Book recommendation method and device for colleges and universities and storage medium

Abstract

The invention discloses a book recommendation method and device for colleges and universities and a storage medium, and belongs to the technical field of education management, and the method comprises the following steps: obtaining information of students before entering a school, inputting the information into a pre-trained machine learning model, and obtaining initial book interest vectors of the students; generating a cumulative borrowing vector of students in school, and obtaining a cumulative borrowing vector of graduated students; similarity comparison is carried out, and a plurality of graduated students with the highest similarity are determined; performing weighted summation on the corresponding actual book interest vectors to obtain predicted book interest vectors of the school students after borrowing; performing weighted summation on the initial book interest vector and the predicted book interest vector to obtain a real-time book interest vector; and recommending the books to the students. According to the invention, book recommendation can be carried out based on the school book borrowing data sparsity condition, the characteristics of a college book system are met, and the real-time recommendation result is accurate.

Description

Technical Field

[0001] The invention relates to a book recommendation method, device and storage medium for colleges and universities, and belongs to the technical field. Background Art

[0002] University libraries have rich collection resources, providing students with the opportunity to obtain the information they need. However, due to the huge resources, it is difficult for students to accurately and effectively find the books they are interested in. Therefore, as an effective way to solve the problem of information overload, the recommendation system deserves extensive research in university libraries.

[0003] There are many types of existing recommendation algorithms, including content-based recommendation, collaborative filtering recommendation, knowledge-based recommendation and hybrid recommendation. Content-based recommendation mainly predicts the items that users may be interested in in the future based on their past preferences. Collaborative filtering makes recommendations based on the historical behaviors of other similar users. Knowledge-based recommendation mainly matches user needs through user personal information and item information. Hybrid recommendation is a method that comprehensively applies multiple recommendation strategies to achieve better recommendation results. In addition, deep learning is gradually becoming an important research direction in the field of recommendation systems.

[0004] However, the library system of colleges and universities is different from that of commercial libraries. The borrowing data of school books is obviously sparse (the borrowing volume is too small relative to the storage volume), because compared with the collection volume, most students only borrow a small number of books each year, and each student only stays in school for a few years. Therefore, the existing commercial library system recommendation method is not applicable to the library system recommendation method of colleges and universities. The accuracy of the recommendation results obtained by the method is not accurate. There is an urgent need for a book recommendation method that can meet the characteristics of the university library system. Summary of the invention

[0005] The technical problem to be solved by the present invention is to overcome the defects of the prior art and provide a book recommendation method, device and storage medium for colleges and universities, which can make book recommendations based on the sparsity of school book borrowing data, conform to the characteristics of university library systems, and provide accurate real-time recommendation results.

[0006] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0007] In a first aspect, the present invention provides a book recommendation method for a university, comprising the following steps:

[0008] Obtain the pre-enrollment information of the students, and then input the pre-enrollment information into the pre-trained machine learning model to obtain the initial book interest vector of the students;

[0009] After each borrowing of a student, a cumulative borrowing vector of the student is generated, and historical borrowing data of graduated students is obtained to obtain the cumulative borrowing vector of the graduated students;

[0010] Compare the cumulative borrowing vectors of current students with the cumulative borrowing vectors of graduated students to determine a number of graduated students with the highest similarity;

[0011] The actual book interest vectors corresponding to several graduated students with the highest similarity are weighted summed to obtain the predicted book interest vectors of the students in school after borrowing;

[0012] Performing a weighted summation of the initial book interest vector of the students and the predicted book interest vector of the students after borrowing, to obtain the real-time book interest vector of the students;

[0013] Recommend books to students based on their real-time book interest vectors.

[0014] The training of the machine learning model includes the following steps:

[0015] Get the historical borrowing data of graduated students and obtain the book borrowing vector of graduated students, which is specifically expressed as: ,

[0016] in, represents the book borrowing vector of the mth graduated student, represents the borrowing volume of the nth category of books by the mth graduated student, and c is the total number of book categories;

[0017] Normalize the book borrowing vector of graduated students to obtain the actual book interest vector of graduated students, which is specifically expressed as: ,

[0018] in, represents the actual book interest vector of the mth graduated student, Indicates the interest of the mth graduated student in the nth category of books;

[0019] The machine learning model is trained using the pre-admission information of graduated students as input and the actual book interest vectors corresponding to the graduated students as output.

[0020] The initial book interest vector includes the student's interest information for books of various categories, which is specifically expressed as: ,

[0021] in, represents the initial book interest vector of the i-th student in school, Indicates the interest of the i-th student in the j-th category of books.

[0022] The cumulative borrowing vector of students in school is specifically expressed as: w i e = [ x i1 e , x i2 e … x ij e … x ic e ] ;

[0023] in, is the cumulative borrowing vector of the i-th student after the e-th borrowing, It is the cumulative borrowing quantity of the j-th category book after the e-th borrowing by the i-th student in school.

[0024] The cumulative borrowing vector of graduated students after borrowing is specifically expressed as: v m e =[ y m1 e , y m2 e … y mn e … y mc e ]

[0025] in, is the cumulative borrowing vector of the mth graduated student after the eth borrowing, It indicates the borrowing amount of books in category n by the mth graduated student after accumulating e borrowing times. If the total number of borrowing times of the mth graduated student does not reach e times, then = .

[0026] The predicted book interest vector after borrowing by students in school is specifically expressed as: u i e =[ z i1 e , z i2 e … z ij e … z ic e ] ,

[0027] in, represents the predicted book interest vector of the i-th student after the e-th borrowing. It represents the predicted interest of the i-th student in the j-th category of books after borrowing them for the e-th time;

[0028] The real-time book interest vector of students in school is specifically expressed as:

[0029]

[0030] in, represents the real-time book interest vector of the i-th student in school, represents the real-time interest of the i-th student in the j-th category of books, is the weight parameter, and λ ∈ [0,1] .

[0031] Weight Parameters The value of the weight is not fixed. As the number of borrowings and the time spent in school increase, The value of the number of books is increasing, and students do not borrow books. The value is fixed to 0.

[0032] The pre-admission information includes one or more of the scores of each subject, region, and the major to be studied; the specific encoding method of the pre-admission information is the scores of each subject in the college entrance examination, and the corresponding subject ranking is divided by the total number of students in the corresponding subject in the region. The region is encoded using the geographic location, and the major to be studied uses an embedded vector.

[0033] According to the real-time book interest vectors of students in school, the specific steps for recommending books to students are as follows: find the g book categories with the largest real-time interest in the real-time book interest vectors, and recommend a corresponding number of the latest or most popular books in the corresponding book categories according to the real-time interest of these book categories, where g is a preset parameter.

[0034] In a second aspect, the present invention provides a book recommendation device for colleges and universities, comprising:

[0035] The initial interest generation module is used to obtain the pre-enrollment information of the students, and then input the pre-enrollment information into the pre-trained machine learning model to obtain the initial book interest vector of the students;

[0036] The cumulative borrowing generation module is used to generate the cumulative borrowing vector of the students in school after each borrowing of the students in school, and obtain the historical borrowing data of the graduated students to obtain the cumulative borrowing vector of the graduated students;

[0037] The highest similarity determination module is used to compare the cumulative borrowing vectors of current students with the cumulative borrowing vectors of graduated students after each borrowing to determine a number of graduated students with the highest similarity;

[0038] The predicted interest generation module is used to perform weighted summation of actual book interest vectors corresponding to several graduated students with the highest similarity, and obtain the predicted book interest vector after borrowing by the students in school;

[0039] A real-time interest generation module is used to perform a weighted summation of the initial book interest vector of the students and the predicted book interest vector after the students borrow the books, so as to obtain the real-time book interest vector of the students;

[0040] The book recommendation module is used to recommend books to students based on their real-time book interest vectors.

[0041] In a third aspect, the present invention provides a computer-readable storage medium having a computer program / instruction stored thereon, which, when executed by a processor, implements the steps of the method for recommending books for colleges and universities.

[0042] Beneficial effects of the present invention: The present invention provides a book recommendation method, device and storage medium for colleges and universities, using the borrowing amount of each category of graduated students as a measure of the students' interest in books of the corresponding category. At the same time, the characteristic that colleges and universities can obtain relatively rich student admission information is utilized. According to the characteristic that colleges and universities can obtain more pre-admission information, this information is reasonably utilized, and a machine learning algorithm is used to automatically learn the relationship between this information and students' book interests. In addition, the book borrowing behavior of students in the past few years in colleges and universities is the information most closely related to students' book interests. Therefore, the present invention draws on the idea of ​​collaborative filtering algorithm, and finds students with similar behaviors among graduated students, and infers the book interests of students in school with the final borrowing information of these students. However, this inference method has the problem of system cold start like the collaborative filtering algorithm. When the number of borrowing times is small, the accuracy of the prediction result of this method is not high. Therefore, the prediction result of machine learning is used to add weighted sum with it. As the number of borrowing times and the time in school increase, the weight of the prediction result of machine learning is gradually reduced, so that this method can always obtain a relatively accurate real-time prediction result.

[0043] Pre-admission information includes one or more of the following: scores of various subjects, region, and the major to be studied. The specific encoding method of pre-admission information is the scores of various subjects in the college entrance examination. The corresponding subject ranking is divided by the total number of students in the corresponding subject in the region. The region is encoded using the geographic location, and the major to be studied is encoded using the embedding vector. Choosing appropriate admission information with a reasonable encoding method can better explore the relationship between admission information and students' interest in books, thereby improving the accuracy of push notifications.

[0044] Find the book categories with the largest real-time interest in the real-time book interest vector, and recommend the corresponding number of latest books or most popular books in the corresponding book categories according to the real-time interest of these book categories. Instead of predicting students' interest in a certain book, we predict students' interest in a certain type of book, and then combine other existing recommendation methods (recommending the most popular books, recommending the latest books) to make book recommendations, with high targeted recommendations. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 The present invention is a flow chart of a method for recommending books for colleges and universities. DETAILED DESCRIPTION

[0046] The present invention will be further described below in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solutions of the present invention and are not intended to limit the protection scope of the present invention.

[0047] Example 1

[0048] like Figure 1 As shown, the present invention discloses a book recommendation method for colleges and universities, comprising the following steps:

[0049] Step 1: Obtain the pre-enrollment information of the students, and then input the pre-enrollment information into the pre-trained machine learning model to obtain the initial book interest vector of the students.

[0050] Step 2: After each borrowing of the current students, a cumulative borrowing vector of the current students is generated, and the historical borrowing data of the graduated students is obtained to obtain the cumulative borrowing vector of the graduated students.

[0051] Step 3: Compare the cumulative borrowing vectors of current students with the cumulative borrowing vectors of graduated students to determine a number of graduated students with the highest similarity.

[0052] Step 4: Perform a weighted sum of the actual book interest vectors corresponding to several graduated students with the highest similarity to obtain the predicted book interest vector of the current students after borrowing.

[0053] Step 5: Perform a weighted summation of the initial book interest vector of the students and the predicted book interest vector after borrowing the books, to obtain the real-time book interest vector of the students.

[0054] Step 6: Recommend books to students based on their real-time book interest vectors.

[0055] The present invention uses the borrowing amount of each category of graduated students as a measure of the students' interest in books of the corresponding category. At the same time, the characteristic that colleges and universities can obtain relatively rich student admission information is utilized. According to the characteristic that colleges and universities can obtain more pre-admission information, this information is reasonably utilized, and a machine learning algorithm is used to automatically learn the relationship between this information and students' book interests. In addition, the book borrowing behavior of students in the past few years in college is the information most closely related to students' book interests. Therefore, the present invention draws on the idea of ​​collaborative filtering algorithm, and finds students with similar behaviors among graduated students, and infers the book interests of students in school based on the final borrowing information of these students. However, this inference method has the problem of system cold start like the collaborative filtering algorithm. When the number of borrowing times is small, the accuracy of the prediction results of this method is not high. Therefore, the prediction results of machine learning are used for weighted summation. As the number of borrowing times and the time in school increase, the weight of the prediction results of machine learning is gradually reduced, so that the method can always obtain a relatively accurate real-time prediction result.

[0056] Example 2

[0057] like Figure 1 As shown, the present invention discloses a book recommendation method for colleges and universities, comprising the following steps:

[0058] Step 1: obtain the pre-admission information of the students, and then input the pre-admission information into the pre-trained machine learning model to obtain the initial book interest vector of the students. The machine learning model is one of a decision tree, a support vector machine, a multi-layer perceptron, or a neural network model.

[0059] The pre-admission information includes one or more of the scores of each subject, region, and the major to be studied; the specific encoding method of the pre-admission information is the scores of each subject in the college entrance examination, and the corresponding subject ranking is divided by the total number of students in the corresponding subject in the region. The region is encoded using the geographic location, and the major to be studied uses an embedded vector.

[0060] Training a machine learning model consists of the following steps:

[0061] Get the historical borrowing data of graduated students and obtain the book borrowing vector of graduated students, which is specifically expressed as: v m =[ y m1 , y m2 … y mn … y mc ] ,

[0062] in, represents the book borrowing vector of the mth graduated student, represents the borrowing volume of the nth category of books by the mth graduated student, and c is the total number of book categories;

[0063] Normalize the book borrowing vector of graduated students to obtain the actual book interest vector of graduated students, which is specifically expressed as: ,

[0064] in, represents the actual book interest vector of the mth graduated student, Indicates the interest of the mth graduated student in the nth category of books;

[0065] The machine learning model is trained using the pre-admission information of graduated students as input and the actual book interest vectors corresponding to the graduated students as output.

[0066] The initial book interest vector includes students' interest information on books of various categories, which can be expressed as: ,

[0067] in, represents the initial book interest vector of the i-th student in school, Indicates the interest of the i-th student in the j-th category of books.

[0068] Step 2: After each borrowing of the current students, a cumulative borrowing vector of the current students is generated, and the historical borrowing data of the graduated students is obtained to obtain the cumulative borrowing vector of the graduated students.

[0069] The cumulative borrowing vector of students in school is specifically expressed as: w i e = [ x i1 e , x i2 e … x ij e … x ic e ] ;

[0070] in, is the cumulative borrowing vector of the i-th student after the e-th borrowing, It is the cumulative borrowing quantity of the j-th category book after the e-th borrowing by the i-th student in school.

[0071] The cumulative borrowing vector of graduated students after borrowing is specifically expressed as: v m e =[ y m1 e , y m2 e … y mn e … y mc e ]

[0072] in, is the cumulative borrowing vector of the mth graduated student after the eth borrowing, It indicates the borrowing amount of books in category n by the mth graduated student after accumulating e borrowing times. If the total number of borrowing times of the mth graduated student does not reach e times, then = .

[0073] Step 3: Whenever a student completes a borrowing, the cumulative borrowing vector after the e-th borrowing is used to compare the similarity with the cumulative borrowing vector of the graduated students, and the k graduated students with the highest similarity are determined, where k is a preset parameter.

[0074] Step 4: Perform a weighted sum of the actual book interest vectors corresponding to several graduated students with the highest similarity to obtain the predicted book interest vector of the current students after borrowing.

[0075] The predicted book interest vector after borrowing by students in school is specifically expressed as: u i e =[ z i1 e , z i2 e … z ij e … z ic e ] ,

[0076] in, represents the predicted book interest vector of the i-th student after the e-th borrowing. It represents the predicted interest of the i-th student in the j-th category of books after borrowing them for the e-th time;

[0077] Step 5: Perform a weighted summation of the initial book interest vector of the students and the predicted book interest vector after borrowing the books, to obtain the real-time book interest vector of the students.

[0078] The real-time book interest vector of students in school is specifically expressed as:

[0079]

[0080] in, represents the real-time book interest vector of the i-th student in school, represents the real-time interest of the i-th student in the j-th category of books, is the weight parameter, and λ ∈ [0,1] . Weight parameter The value of the weight is not fixed. As the number of borrowings and the time spent in school increase, The value of the number of books is increasing, and students do not borrow books. The value is fixed to 0.

[0081] Step 6: Recommend books to students based on their real-time book interest vectors. Recommending books to students based on their real-time book interest vectors is as follows: find the g book categories with the largest real-time interest in the real-time book interest vectors, and recommend a corresponding number of the latest books or the most popular books in the corresponding book categories according to the real-time interest of these book categories, where g is a preset parameter.

[0082] Example 3

[0083] This embodiment discloses a book recommendation device for colleges and universities, comprising:

[0084] The initial interest generation module is used to obtain the pre-enrollment information of the students, and then input the pre-enrollment information into the pre-trained machine learning model to obtain the initial book interest vector of the students;

[0085] The cumulative borrowing generation module is used to generate the cumulative borrowing vector of the students in school after each borrowing of the students in school, and obtain the historical borrowing data of the graduated students to obtain the cumulative borrowing vector of the graduated students;

[0086] The highest similarity determination module is used to compare the cumulative borrowing vectors of current students with the cumulative borrowing vectors of graduated students after each borrowing to determine a number of graduated students with the highest similarity;

[0087] The predicted interest generation module is used to perform weighted summation of actual book interest vectors corresponding to several graduated students with the highest similarity, and obtain the predicted book interest vector after borrowing by the students in school;

[0088] A real-time interest generation module is used to perform a weighted summation of the initial book interest vector of the students and the predicted book interest vector after the students borrow the books, so as to obtain the real-time book interest vector of the students;

[0089] The book recommendation module is used to recommend books to students based on their real-time book interest vectors.

[0090] Example 4

[0091] This embodiment introduces a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the steps of the method described in Embodiment 1 or 2 are executed.

[0092] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0093] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0094] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0095] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0096] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A book recommendation method for colleges and universities, characterized in that: The following steps are involved: Obtain the pre-enrollment information of the students, and then input the pre-enrollment information into the pre-trained machine learning model to obtain the initial book interest vector of the students; After each borrowing of a student, a cumulative borrowing vector of the student is generated, and historical borrowing data of graduated students is obtained to obtain the cumulative borrowing vector of the graduated students; Compare the cumulative borrowing vectors of current students with the cumulative borrowing vectors of graduated students to determine a number of graduated students with the highest similarity; The actual book interest vectors corresponding to several graduated students with the highest similarity are weighted summed to obtain the predicted book interest vectors of the students in school after borrowing; Performing a weighted summation of the initial book interest vector of the students and the predicted book interest vector of the students after borrowing, to obtain the real-time book interest vector of the students; Recommend books to students based on their real-time book interest vectors.

2. The book recommendation method for colleges and universities according to claim 1, characterized in that: The training of the machine learning model includes the following steps: Get the historical borrowing data of graduated students and obtain the book borrowing vector of graduated students, which is specifically expressed as: , in, represents the book borrowing vector of the mth graduated student, represents the borrowing volume of the nth category of books by the mth graduated student, and c is the total number of book categories; Normalize the book borrowing vector of graduated students to obtain the actual book interest vector of graduated students, which is specifically expressed as: , in, represents the actual book interest vector of the mth graduated student, Indicates the interest of the mth graduated student in the nth category of books; The machine learning model is trained using the pre-admission information of graduated students as input and the actual book interest vectors corresponding to the graduated students as output.

3. The book recommendation method for colleges and universities according to claim 2, characterized in that: The initial book interest vector includes the student's interest information for books of various categories, which is specifically expressed as: , in, represents the initial book interest vector of the i-th student in school, Indicates the interest of the i-th student in the j-th category of books.

4. The method for recommending books for colleges and universities according to claim 3, characterized in that: The cumulative borrowing vector of students in school is specifically expressed as: ; in, is the cumulative borrowing vector of the i-th student after the e-th borrowing, is the cumulative borrowing volume of the j-th category of books after the i-th student borrowed the e-th time; The cumulative borrowing vector of graduated students after borrowing is specifically expressed as: ; in, is the cumulative borrowing vector of the mth graduated student after the eth borrowing, It indicates the borrowing amount of books in category n by the mth graduated student after accumulating e borrowing times. If the total number of borrowing times of the mth graduated student does not reach e times, then = ; The predicted book interest vector after borrowing by students in school is specifically expressed as: , in, represents the predicted book interest vector of the i-th student after the e-th borrowing. It represents the predicted interest of the i-th student in the j-th category of books after borrowing them for the e-th time.

5. The method for recommending books for colleges and universities according to claim 4, characterized in that: The real-time book interest vector of students in school is specifically expressed as: ； in, represents the real-time book interest vector of the i-th student in school, represents the real-time interest of the i-th student in the j-th category of books, is the weight parameter, and .

6. The method for recommending books for colleges and universities according to claim 5, characterized in that: Weight Parameters The value of the weight is not fixed. As the number of borrowings and the time spent in school increase, The value of the number of books is increasing, and students do not borrow books. The value is fixed to 0.

7. The book recommendation method for colleges and universities according to claim 1, characterized in that: The pre-admission information includes one or more of the scores of each subject, region, and the major to be studied; the specific encoding method of the pre-admission information is the scores of each subject in the college entrance examination, and the corresponding subject ranking is divided by the total number of students in the corresponding subject in the region. The region is encoded using the geographic location, and the major to be studied uses an embedded vector.

8. The method for recommending books for colleges and universities according to claim 6, characterized in that: According to the real-time book interest vectors of students in school, the specific steps for recommending books to students are as follows: find the g book categories with the largest real-time interest in the real-time book interest vectors, and recommend a corresponding number of the latest or most popular books in the corresponding book categories according to the real-time interest of these book categories, where g is a preset parameter.

9. A book recommendation device for colleges and universities, characterized in that: include: The initial interest generation module is used to obtain the pre-enrollment information of the students, and then input the pre-enrollment information into the pre-trained machine learning model to obtain the initial book interest vector of the students; The cumulative borrowing generation module is used to generate the cumulative borrowing vector of the students in school after each borrowing of the students in school, and obtain the historical borrowing data of the graduated students to obtain the cumulative borrowing vector of the graduated students; The highest similarity determination module is used to compare the cumulative borrowing vectors of current students with the cumulative borrowing vectors of graduated students after each borrowing to determine a number of graduated students with the highest similarity; The predicted interest generation module is used to perform weighted summation of actual book interest vectors corresponding to several graduated students with the highest similarity, and obtain the predicted book interest vector after borrowing by the students in school; A real-time interest generation module is used to perform a weighted summation of the initial book interest vector of the students and the predicted book interest vector after the students borrow the books, so as to obtain the real-time book interest vector of the students; The book recommendation module is used to recommend books to students based on their real-time book interest vectors.

10. A computer-readable storage medium having a computer program / instruction stored thereon, characterized in that: When the computer program / instruction is executed by a processor, the steps of the method for recommending books for colleges and universities described in any one of claims 1-8 are implemented.

Images