-time demand matching; at the feedback optimization layer, the model is iterated daily through like / disinterest feedback, and the book difficulty threshold is optimized through comment text semantic analysis, forming a "recommendation-feedback-optimization" self-learning cycle; at the system security layer, relying on slow query log analysis and load balancing strategy to ensure SQL execution efficiency in high-concurrency scenarios, and the book recommendation review mechanism is combined with collection policy to achieve precise resource allocation. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a schematic diagram of the overall use process of the present invention. DETAILED DESCRIPTION

[0016] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0017] like Figure 1 As shown, the embodiment of the present invention provides a personalized book recommendation method based on big data analysis. The specific steps of the recommendation method are as follows: S1: Data integration and cleaning: The system integrates multi-source data from the academic affairs department, library, and community, connects data through interfaces in real time, cleans and standardizes it, and stores it in the data warehouse. For example, it extracts information about course names and borrowed books. S2: Data storage and index optimization: The system uses MySQL to store structured data and MongoDB to process unstructured data. It also creates indexes for key fields such as student ID and book ID to improve query efficiency. S3: Personalized recommendation algorithm design: The system integrates collaborative filtering and content recommendation, analyzes the class schedule and associated textbooks, analyzes borrowing behavior for weighted recommendations, and expands the book list based on the community theme to achieve personalized book push.

[0018] S4: Keyword interaction mechanism: Students dynamically manage keywords (such as authors, subjects), the system adjusts the recommendation weights in real time, and regularly analyzes trend optimization models; if "artificial intelligence" is frequently added, relevant book recommendations will be enhanced to achieve demand-driven precise iteration.

[0019] S5: Student book recommendation portal and review mechanism: Students submit recommendations (including book title, author and reason for recommendation) through the book recommendation interface. When reviewing, the administrator will consider the library's collection, policies and funding situation to decide whether to adopt it.

[0020] S6: Recommendation result display and feedback: The system displays the recommended book details, and students give feedback by likes / comments / disinterest. Algorithm optimization improves accuracy and satisfaction. S7: System performance optimization and maintenance: Regularly optimize database queries, indexes, and models, and monitor performance to ensure high-concurrency and stable operation.

[0021] Example 1: Accurate portrait construction driven by multi-source data fusion After deploying this system at a university, it synchronized course schedule data hourly through the academic affairs API, parsing out course keywords such as "data structure" and "machine learning." Library middleware captured borrowing records in real time, linking them to textbooks like "Introduction to Algorithms" through ISBN mapping. Club webhooks captured activity tags for programming clubs. The system constructed a star-shaped warehouse, generating a "Computer Science + High Activity" feature matrix. This enabled the targeted push of "Using Python for Data Analysis" to students taking the Python elective, resulting in a 42% increase in textbook borrowing rates.

[0022] The specific steps of data integration and cleaning in S1 are as follows: Step 1: Multi-source data interface docking and real-time collection The system configures the academic affairs API to periodically pull course schedule data, extracting core fields such as course number and name; deploys library middleware, uses JDBC to capture borrowing records in real time, and calculates 30-day borrowing activity; develops a club webhook to monitor registration events, parses JSON to obtain information such as club ID and skill tags, and implements automated multi-source data collection; Step 2: Heterogeneous data cleaning and standardization Course data cleaning: Analyze the syllabus using regular expressions to extract core elements such as teaching objectives and assessment methods; construct a standardized dictionary of course names and merge courses with synonymous names such as advanced mathematics AI; verify the validity of reference book ISBNs and connect to the National Version Data Center API to automatically complete publisher information to ensure the standardization and integrity of course data; Borrowing record processing: Unify borrowing time as a standard timestamp and calculate the average borrowing period (total days / number of times); link OpenLibrary metadata through ISBN and add book subject classification tags; establish an anomaly detection model to trigger manual review when the daily borrowing volume exceeds 20 books to ensure the quality and rationality of borrowing data; Step 3: Create a star schema data warehouse Build a star-shaped data warehouse: With student behavior as the core fact table, associate course schedule ID, borrowing ID, and activity ID, and provide four-dimensional tables for time (semester / week), space (classroom location), book subject, and course knowledge domain. Use student IDs to connect borrowing and club data, and ISBN mapping enables linkage between course bibliographies and library collections, generating cross-domain feature matrices such as "excellent math students + high activity." Deploy triple quality control: integrity checks ensure a fill rate of over 95% for core fields, a timeliness dashboard monitors data synchronization delays (less than 10 minutes for academic affairs data), and consistency checks compare the distribution of student IDs for course selection and borrowing, intelligently identifying outliers.

[0023] This data integration and cleaning solution significantly enhances data value through triple optimization: automated collection enables real-time synchronization of academic affairs, library, and community data to ensure information timeliness; standardized processing integrates regular analysis, ISBN verification, and anomaly detection to improve the uniformity of course names and the accuracy of borrowing data; a star-shaped warehouse constructs a cross-domain feature matrix to support the generation of precise portraits such as "excellent math students + high activity levels", and triple quality control ensures improved integrity of core fields and data synchronization delays within 10 minutes; this system not only achieves efficient integration of heterogeneous data, but also connects course bibliographies with library collections through ISBN mapping, providing a high-quality data base for the recommendation system and supporting subsequent personalized services and intelligent analysis.

[0024] The specific steps of data storage and index optimization in S2 are as follows: Step 1: MySQL structured data storage and index optimization MySQL is used to store structured data. A student behavior fact table is designed to integrate foreign key fields such as student ID and course ID, and transaction consistency is ensured through the InnoDB engine. To meet high-frequency query requirements, a B+ tree joint index of student ID and borrowing time is created in the borrowing record table, significantly improving the efficiency of searches by student ID and time range. Slow query log monitoring is also deployed, and execution plans are regularly analyzed. Index completion or query rewriting is performed for SQL statements that take more than 1 second to scan the entire table, continuously optimizing database performance. Step 2: MongoDB unstructured data storage and index acceleration MongoDB is used to store unstructured data, and nested documents are used to flexibly carry community activity metadata (such as dynamic content and participation records). The BSON format supports free field expansion. A composite index of {student ID: 1, personalized book recommendation ID: 1, personalized book recommendation borrowing time: -1} is created in the borrowing record collection to optimize the efficiency of multi-dimensional retrieval by exact matching by student ID, ISBN range query, and reverse chronological order. A full-text index is established for the syllabus text field, and keyword search is implemented in combination with the text operator to quickly locate course documents containing terms such as "machine learning", significantly improving the retrieval performance of unstructured data.

[0025] The specific steps for designing the personalized recommendation algorithm in S3 are as follows: Step 1: Recommend textbooks based on the curriculum Extract student class schedules from the data warehouse, parse the syllabus, and extract course keywords using NLP. Then, link the course dimension table to match reference book topics. Retrieve relevant textbooks (such as "Deep Learning") through the library API, and generate a supporting book list based on course priority, which is then pushed to the student end. Step 2: Collaborative filtering recommendation based on borrowing behavior A user-book borrowing matrix is constructed and cosine similarity between students is calculated to identify similar user groups for personalized book recommendations for the top 20 books. Time-decay weighting is then applied to the borrowing records of similar users to filter out books they have borrowed. Finally, a "recommended book list for fellow enthusiasts" is generated based on the circulation popularity scores in the book dimension table and displayed in descending order by weighted score, achieving precise recommendations. Step 3: Cross-domain fusion recommendation based on community themes Through the club participation records, we can identify the community to which students belong (such as "Photography Workshop"), extract the high-frequency borrowing themes in the history of the community (such as "Composition Skills" and "Post-Processing"), and then expand the associated book tag system to recommend to adjacent fields (such as from "Photography" to "Travel Literature"). Finally, we combine the borrowing frequency in the student portrait (weight 0.6) and the club contribution (weight 0.4) to calculate the activity coefficient, generate a personalized "Club Theme Advanced Book List", and achieve precise penetration of interest circles.

[0026] Example 2: Cross-domain recommendation practice with dynamic weight adjustment During the first semester of the new semester, the system detected a surge in enrollment in the "Foundations of Artificial Intelligence" course. NLP parsed the syllabus and extracted the keyword "neural network," triggering a collaborative filtering model to dynamically adjust its weights. The model, combined with community data, discovered that a student participated in a robotics club and expanded its recommendations to "Deep Learning" and "Hands-on Reinforcement Learning." Using MongoDB's full-text index to match the "reinforcement learning" section in the syllabus, a cross-domain book list was generated, resulting in a 68% adoption rate for new students' majors.

[0027] The keyword interaction mechanism in S4 refers to the system's ability to dynamically manage personalized keywords (such as authors and subject areas) through an interactive interface, triggering real-time adjustments to the recommendation algorithm's weights to precisely match students' needs and preferences. Furthermore, the system periodically analyzes keyword usage frequency and trends, applying machine learning to dynamically optimize recommendation model parameters. For example, if the system detects a surge in the use of the keyword "artificial intelligence," it automatically increases the recommendation priority and exposure of books in that field, forming a demand-driven intelligent iteration mechanism.

[0028] Among them, the S5 middle school student book recommendation entrance and review mechanism refers to students submitting book recommendation requests through the system's book recommendation interface, including information such as the book title, author, and reason for recommendation; when library administrators review these requests in the background, they will decide whether to adopt students' suggestions based on factors such as the library's collection situation, procurement policies, and funding.

[0029] The student recommendation portal and review mechanism enhances participation through open recommendation channels. This, combined with a multi-dimensional assessment of collections, policies, and funding, enables precise resource allocation. Student demand directly drives purchasing decisions, optimizing collection structure and improving funding efficiency. This creates a closed-loop demand response, enhancing the relevance of library services and user satisfaction.

[0030] The specific steps of displaying and providing feedback on the recommendation results in S6 are as follows: Step 1: Visualization of recommendation results After logging into the system, students enter the personalized recommendation area, where a list of books is displayed in the form of cards. Each card contains the title, author, introduction, collection status, and user ratings. Books are sorted by recommendation priority, support filtering by subject classification / popularity, and provide a quick entry for "trial reading chapters" and an extended recommendation module for "similar books"; Interactive buttons are set under each book: Like, Comment, Not Interested. Clicking them will pop up a lightweight feedback form (with an optional reason label). Step 2: Feedback-driven model iteration The system captures user feedback behavior in real time, injects likes data into the collaborative filtering model as positive samples, and marks as uninterested triggers weight decay of negative samples; Daily incremental training of the recommendation model, adjusting the weights of relevant dimensions in the user feature vector, and optimizing content filtering rules; When the model is fully updated every week, the book difficulty label matching threshold is dynamically adjusted based on the semantic analysis of the review text (such as comments such as "too difficult") to form a feedback loop.

[0031] Example 3: Feedback closed-loop optimization for high-concurrency scenarios During final exam preparation, the system received a large number of comments stating "too difficult." The semantic analysis model identified a need to lower the difficulty level of the "Probability Theory and Mathematical Statistics" textbook. The development team analyzed slow query logs and discovered that a full table scan of borrowing records took too long. They urgently added a joint index (student ID, borrowing time) to the query.

[0032] The system utilizes a card-style recommendation interface and real-time feedback mechanism to achieve precise push notifications and self-optimization. A visual layout improves browsing efficiency, while interactive buttons instantly capture preferences. This drives collaborative filtering and daily incremental adjustments to content models (for example, increasing AI preference weights by 15%). Combined with semantic analysis, the system dynamically adjusts recommendation thresholds, forming a closed loop of "recommendation-feedback-optimization" to continuously enhance recommendation relevance and user engagement.

[0033] Among them, the system performance optimization and maintenance in S7 refers to the development team regularly optimizing the database (adjusting queries, indexes, and models), implementing full-link performance monitoring, and combining load balancing and caching strategies to ensure stable and efficient operation of the system in high-concurrency scenarios.

[0034] The system regularly optimizes database queries, indexes, and models, combined with full-link performance monitoring and load balancing strategies, to ensure stable and efficient operation in high-concurrency scenarios. Real-time monitoring and intelligent caching reduce latency, improve response speed, avoid system bottlenecks, ensure a smooth user experience, and support continuous business expansion.

[0035] Among them, the system includes multi-source data integration and preprocessing module, distributed data storage and index optimization module, hybrid recommendation engine module, user interaction and feedback closed-loop module, intelligent model tuning and monitoring module, and user portrait and behavior analysis module.

[0036] The multi-source data integration and pre-processing module is used to connect multi-source data in real time, build a star-shaped warehouse after cleaning and standardization, and implement triple quality control verification; Real-time connection to multi-source data interfaces such as academic affairs, library, and community, complete data cleaning (such as ISBN completion and borrowing quantity verification) through regular analysis, standardized dictionary, and anomaly detection models, build a star data warehouse and implement triple quality control (completeness, timeliness, and consistency verification).

[0037] The distributed data storage and index optimization module uses MySQL+InnoDB to store structured data, and B+ tree index to accelerate query; MongoDB uses BSON to store unstructured data, and composite index and full-text index support multi-dimensional search; MySQL+InnoDB is used to store structured data (such as student behavior fact tables), and B+ tree joint indexes are used to accelerate high-frequency queries. MongoDB's BSON format is used to flexibly store unstructured data (such as club activity metadata), and composite indexes (such as {student ID: 1, borrowing time: -1}) and full-text indexes of teaching syllabi are established to support multi-dimensional retrieval.

[0038] The hybrid recommendation engine module is used to integrate collaborative filtering, content recommendation, cross-domain recommendation, and dynamically adjust weights to achieve accurate push; Accurate push notifications are achieved by integrating collaborative filtering (user similarity calculation + time decay weighting), content recommendation (NLP analysis of course schedule keywords to associate teaching materials) and cross-domain recommendation (expanding club topics to adjacent fields), combined with dynamic keyword weight adjustment (such as increasing recommendation priority when the popularity of "artificial intelligence" surges).

[0039] The user interaction and feedback closed-loop module is used to submit recommendations at the book recommendation portal. Administrators review and decide whether to adopt the recommendations based on the library's collection, policies, and funding. Cards display the results, and feedback drives model optimization.

[0040] It provides a book recommendation portal (submit book title / author / recommendation reason) and an audit mechanism (combined with collections / policies / funding decisions), and displays recommendation results (including ratings and trial reading portals) through a card-style interface, captures likes / comments / disinterest feedback, and drives daily incremental training of the model (such as adjusting the AI preference score by 15%) and weekly full optimization.

[0041] The intelligent model tuning and monitoring module is used to deploy slow query analysis, load balancing and caching, optimize indexes and models, tune thresholds, and ensure high concurrency and stability; Deploy slow query log analysis, load balancing strategies, and caching mechanisms, regularly optimize database queries (such as completing time-consuming SQL indexes) and recommendation model parameters, and dynamically adjust matching thresholds based on comment text semantic analysis (such as the "too difficult" tag) to ensure system stability in high-concurrency scenarios.

[0042] The user portrait and behavior analysis module builds user portraits based on borrowing, class schedule, and community data, and calculates activity to support recommendation optimization.

[0043] A multi-dimensional feature matrix (e.g., "Excellent Math Student + Highly Active") is constructed based on borrowing records, class schedule data, and club participation. The activity coefficient (borrowing frequency weighted 0.6 + club contribution weighted 0.4) is calculated to support the generation of personalized book lists and feedback loop iteration. This personalized book recommendation system significantly improves recommendation accuracy and system efficiency through the collaborative operation of six major modules. The multi-source data integration module integrates academic affairs, library, and community data in real time, and constructs a star-shaped warehouse after cleaning and standardization to ensure data quality. The distributed storage module uses a combination of MySQL+InnoDB and MongoDB, and optimizes query efficiency through B+ trees, composite indexes, and full-text indexes to support high-frequency access scenarios. The hybrid recommendation engine integrates collaborative filtering, content recommendation, and cross-domain recommendation technologies, combined with a dynamic keyword weighting mechanism to achieve demand-driven precision push. The user interaction module provides a book recommendation portal and a card-style feedback interface, and uses likes / comments data to drive daily model iterations. The intelligent monitoring module deploys slow query analysis and load balancing strategies, and dynamically adjusts thresholds based on comment text semantic analysis to ensure high concurrency and stability. The user portrait module calculates the activity coefficient based on a multi-dimensional feature matrix to support the generation of personalized book lists. The entire system forms a closed-loop ecosystem driven by data, algorithm optimization, and user feedback, ensuring efficient and stable system operation while improving recommendation satisfaction.

[0044] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0045] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A personalized book recommendation method based on big data analysis, characterized by: The specific steps of this recommended method are as follows: S1: Data integration and cleaning: The system integrates multi-source data from the academic affairs department, library, and community, connects data through interfaces in real time, cleans and standardizes it, and stores it in the data warehouse. It also extracts information about course names and borrowed books. S2: Data storage and index optimization: The system uses MySQL to store structured data and MongoDB to process unstructured data. It also creates indexes for key fields such as student ID and book ID to improve query efficiency. S3: Personalized recommendation algorithm design: The system integrates collaborative filtering and content recommendation, analyzes the curriculum and textbook associations, analyzes borrowing behavior for weighted recommendations, and expands the book list based on community themes to achieve personalized book push; S4: Keyword interaction mechanism: Students dynamically manage keywords, the system adjusts recommendation weights in real time, and regularly analyzes trends to optimize the model; S5: Student book recommendation portal and review mechanism: Students submit their recommendations through the book recommendation interface, and administrators will review and decide whether to adopt them based on a comprehensive consideration of the library's collection, policies, and funding. S6: Recommendation result display and feedback: The system displays the details of recommended books, and students give feedback that they are not interested. Algorithm optimization improves accuracy and satisfaction; S7: System performance optimization and maintenance: Regularly optimize database queries, indexes, and models, and monitor performance to ensure high-concurrency and stable operation.

2. The personalized book recommendation method based on big data analysis according to claim 1, characterized in that: The specific steps of data integration and cleaning in S1 are as follows: Step 1: Multi-source data interface docking and real-time collection The system configured the academic affairs API to periodically pull course schedule data, extracting core fields such as course number and name; deployed library middleware, used JDBC to capture borrowing records in real time, and calculated 30-day borrowing activity; developed a club webhook to monitor registration events and parse JSON to obtain club ID and skill tag information; Step 2: Heterogeneous data cleaning and standardization Course schedule data cleaning: Analyze the syllabus using regular expressions to extract the core elements of teaching objectives and assessment methods; build a standardized dictionary of course names and merge courses with different names for advanced mathematics AI; verify the validity of reference book ISBNs and connect to the National Version Data Center API to automatically complete publisher information; Borrowing record processing: Unify borrowing time as a standard timestamp and calculate the average borrowing period of books; associate OpenLibrary metadata through ISBN and add book subject classification tags; establish an anomaly detection model to trigger manual review when the daily borrowing volume exceeds 20 books; Step 3: Create a star schema data warehouse Build a star-shaped data warehouse: take student behavior as the core fact table, associate the course schedule ID, borrowing ID and activity ID, and provide four-dimensional tables for time, space, book subject, and course knowledge field; connect borrowing and club data through student ID, and realize the linkage between course bibliography and library collection through ISBN mapping, and generate a cross-domain feature matrix of excellent mathematics students + high activity.

3. The personalized book recommendation method based on big data analysis according to claim 1, characterized in that: The specific steps for data storage and index optimization in S2 are as follows: Step 1: MySQL structured data storage and index optimization MySQL is used to store structured data. A student behavior fact table is designed to integrate student ID and course ID foreign key fields, and the InnoDB engine is used to ensure transaction consistency. To meet high-frequency query requirements, a B+ tree joint index of student ID and borrowing time is created in the borrowing record table. Slow query log monitoring is also deployed, and execution plans are regularly analyzed. Index completion or query rewriting is performed for SQL statements that take more than 1 second to scan the entire table. Step 2: MongoDB unstructured data storage and index acceleration MongoDB is used to store unstructured data, and nested documents are used to flexibly carry metadata about community activities. The BSON format feature is used to support free field expansion. Composite indexes are created in the borrowing record collection to optimize the efficiency of multi-dimensional retrieval by exact matching by student ID, ISBN range query, and reverse chronological order. Full-text indexes are established for the syllabus text fields, and keyword searches are implemented in combination with the text operator to quickly locate course documents containing machine learning terms, significantly improving the retrieval performance of unstructured data.

4. The method for personalized book recommendation based on big data analysis according to claim 1, characterized in that: The specific steps of designing the personalized recommendation algorithm in S3 are as follows: Step 1: Recommend textbooks based on the curriculum Extract student class schedules from the data warehouse, parse the syllabus, and extract course keywords using NLP. Then, link the course dimension table to match reference book topics. Retrieve relevant textbooks through the library API, and generate a supporting book list based on course priority, which is then pushed to the student end. Step 2: Collaborative filtering recommendation based on borrowing behavior A user-book borrowing matrix is constructed and cosine similarity between students is calculated to identify similar user groups for personalized book recommendations for the top 20 books. Time-decay weighting is then applied to the borrowing records of similar users to filter out books they have borrowed. Finally, a "recommended book list for fellow enthusiasts" is generated based on the circulation popularity scores in the book dimension table and displayed in descending order by weighted score, achieving precise recommendations. Step 3: Cross-domain fusion recommendation based on community themes The community to which the student belongs is identified through the community participation record, and the high-frequency borrowing topics in the community's history are extracted. Then, the associated book tag system is expanded to recommend to adjacent fields. Finally, the activity coefficient is calculated by combining the borrowing frequency in the student portrait and the community contribution.

5. The personalized book recommendation method based on big data analysis according to claim 1, characterized in that: The keyword interaction mechanism in S4 refers to the system supporting students to dynamically manage personalized keywords through the interactive interface, triggering real-time adjustment of the recommendation algorithm weights, and accurately matching demand preferences; at the same time, the system periodically analyzes the frequency and trend of keyword usage, and uses machine learning to dynamically optimize the recommendation model parameters.

6. The personalized book recommendation method based on big data analysis according to claim 1, characterized in that: The S5 student book recommendation portal and review mechanism refers to students submitting book recommendation requests through the system's book recommendation interface, including information such as the title, author, and reason for recommendation; when library administrators review these requests in the background, they will decide whether to adopt students' suggestions based on the library's collection situation, procurement policies, and funding factors.

7. The method for personalized book recommendation based on big data analysis according to claim 1, characterized in that: The specific steps of displaying and providing feedback on the recommendation results in S6 are as follows: Step 1: Visualization of recommendation results After logging into the system, students enter the personalized recommendation area, where a list of books is displayed in the form of cards. Each card contains the title, author, introduction, collection status, and user ratings. Books are sorted by recommendation priority, support filtering by subject classification / popularity, and provide a quick entry for "trial reading chapters" and an extended recommendation module for "similar books"; Interactive buttons are set under each book: Like, Comment and Not Interested. Clicking them will pop up a lightweight feedback form. Step 2: Feedback-driven model iteration The system captures user feedback behavior in real time, injects likes data into the collaborative filtering model as positive samples, and marks as uninterested triggers weight decay of negative samples; Daily incremental training of the recommendation model, adjusting the weights of relevant dimensions in the user feature vector, and optimizing content filtering rules; When the model is fully updated every week, the semantic analysis of the review text is combined to dynamically adjust the book difficulty label matching threshold to form a feedback loop.

8. The method for personalized book recommendation based on big data analysis according to claim 1, characterized in that: The system performance optimization and maintenance in S7 refers to the development team regularly optimizing the database and implementing full-link performance monitoring, combined with load balancing and caching strategies, to ensure stable and efficient operation of the system in high-concurrency scenarios.

9. A personalized book recommendation system based on big data analysis, characterized by: The system includes a multi-source data integration and preprocessing module, a distributed data storage and index optimization module, a hybrid recommendation engine module, a user interaction and feedback closed-loop module, an intelligent model tuning and monitoring module, and a user portrait and behavior analysis module; The multi-source data integration and pre-processing module is used to connect multi-source data in real time, build a star-shaped warehouse after cleaning and standardization, and implement triple quality control verification; The distributed data storage and index optimization module uses MySQL+InnoDB to store structured data, and B+ tree index to accelerate query; MongoDB uses BSON to store unstructured data, and composite index and full-text index support multi-dimensional search; The hybrid recommendation engine module is used to integrate collaborative filtering, content recommendation, and cross-domain recommendation, and dynamically adjust weights to achieve accurate push notifications. The user interaction and feedback closed-loop module is used to submit recommendations at the book recommendation portal, and administrators review and decide whether to adopt them based on the library's collection, policies, and funding. The results are displayed on cards, and feedback drives model optimization. The intelligent model tuning and monitoring module is used to deploy slow query analysis, load balancing and caching, optimize indexes and models, tune thresholds, and ensure high concurrency and stability; The user portrait and behavior analysis module constructs user portraits based on borrowing, class schedule, and community data, and calculates activity to support recommendation optimization.

Images