A Library Book Push Method and System Based on Behavior Analysis

Abstract

The present invention relates to a method and system for pushing library books based on behavior analysis, belonging to the technical field of information processing. Through the borrowing system of the library, the behavior data of readers is collected in real time, and an interest model is constructed based on the reader behavior data to analyze the interest preferences and behavior patterns of readers; clustering analysis is performed according to the analyzed reader interest preferences and behavior patterns to generate clustering portraits of different readers; based on the collected reader behavior data and the local resources of the library, the quality index data of each book is calculated, and a relationship graph between the clustering portraits and the book quality index data is constructed; question interaction is carried out with readers through the Q&A interface of the borrowing system, search keywords in the Q&A are extracted, and a personalized book recommendation list is generated according to the extracted search keywords; the integrated recommendation scores of the books in the book recommendation list are calculated, and the target recommended books are determined according to the integrated recommendation scores, so as to realize personalized book recommendation.

Description

Technical Field

[0001] The present invention relates to the technical field of information processing, and in particular, to a method and system for pushing library books based on behavior analysis. Background Art

[0002] With the continuous progress of information technology, the functions of traditional libraries have undergone great changes. Modern libraries are not only a place for storing and borrowing books, but also an important platform for information acquisition and knowledge exchange. In order to improve the borrowing experience of readers and enhance the service quality of libraries, modern libraries need to provide more refined and personalized services to improve the usage experience and satisfaction of readers.

[0003] Modern libraries not only provide traditional book borrowing services, but also provide various services such as electronic resources, information retrieval, reading promotion, and cultural activities. The diversification of these services requires libraries to have stronger management and service capabilities. With the increasing demand for knowledge and information from readers, the need for personalized services has become increasingly prominent. Different readers have differences in reading interests, information acquisition methods, knowledge fields, etc. How to meet the personalized needs of readers has become an important challenge for libraries. Especially the explosive growth of information volume makes it difficult for readers to quickly find the required resources in the vast amount of information. Libraries need to provide efficient information screening and recommendation services to help readers quickly obtain the required information. Summary of the Invention

[0004] In view of this, the present invention provides a method and system for pushing library books based on behavior analysis, aiming to collect and analyze the behavior data of readers, and combine natural language processing and pushing technology to achieve personalized book recommendations, thereby improving the borrowing experience of readers and the service quality of libraries, and having broad application prospects and practical value.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] In the first aspect, the present invention provides a method for pushing library books based on behavior analysis, including the following steps:

[0007] Collect the behavior data of readers in real time through the borrowing system of the library, and construct an interest model based on the reader behavior data to analyze the interest preferences and behavior patterns of readers;

[0008] Perform clustering analysis according to the analyzed interest preferences and behavior patterns of readers to generate clustering portraits of different readers;

[0009] Based on the collected reader behavior data and the local resources of the library, calculate the quality index data of each book, and construct a relationship map between the clustering portrait and the book quality index data;

[0010] Interact with readers through the Q&A interface of the borrowing system, extract search keywords in the Q&A, and generate a personalized book recommendation list based on the extracted search keywords, combined with the readers' interest models and behavior data;

[0011] Calculate the integrated recommendation scores of each book in the book recommendation list according to the quality index data of the books and the recommendation scores output by the recommendation model, and determine the target recommended books according to the integrated recommendation scores.

[0012] As a further solution of the present invention, the behavior data of readers is collected in real time, including the borrowing behavior data, access behavior data, search behavior data, and payment behavior data of readers; wherein, the borrowing behavior data includes borrowing time, return time, and renewal times; the access behavior data includes arrival time and frequency; the search behavior data includes search keywords and query records; the payment behavior data includes the download records of paid electronic resources.

[0013] As a further solution of the present invention, when constructing an interest model according to the readers' behavior data, the following steps are included:

[0014] Collect behavior data: Collect the borrowing behavior data, access behavior data, search behavior data, and payment behavior data of readers;

[0015] Extract features: Analyze the collected behavior data, extract borrowing behavior features, and extract borrowing keywords and themes from the content of the borrowed books;

[0016] Construct an interest vector: According to the extracted borrowing behavior features and the borrowing keywords and themes of the content of the borrowed books, construct an interest vector for each reader; wherein, each dimension of the interest vector represents a theme, and calculate the interest weights of each reader for different themes and the weights of the borrowing keywords;

[0017] Integrate the interest vectors: Integrate the calculated weights of each theme and borrowing keyword into an interest vector to form a comprehensive interest vector, and use the integrated interest vector to construct the readers' interest model;

[0018] Optimize the interest model: Use matrix factorization to factorize the reader and book rating matrices into feature matrices, mine potential interest features, and optimize the interest model.

[0019] As a further solution of the present invention, the extracted borrowing behavior characteristics include borrowing frequency, borrowing duration, search keywords, and access behavior; among them, the borrowing frequency is calculated by counting the number of times each reader borrows different categories of books within different time periods; the borrowing duration is the counted borrowing duration of each book; the search keywords are the retrieval terms input by the reader into the borrowing system; the access behavior is the counted frequency and duration of the reader's access to the page.

[0020] As a further solution of the present invention, when calculating the interest weight of each reader for different book categories, calculate the weighted sum of the borrowing frequency and borrowing duration of each theme as the interest weight of this theme; the calculation formula of the interest weight is:

[0021] Interest weight 历史 = α × borrowing frequency 历史 + b × borrowing duration 历史

[0022] Among them, α is the borrowing frequency weight, and b is the borrowing duration weight;

[0023] When calculating the weight of the reader's borrowing keywords, calculate the weight of each borrowing keyword in the reader behavior data. Among them, the calculation formula of the borrowing keyword weight is:

[0024] Keyword weight k = T k × I k

[0025] Among them, T k is the frequency of keyword k in the reader behavior data, and I k is the inverse document frequency of keyword k in all reader behavior data.

[0026] As a further solution of the present invention, when generating the clustering portraits of different readers, the following steps are included:

[0027] Standardize the reader's interest preferences and behavior patterns, and perform clustering analysis on the obtained reader feature data to calculate the clustering label of each reader;

[0028] According to the clustering labels, perform feature analysis on each cluster, extract common features, and generate clustering portraits.

[0029] As a further solution of the present invention, the quality index data of each book includes the borrowing times and borrowing and returning times.

[0030] As a further solution of the present invention, when calculating the quality index data of each book, the following steps are included:

[0031] Extract the borrowing times, borrowing and returning times, and user rating data of each book from the library system; among them, the total number of times each book is borrowed is recorded as Bi ; The average borrowing duration of all books is recorded as T i ; The average rating of the books is recorded as S i ;

[0032] The borrowing times, borrowing and returning times, and user rating data are weighted to calculate the quality score Q i , and the calculation formula is:

[0033] Q i = ω1·B i + ω2·T i + ω3·S i

[0034] Among them, ω1, ω2, and ω3 are the weights of borrowing times, borrowing and returning times, and average rating respectively

[0035] As a further solution of the present invention, when constructing the relationship graph of the clustering portrait and the book quality index data, the following steps are included:

[0036] Perform clustering analysis based on the reader's interest model and behavior data to generate the average feature vector of each cluster, representing the portraits of different types of readers;

[0037] Associate each cluster portrait with the book quality index data and draw a relationship graph. In the relationship graph, each node represents a reader cluster or a book, and the weight of the edge represents the degree of interest of the readers in the cluster in the book;

[0038] Use the calculated quality index data Q i and the feature vector in the reader cluster portrait to determine the weight of the edge in the graph through cosine similarity, and obtain the relationship graph of the clustering portrait and the book quality index data

[0039] As a further solution of the present invention, when extracting search keywords and generating a personalized book recommendation list, the following steps are included:

[0040] Analyze the text input by the reader through the Q&A interface of the borrowing system to extract search keywords;

[0041] According to the extracted search keywords, combine the reader's interest model and behavior data to calculate the similarity between the search keywords and the reader's interest vector. The calculation formula is:

[0042]

[0043] Among them, θ is the search keyword vector, and γ is the reader's interest vector;

[0044] Select books that are highly relevant to the search keywords and readers' interests from library resources according to the similarity score, and generate a personalized recommendation list.

[0045] As a further solution of the present invention, when calculating the integrated recommendation score, the following steps are included:

[0046] Calculate the recommendation score R of each book according to the reader's behavior data and interest model i ;

[0047] Integrate the quality index data Q of the book i and the recommendation score R output by the recommendation model i , calculate the integrated recommendation score F of each book i , and the calculation formula is:

[0048] F i = σ·Q i + τR i

[0049] where σ is the weight of the book quality index, and τ is the weight of the recommendation score output by the recommendation model;

[0050] Sort the books according to the integrated recommendation score F i , and select the book with the highest score as the target recommended book.

[0051] In a second aspect, the present invention also provides a library book push system based on behavior analysis, including the following modules:

[0052] A data collection module for collecting readers' behavior data in real time, including: a borrowing system interface connected to the library borrowing system for collecting borrowing data; and a data collector for monitoring and recording readers' click and browsing behaviors in the library system;

[0053] An interest model construction module for constructing an interest model for each reader according to the collected readers' behavior data, and analyzing interest preferences and behavior patterns;

[0054] A clustering analysis module for performing clustering analysis on the readers' interest models to generate clustering portraits of different readers;

[0055] A book quality evaluation module for calculating the quality index data of each book;

[0056] A relationship graph construction module for constructing a relationship graph between the clustering portraits and the book quality index data, and associating the reader groups with the book quality;

[0057] A question-and-answer interaction module for interacting with readers through the question-and-answer interface of the borrowing system, and extracting search keywords in the question-and-answer;

[0058] A list generation module, configured to generate a personalized book recommendation list according to the extracted search keywords, in combination with the reader's interest model and behavior data;

[0059] A fusion recommendation score calculation module, configured to calculate the fusion recommendation scores of each book in the book recommendation list, and determine the target recommended books according to the quality index data of the books and the recommendation scores output by the recommendation model.

[0060] As a further solution of the present invention, the data collection module includes: a borrowing system interface connected to the library borrowing system for collecting borrowing data; and a data collector for monitoring and recording the click and browsing behaviors of readers in the library system.

[0061] Compared with the prior art, the method and system for pushing library books based on behavior analysis of the present invention comprehensively collect and analyze the behavior data of readers, construct a reader interest model, and combine the quality indicators of library books to generate a personalized book recommendation list, and has the following beneficial effects:

[0062] 1. By collecting the behavior data of readers in real time (including borrowing, accessing, searching, and payment behaviors), accurately constructing the reader's interest model, ensuring that the recommended books are more in line with the reader's interests, and improving the reader's reading experience; using multi-dimensional data such as borrowing frequency, borrowing duration, search keywords, and access behaviors to comprehensively analyze the reading habits and interest preferences of readers, ensuring the accuracy and comprehensiveness of the interest model.

[0063] 2. Achieved dynamic adjustment and accurate recommendation: The interest model and behavior pattern of readers are dynamically changing. The present invention can update the behavior data and interest model of readers in real time, ensuring that the recommendation system can be adjusted in time and reflect the latest reader interests, increasing the timeliness and accuracy of recommendations; by performing cluster analysis on the reader interest model, generating cluster portraits of different reader groups, and further refining the recommendation strategy, ensuring that readers within the same group can receive more accurate and common recommendations.

[0064] 3. By calculating the quality index data of each book, providing an objective basis for evaluating the quality of books for the recommendation system. Combining the recommendation scores output by the recommendation model, ensuring that the recommended books not only meet the reader's interests; by constructing a relationship map between the cluster portrait and the book quality index data, intuitively showing the association between the reader group and the books, and using cosine similarity to calculate the edge weights in the map, further optimizing the recommendation results and improving the scientificity and rationality of the recommendation algorithm.

[0065] 4. Intelligent interaction is realized: With the help of the Q&A interaction module, interact with readers through the Q&A interface, extract search keywords, and further combine the reader's interest model and behavior data to generate a personalized book recommendation list, making the recommendation process more intelligent and interactive; through the fusion recommendation score calculation module, comprehensively consider the quality index data of books and the recommendation scores output by the recommendation model to ensure that the finally recommended books not only meet the readers' interests but also have high reading value, improving the satisfaction of the recommended books.

[0066] In summary, the present invention accurately collects and analyzes readers' behavior data, constructs a dynamically updated interest model, and combines the quality indicators of books to generate a personalized recommendation list, greatly improving the accuracy of library book recommendations and user satisfaction, and having significant application value and promotion prospects.

[0067] These aspects or other aspects of the present invention will be more clearly understood in the following description of the embodiments. It should be understood that the above general description and the following detailed description are only exemplary and explanatory, and cannot limit the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or related technologies, the following will briefly introduce the drawings required for the description of the exemplary embodiments or related technologies. The drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation to the present invention. In the drawings:

[0069] Figure 1 It is a flowchart of a method for pushing library books based on behavior analysis according to an embodiment of the present invention.

[0070] Figure 2 It is a flowchart of constructing an interest model in a method for pushing library books based on behavior analysis according to an embodiment of the present invention.

[0071] Figure 3 It is a flowchart of calculating the quality index data of each book in a method for pushing library books based on behavior analysis according to an embodiment of the present invention.

[0072] Figure 4 It is a flowchart of constructing a relationship graph in a method for pushing library books based on behavior analysis according to an embodiment of the present invention.

[0073] Figure 5 It is a flowchart of generating a personalized book recommendation list in a method for pushing library books based on behavior analysis according to an embodiment of the present invention.

[0074] Figure 6It is a flowchart for calculating the integrated recommendation score in a library book push method based on behavior analysis according to an embodiment of the present invention. Detailed implementation manners

[0075] In order to make the objectives, technical solutions and advantages of the present invention clearer, the present 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 only used to explain the present invention, but not to limit the present invention.

[0076] In some processes described in the specification, claims and above-mentioned drawings of the present invention, a plurality of operations appear in a specific order. However, it should be clearly understood that these operations may not be executed in the order in which they appear herein or may be executed in parallel. The serial numbers of the operations, such as 101, 102, etc., are only used to distinguish different operations, and the serial numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions such as "first" and "second" in this article are used to distinguish different messages, devices, modules, etc., and do not represent a sequence, nor do they limit that "first" and "second" are of different types.

[0077] Next, the technical solutions in the exemplary embodiments of the present invention will be clearly and completely described with reference to the accompanying drawings in the exemplary embodiments of the present invention. Obviously, the described exemplary embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative efforts fall within the protection scope of the present invention.

[0078] Due to the explosive growth of information volume, it is difficult for readers to quickly find the required resources in the vast amount of information. Libraries need to provide efficient information screening and recommendation services to help readers quickly obtain the required information. In view of this, the present invention provides a library book push method and system based on behavior analysis, aiming to collect and analyze readers' behavior data, and combine natural language processing and push technology to achieve personalized book recommendation, thereby improving readers' borrowing experience and the service quality of libraries, and having broad application prospects and practical values.

[0079] The technical solutions of the present invention will be further described below with reference to specific embodiments:

[0080] See Figure 1 As shown, a library book push method based on behavior analysis provided by an embodiment of the present invention includes the following steps:

[0081] Step S10: Real-time collect readers' behavior data through the library borrowing system, construct an interest model based on the readers' behavior data, and analyze the readers' interest preferences and behavior patterns;

[0082] Step S20: Conduct clustering analysis based on the analyzed readers' interest preferences and behavior patterns to generate clustering portraits of different readers;

[0083] Step S30: Based on the collected readers' behavior data and the local resources of the library, calculate the quality index data of each book, and construct a relationship map between the clustering portraits and the quality index data of the books; among them, the quality index data of each book includes the borrowing times and the borrowing and returning times;

[0084] Step S40: Conduct question interaction with readers through the Q&A interface of the borrowing system, extract the search keywords in the Q&A, and generate a personalized book recommendation list according to the extracted search keywords, combined with the readers' interest model and behavior data;

[0085] Step S50: Calculate the integrated recommendation scores of the books in the book recommendation list according to the quality index data of the books and the recommendation scores output by the recommendation model, and determine the target recommended books according to the integrated recommendation scores.

[0086] In this embodiment, in step S10, the real-time collection of readers' behavior data includes readers' borrowing behavior data, access behavior data, search behavior data, and payment behavior data; among them, the borrowing behavior data includes borrowing time, returning time, and renewal times; the access behavior data includes arrival time and frequency; the search behavior data includes search keywords and query records; the payment behavior data includes the download records of paid electronic resources.

[0087] The embodiment of the present invention provides a method for pushing library books based on behavior analysis. By real-time collecting various behavior data of readers, constructing an interest model and behavior patterns, conducting clustering analysis to generate reader portraits, and constructing a relationship map based on the quality index data of books. In the interaction with readers, the system extracts search keywords, combines with the readers' interest model to generate a personalized recommendation list, and finally determines the target recommended books through the integrated recommendation scores. This method not only improves the accuracy of recommendations and user satisfaction, but also improves the utilization efficiency of library resources, and has significant application value and promotion prospects.

[0088] When the library book push method based on behavior analysis is executed, first, the behavior data of readers is collected in real time. Through the library borrowing system, the borrowing time, return time, and renewal times of the borrowing behavior data of each reader are recorded in real time to reflect the usage frequency and duration of different books by readers; and access behavior data, search behavior data, and payment behavior data are collected, respectively recording the time and frequency of readers' visits to the library, the search keywords and query records of readers in the library system, and the records of readers' downloads of paid electronic resources.

[0089] If the behavior data of three readers is collected through the library book push method based on behavior analysis in the embodiments of the present invention, assuming that the behavior data of readers collected in real time is as follows:

[0090] (1) Reader A borrowed 5 books in a certain month, with an average borrowing time of 15 days for each book, and also included 2 renewal behaviors.

[0091] (2) Reader B visited the library 3 times in a certain week, and each visit lasted about 2 hours.

[0092] (3) Reader C searched for "artificial intelligence technology" and "machine learning algorithms" many times in the library system and downloaded several related paid electronic resources.

[0093] Then, when performing cluster analysis to generate the cluster portraits of different readers, according to the collected behavior data, the readers are classified using a clustering algorithm to form the cluster portraits of different interest groups. Through analysis, it is found that Reader A and several other readers belong to the "science and technology book lovers" group because they frequently borrow and search for science and technology books. Reader B is classified into the "literature book lovers" group because his borrowing and visit records show a high interest in literary works.

[0094] Then, when calculating the book quality index data and constructing the relationship graph, the borrowing times and borrowing and returning times of each book are calculated as its quality index data; a relationship graph between the reader cluster portraits and the book quality index data is constructed to improve the accuracy of recommendations through these associated data. Then, the relationship graph that can be constructed is as follows:

[0095] ① Book X was borrowed 30 times in the past year, with an average borrowing time of 20 days.

[0096] ② The borrowing times of Book Y are 10 times, and the average borrowing time is 15 days.

[0097] Through the relationship graph, it can be found that among the "science and technology book lovers" group, the borrowing times and borrowing time of Book X are both relatively high, indicating that this book is popular among this group.

[0098] When interacting with readers through the Q&A interface and generating a personalized recommendation list, in the Q&A interface of the borrowing system, readers input the keywords they are interested in. The system extracts these keywords and combines them with the aforementioned interest model and behavioral data to generate a personalized recommendation list. For example, reader D inputs "blockchain technology" in the Q&A interface. Combining their past borrowing and search records, the system recommends several relevant books, such as *Blockchain Revolution*, *Detailed Explanation of Blockchain Technology*, etc.

[0099] When calculating the integrated recommendation score and determining the target recommended book, according to the quality index data of the book and the recommendation score output by the recommendation model, calculate the integrated recommendation score for each book, and select the book with the highest score as the target recommended book. Finally, through calculation, the book *Blockchain Revolution* has the highest integrated recommendation score, and the system finally recommends this book to reader D.

[0100] In step S10 of this embodiment, as shown in Figure 2 When constructing an interest model based on reader behavior data, the following steps are included:

[0101] Step S101, collect behavioral data: Collect readers' borrowing behavior data, access behavior data, search behavior data, and payment behavior data;

[0102] Step S102, extract features: Analyze the collected behavioral data, extract borrowing behavior features, and extract borrowing keywords and themes from the content of the borrowed books;

[0103] Step S103, construct an interest vector: According to the extracted borrowing behavior features and the borrowing keywords and themes of the content of the borrowed books, construct an interest vector for each reader; where each dimension of the interest vector represents a theme, and calculate the interest weight of each reader for different themes and the weight of the borrowing keywords.

[0104] Step S104, integrate the interest vector: Integrate the calculated weights of each theme and borrowing keywords into an interest vector to form a comprehensive interest vector, and use the integrated interest vector to construct a reader's interest model;

[0105] Step S105, optimize the interest model: Use matrix factorization to factorize the reader and book rating matrix into a feature matrix, mine potential interest features, and optimize the interest model.

[0106] In this embodiment, through the borrowing system of the library, the following reader behavior data is collected in real time:

[0107] Borrowing behavior data: including borrowing time, return time, renewal times, categories of borrowed books, borrowing frequency, and borrowing duration.

[0108] Access behavior data: including arrival time, access frequency, accessed pages, and access duration.

[0109] Search behavior data: including search keywords and query records.

[0110] Payment behavior data: including download records of paid e-resources.

[0111] The library's borrowing system collects the above-mentioned behavior data through log records and database queries, and analyzes the collected behavior data. Among them, the extracted borrowing behavior characteristics include borrowing frequency, borrowing duration, search keywords, and access behavior; among them, the borrowing frequency is calculated by counting the number of times each reader borrows different categories of books within different time periods; the borrowing duration is the statistically calculated borrowing duration of each book; the search keywords are the retrieval terms input by the reader into the borrowing system; the access behavior is the frequency and access duration of the reader's accessed pages counted.

[0112] In step S103 of this embodiment, when calculating the interest weight of each reader for different book categories, calculate the weighted sum of the borrowing frequency and borrowing duration of each theme as the interest weight of this theme; the calculation formula for the interest weight is:

[0113] Interest weight 历史 = α × borrowing frequency 历史 + b × borrowing duration 历史

[0114] Among them, α is the borrowing frequency weight, and b is the borrowing duration weight;

[0115] When calculating the weight of the reader's borrowing keywords, calculate the weight of each borrowing keyword in the reader's behavior data. Among them, the calculation formula for the borrowing keyword weight is:

[0116] Keyword weight k = T k × I k

[0117] Among them, T k is the frequency of keyword k in the reader's behavior data, and I k is the inverse document frequency of keyword k in all readers' behavior data.

[0118] In step S20 of this embodiment, when generating the clustering portraits of different readers, the following steps are included:

[0119] Standardize the readers' interest preferences and behavior patterns, and perform clustering analysis on the obtained reader feature data to calculate the clustering label of each reader;

[0120] According to the clustering labels, perform feature analysis on each cluster, extract common features, and generate clustering portraits.

[0121] In step S30 of this embodiment, as shown in Figure 3 the following steps are included when calculating the quality index data of each book:

[0122] Step S301: Extract the borrowing times, borrowing and returning times, and user rating data of each book from the library system; among them, the total number of times each book is borrowed is recorded as B i ; the average value of all borrowing durations of the book is recorded as T i ; the average rating of the book is recorded as S i ;

[0123] Step S302: Perform weighted calculation on the borrowing times, borrowing and returning times, and user rating data to obtain the quality score Q i , and the calculation formula is:

[0124] Q i = ω1·B i + ω2·T i + ω3·S i

[0125] where ω1, ω2, and ω3 are the borrowing times weight, borrowing and returning times weight, and average rating weight respectively.

[0126] In step S30 of this embodiment, as shown in Figure 4 the following steps are included when constructing the relationship graph between the clustering portrait and the book quality index data:

[0127] Step S311: Perform clustering analysis based on the reader's interest model and behavior data to generate the average feature vector of each cluster, representing the portraits of different types of readers;

[0128] Step S312: Associate each clustering portrait with the book quality index data and draw the relationship graph. In the relationship graph, each node represents a reader cluster or a book, and the weight of the edge represents the degree of interest of the readers in the cluster in the book;

[0129] Step S313: Use the calculated quality index data Q i and the feature vector in the reader clustering portrait to determine the weight of the edge in the graph through cosine similarity, and obtain the relationship graph between the clustering portrait and the book quality index data.

[0130] Generate reader portraits through cluster analysis and calculation of quality metrics, and establish a relationship graph with book data to better understand readers' interest preferences and optimize the library's book recommendation system. When generating cluster portraits of different readers and calculating quality metric data for each book, following the above steps, if readers' preferences, borrowing frequencies, ratings, and data on the borrowing times, borrowing and returning times, and user ratings of each book extracted from the library system are as follows:

[0131] Reader A: Prefers fictional books, has a high borrowing frequency, and gives positive ratings.

[0132] Reader B: Prefers science and technology books, has a low borrowing frequency, and gives strict ratings.

[0133] Book X: Fictional, borrowed 50 times, average borrowing duration of 7 days, average rating of 4.5.

[0134] Book Y: Science and technology, borrowed 30 times, average borrowing duration of 14 days, average rating of 3.8.

[0135] Then, when generating cluster portraits of different readers, standardize the interest preferences and behavior pattern data of Readers A and B, use K-means clustering to cluster Readers A and B into two different categories respectively, extract features, and generate two reader portraits (A is a high-frequency fictional book lover, and B is a low-frequency science and technology lover).

[0136] Then, when calculating the quality metric data for each book, extract the borrowing times, average duration, and ratings of Books X and Y; if the weights are set as: ω1 = 0.5, ω2 = 0.3, ω3 = 0.2; calculate the quality score Q i as:

[0137] Book X: Q X = 0.5·50 + 0.3·7 + 0.2·4.5 ≈ 27.4;

[0138] Book Y: Q Y = 0.5·30 + 0.3·14 + 0.2·3.8 ≈ 19.76.

[0139] Through the above steps, accurate reader portraits can be generated and the quality metrics of books can be calculated. Using these data to construct a relationship graph realizes a comprehensive analysis of readers' interests and book quality. The method of the present invention can not only help the library optimize the book recommendation system, but also provide more personalized reading suggestions for readers, improving the borrowing rate and reader satisfaction.

[0140] In step S40 of this embodiment, as shown in Figure 5 when extracting search keywords and generating a personalized book recommendation list, the following steps are included:

[0141] Step S401: Analyze the text input by the reader through the Q&A interface of the borrowing system, and extract search keywords. Specifically, the system first receives the text input by the reader through the Q&A interface of the borrowing system, uses NLP to parse the input text, and extracts keywords related to book search. These keywords can be topic words, book titles, author names, or other relevant vocabulary.

[0142] Step S402: According to the extracted search keywords, combined with the reader's interest model and behavior data, calculate the similarity between the search keywords and the reader's interest vector. The calculation formula is:

[0143]

[0144] where θ is the search keyword vector and γ is the reader's interest vector.

[0145] When calculating the similarity between the search keywords and the reader's interest vector, the system converts the extracted search keywords into vector representation, denoted as θ. At the same time, the system calls the reader's interest model from the database to obtain the reader's interest vector γ, and calculates the similarity between the search keyword vector and the reader's interest vector through the above similarity calculation formula.

[0146] Step S403: According to the similarity score, select books highly relevant to the search keywords and the reader's interest from the library resources, and generate a personalized recommendation list.

[0147] In step S50 of this embodiment, as shown in Figure 6 When calculating the integrated recommendation score, the following steps are included:

[0148] Step S501: Calculate the recommendation score R of each book according to the reader's behavior data and interest model i ;

[0149] Step S502: Integrate the quality index data Q of the book i and the recommendation score E output by the recommendation model i , and calculate the integrated recommendation score F of each book i . The calculation formula is:

[0150] F i =σ·Q i +τR i

[0151] where σ is the weight of the book quality index and τ is the weight of the recommendation score output by the recommendation model;

[0152] Step S503: Sort the books according to the integrated recommendation score F i and select the book with the highest score as the target recommended book.

[0153] Exemplarily, if the text input by a reader in the Q&A interface of the borrowing system is: "I'm very interested in science fiction recently. Can you recommend some?", according to the above steps, the system will perform the following operations:

[0154] (1) Keyword extraction: The system extracts the keyword "science fiction" from the input text through NLP.

[0155] (2) Calculate similarity: Convert "science fiction" into a vector θ; extract the reader's interest vector γ from the database (assuming the reader's interests include science fiction); then calculate the similarity, and assume the result is 0.85.

[0156] (3) Generate a recommendation list: The system screens out books related to "science fiction" from the library resources and generates a preliminary recommendation list in combination with the similarity score.

[0157] (4) Integrate recommendation scores: First, according to the reader's behavior data and interest model, the system calculates the recommendation score for each book. Then, by synthesizing the quality indicators of each book (such as high-rated science fiction novels) and the recommendation scores, the integrated recommendation score is calculated. Assume that the quality indicator score of a certain book is 0.9, the recommendation score is 0.8, and the weights are σ = 0.6 and τ = 0.4. Then the integrated recommendation score F = 0.6·0.9 + 0.4·0.8 = 0.86. Then, the books can be sorted according to the integrated recommendation score, and the books with the highest scores are selected as the final recommendations.

[0158] This embodiment realizes personalized book recommendations based on reader input, interest models, and behavior data. The system not only considers keyword matching but also comprehensively takes into account the reader's interests and the quality of the books, ensuring that the recommended books not only meet the reader's interests but also have high quality, effectively improving the user's reading experience and satisfaction.

[0159] Through the above-mentioned manner, the library book push method based on behavior analysis of the present invention can comprehensively collect readers' borrowing behavior, access behavior, search behavior and payment behavior data, extract borrowing frequency, borrowing time, search keywords, access behavior and other features from the collected data, and extract borrowing keywords and themes from the book content, based on the extracted features, calculate the weight of each theme and keyword, form the reader's interest vector, integrate the weights of each theme and keyword into a comprehensive interest vector, and optimize the interest model through a matrix decomposition method, cluster readers according to the interest model, generate cluster portraits, calculate book quality indicators, and construct a relationship map between cluster portraits and book quality indicators, so as to facilitate interaction with readers through a question-and-answer interface, extract search keywords, generate a personalized recommendation list in combination with interest models and behavior data, integrate book quality indicator data and recommendation model scores, calculate fusion recommendation scores, and determine recommended books. Not only does it improve the accuracy of recommendations, but it can also dynamically adapt to changes in readers' interests and provide personalized and high-quality book recommendation services.

[0160] It should be understood that, although described in a certain order, these steps are not necessarily performed in sequence in the above order. Unless there is clear explanation in this article, the execution of these steps does not have strict order restriction, and these steps can be performed in other orders. Moreover, a part of the steps of the present embodiment may include a plurality of steps or a plurality of stages, and these steps or stages are not necessarily performed at the same time, but can be performed at different times, and the execution order of these steps or stages is not necessarily performed in sequence, but can be performed in turn or alternately with at least a part of the steps or stages in other steps or other steps.

[0161] In an embodiment, the present invention also provides a library book push system based on behavior analysis, which aims to build a personalized book recommendation mechanism by collecting and analyzing reader behavior data. The system includes the following components:

[0162] The data collection module is used to collect readers' behavior data in real time, including: a borrowing system interface connected to the library's borrowing system for collecting borrowing data; and a data collector for monitoring and recording readers' clicks and browsing behaviors in the library system; wherein, the borrowing system interface collects readers' borrowing records including borrowing time, book type, borrowing frequency and other data; the data collector is used to obtain detailed behavior data such as visited pages, browsing time, number of clicks, etc. During data collection, a reader logged into the library system, borrowed several history books and browsed several articles on world history in the system. The data collection module can record this information in real time and transmit it to the system's central database;

[0163] An interest model construction module, which is used to construct an interest model for each reader based on the collected reader behavior data, and analyze the interest preferences and behavior patterns;

[0164] A clustering analysis module, which is used to perform clustering analysis on the interest models of readers, generate clustering portraits of different readers, be able to identify reader groups with similar interest preferences, and describe the common characteristics of each reader group through the generated clustering portraits;

[0165] A book quality evaluation module, which is used to calculate the quality index data of each book, including user ratings, borrowing times, number of comments, etc., and comprehensively evaluate the overall quality of the book based on various indicators; For example: A history book "A Brief History of the World" has a high rating among readers and also has a large number of borrowing times, and the system calculates high-quality index data for it.

[0166] A relationship graph construction module, which is used to construct a relationship graph between the clustering portraits and the book quality index data, and associate the reader groups with the book quality; Exemplarily, the system associates the history enthusiast group with the high-quality history book "A Brief History of the World" in the relationship graph.

[0167] A question-and-answer interaction module, which interacts with readers through the question-and-answer interface of the borrowing system, extracts the search keywords in the question and answer, and further understands the immediate interests and needs of readers; Suppose a reader asks "Recommend books about World War II" on the question-and-answer interface, and the system can extract the keyword "World War II".

[0168] A list generation module, which is used to generate a personalized book recommendation list according to the extracted search keywords, combined with the reader's interest model and behavior data, and give priority to recommending high-quality books related to the reader's current interests; For example, the system recommends high-quality books about World War II, such as "The Complete History of World War II", according to the keyword "World War II" and the reader's interest model.

[0169] A fusion recommendation score calculation module, which is used to calculate the fusion recommendation scores of each book in the book recommendation list, and determine the target recommended books according to the quality index data of the books and the recommendation scores output by the recommendation model. Suppose the final target recommended book is "The Complete History of World War II". "The Complete History of World War II" not only meets the current search needs of reader A (keyword "World War II"), but also has high quality indicators, and is finally determined as a recommended book and pushed to the reader.

[0170] The library book recommendation system based on behavior analysis in this embodiment constructs a personalized interest model based on the accurate analysis of readers' behavior data, realizes personalized book recommendations for each reader, and through the Q&A interaction module, the system can obtain the immediate needs of readers in real time and provide more accurate recommendations. Combining with book quality evaluation, the system ensures that the recommended books not only meet the readers' interests but also have high quality. Using the clustering analysis module, the system can identify groups of readers with similar interests and provide support for group recommendations. Through this library book recommendation system, the library can better meet the personalized needs of readers, improve the reading experience and satisfaction of readers, and thus improve the utilization efficiency of library resources.

[0171] In this embodiment, the library book recommendation system based on behavior analysis adopts the steps of a library book recommendation method based on behavior analysis as described above when executing. Therefore, the operation process of the library book recommendation system based on behavior analysis in this embodiment will not be introduced in detail.

[0172] In an embodiment of the present invention, there is also provided a computer device, including at least one processor, and a memory communicatively connected to the at least one processor. The memory stores instructions executable by the at least one processor. The instructions are executed by the at least one processor to cause the at least one processor to execute the library book recommendation method based on behavior analysis, and when the processor executes the instructions, it implements the steps in the above-mentioned embodiment of the library book recommendation method based on behavior analysis.

[0173] In an embodiment of the present invention, there is provided a computer-readable storage medium storing computer instructions for causing the computer to execute the steps of the library book recommendation method based on behavior analysis.

[0174] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by a computer program represented by computer instructions instructing relevant hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the above-mentioned method embodiments. Among them, any reference to a memory, storage, database, or other medium used in the various embodiments provided in this application can include at least one of non-volatile and volatile memories.

[0175] Non-volatile memory may include read-only memory, magnetic tape, floppy disk, flash memory, or optical memory, etc. Volatile memory may include random access memory or external cache memory. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory or dynamic random access memory, etc.

[0176] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention shall fall within the protection scope of the present invention.

Claims

1. A library book push method based on behavior analysis, characterized in that: The method comprises the following steps: Through the library's lending system, readers' behavior data is collected in real time, and interest models are built based on the readers' behavior data to analyze readers' interest preferences and behavior patterns; Perform cluster analysis based on the readers' interest preferences and behavior patterns to generate cluster portraits of different readers; Based on the collected reader behavior data and the library's local resources, the quality index data of each book is calculated, and a relationship map between clustering portraits and book quality index data is constructed; Interact with readers through the Q&A interface of the lending system, extract search keywords from the Q&A, and generate a personalized book recommendation list based on the extracted search keywords and combined with the readers' interest model and behavior data; According to the quality index data of the books and the recommendation scores output by the recommendation model, the fusion recommendation scores of each book in the book recommendation list are calculated, and the target recommended books are determined according to the fusion recommendation scores; Among them, when constructing an interest model based on reader behavior data, the following steps are included: Collect readers' borrowing behavior data, access behavior data, search behavior data, and payment behavior data; Analyze the collected behavioral data, extract borrowing behavior characteristics, and extract borrowing keywords and themes from the content of borrowed books; Based on the extracted borrowing behavior characteristics and the borrowing keywords and themes of the borrowed book content, an interest vector for each reader is constructed; each dimension of the interest vector represents a theme, and the interest weights of each reader on different themes and the weights of the borrowing keywords are calculated; Integrate the calculated weights of each topic and borrowing keyword into an interest vector to form a comprehensive interest vector, and use the integrated interest vector to build a reader's interest model; Use matrix decomposition to decompose reader and book rating matrices into feature matrices, mine potential interest features, and optimize interest models; When calculating the interest weight of each reader for different book categories, the weighted sum of borrowing frequency and borrowing duration of each subject is calculated as the interest weight of the subject; the calculation formula of interest weight is: Interest Weight 历史 =α×borrowing frequency 历史 +b×borrowing time 历史 Among them, α is the borrowing frequency weight, and b is the borrowing duration weight; When calculating the weight of a reader's borrowing keyword, the weight of each borrowing keyword in the reader's behavior data is calculated. The calculation formula for the borrowing keyword weight is: Keyword weight k =T k ×I k Among them, T k is the frequency of keyword k in the reader behavior data, I k is the inverse document frequency of keyword k in all reader behavior data; The calculation of the quality index data of each book includes the following steps: Extract the borrowing times, borrowing and returning time and user rating data of each book from the library system; the total number of times each book is borrowed is recorded as B i ; The average of all borrowing time of books is recorded as T i ; The average rating of the book is recorded as S i ; The quality score Q is obtained by weighting the number of borrowing times, borrowing and returning time, and user rating data. i , the calculation formula is: Q i =ω1·B i +ω2·T i +ω3·S i Among them, ω1, ω2 and ω3 are the weight of borrowing times, the weight of borrowing and returning time and the weight of the average score respectively; When constructing the relationship map between clustering portraits and book quality index data, the following steps are included: Perform cluster analysis based on readers’ interest models and behavior data to generate the average feature vector of each cluster, representing the portraits of different types of readers; Associate each cluster portrait with the book quality index data and draw a relationship map. In the relationship map, each node represents a reader cluster or a book, and the edge weight represents the degree of interest of the readers in the cluster in the book. Using the calculated quality index data Q i The characteristic vectors in the reader clustering portrait are used to determine the weights of the edges in the graph through cosine similarity, and the relationship graph between the clustering portrait and the book quality index data is obtained.

2. The method for pushing books to a library based on behavior analysis as claimed in claim 1, characterized in that: The real-time collection of readers' behavior data includes readers' borrowing behavior data, access behavior data, search behavior data and payment behavior data; wherein the borrowing behavior data includes borrowing time, return time and renewal times; the access behavior data includes visit time and frequency; the search behavior data includes search keywords and query records; and the payment behavior data includes download records of paid electronic resources.

3. The method for pushing books to a library based on behavior analysis as claimed in claim 2, characterized in that: The extracted borrowing behavior features include borrowing frequency, borrowing duration, search keywords and access behavior; wherein, the borrowing frequency is calculated by calculating the number of times each reader borrows books of different categories in different time periods; the borrowing duration is the statistical borrowing duration of each book; the search keywords are the search terms entered by readers into the borrowing system; and the access behavior is the statistical frequency and access duration of readers visiting the page.

4. The method for pushing books to a library based on behavior analysis as claimed in claim 3, characterized in that: When extracting search keywords and generating a personalized book recommendation list, the following steps are included: Through the question-and-answer interface of the borrowing system, the text entered by the reader is analyzed to extract the search keywords; Based on the extracted search keywords, combined with the reader's interest model and behavior data, the similarity between the search keywords and the reader's interest vector is calculated. The calculation formula is: Among them, θ is the search keyword vector, γ is the reader's interest vector; Based on the similarity score, books that are highly relevant to the search keywords and readers' interests are selected from the library resources to generate a personalized recommendation list.

5. The method for pushing books to a library based on behavior analysis as claimed in claim 4, characterized in that: The calculation of the fusion recommendation score includes the following steps: Calculate the recommendation score R of each book based on the reader's behavior data and interest model i ; Quality index data Q of comprehensive books i And the recommendation score R output by the recommendation model i , calculate the fusion recommendation score F of each book i , the calculation formula is: F i =σ·Q i +τR i Among them, σ is the weight of the book quality index, and τ is the weight of the recommendation model output score; According to the fusion recommendation score F i Sort the books and select the books with the highest scores as the target recommended books.

6. A library book push system based on behavior analysis, characterized in that: The system is used to execute the library book push method based on behavior analysis as described in any one of claims 1 to 5, comprising: The data collection module is used to collect the readers' behavior data in real time, including: a borrowing system interface connected to the library's borrowing system for collecting borrowing data; and a data collector for monitoring and recording the readers' click and browsing behaviors in the library system; The interest model building module is used to build an interest model for each reader based on the collected reader behavior data, and analyze interest preferences and behavior patterns; Cluster analysis module, used to perform cluster analysis on readers' interest models and generate cluster portraits of different readers; Book quality assessment module, used to calculate the quality index data of each book; The relationship map construction module is used to construct a relationship map between clustering portraits and book quality index data, and associate reader groups with book quality; The question-and-answer interaction module interacts with readers through the question-and-answer interface of the borrowing system and extracts search keywords in the questions and answers; The list generation module is used to generate a personalized book recommendation list based on the extracted search keywords and combined with the reader's interest model and behavior data; The fusion recommendation score calculation module is used to calculate the fusion recommendation score of each book in the book recommendation list, and determine the target recommended book based on the book's quality indicator data and the recommendation score output by the recommendation model.

Images