Book recommendation system based on artificial intelligence
By introducing artificial intelligence and machine learning technology into the book recommendation system, identifying and adjusting the deviation of users' borrowing behavior, the problem of traditional recommendation systems decreasing when user data is scarce or behavior changes is solved, personalized and diversified recommendation effects are achieved, and users' reading experience is enriched.
Patent Information
- Application Number
- CN202411415063.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-11
- Publication Date
- 2025-03-07
AI Technical Summary
The traditional book recommendation system has a decrease in the effectiveness of user data when user data is scarce or user behavior changes, which is prone to biased recommendations, and cannot effectively adapt to the user's reading habits and interests, and does not consider the user's cognitive bias.
A book recommendation system based on artificial intelligence was designed, through user identity recognition, borrowing historical data extraction, cognitive bias recognition and diversified recommendation generation modules, combined with machine learning and cluster analysis, users' borrowing behavior deviations are identified, and by dynamically adjusting the feature weights in the user preference vector, a personalized and diversified recommendation list is generated.
Effectively identify and adjust users' cognitive biases, promote users to contact more book categories, enrich reading experience, improve the relevance and diversity of recommended results, and reduce users' dependence on specific categories of books.
Smart Images
Figure CN119577233A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of recommendation analysis, and in particular to a book recommendation system based on artificial intelligence. Background Art
[0002] With the popularity of digital reading, book recommendation systems have played an important role in helping users discover new books. Traditional recommendation systems are mainly based on users' borrowing history and user evaluation data, and make recommendations through collaborative filtering methods. However, these methods have some shortcomings in practical applications.
[0003] First, the collaborative filtering method relies on a large amount of user data to identify similar users, but the effectiveness of this method will drop significantly when user data is scarce. In addition, collaborative filtering is often susceptible to "biased recommendations", that is, the recommendation system may tend to recommend book types that users are already familiar with, resulting in a lack of diversity in recommendation results.
[0004] Secondly, content-based recommendation methods focus on analyzing the characteristics of books, such as themes, authors, and keywords. Although they can provide highly relevant recommendations, they often ignore users' personalized preferences and behavioral changes and cannot effectively adapt to users' reading habits and changes in interests.
[0005] In addition, traditional recommendation systems do not take into account users' cognitive biases, such as over-reliance on a certain type of book, which leads to limitations in recommendation results. The recommendation results may be biased towards certain specific types of books, further limiting users' reading scope and experience. Summary of the invention
[0006] The invention provides a book recommendation system based on artificial intelligence.
[0007] A book recommendation system based on artificial intelligence, comprising:
[0008] User identification module: authenticates the user through the user ID and extracts the user's basic information and borrowing history data after the user logs in;
[0009] Borrowing history data extraction module: Analyzes the borrowing history data obtained from the user identification module, records the borrowing time, borrowing frequency and book type of each book, and forms a user behavior feature data set;
[0010] Cognitive bias identification module: Analyzes user behavior feature data sets through machine learning, including using cluster analysis to classify user borrowing behaviors into different categories to identify user selection trends in specific book types, and then uses anomaly detection to analyze the degree of bias in user preferences (i.e., preference level), such as over-reliance on a certain type of book or neglect of other types;
[0011] Diversified recommendation generation module: Based on the potential bias pattern analysis results of the cognitive bias identification module, it actively recommends diversified books to users. The recommendation strategy adopts a combination of content-based recommendation and collaborative filtering to ensure that the recommendation results meet the user's potential interests while increasing books of different categories and themes, generating personalized recommendation lists, ensuring the relevance and diversity of the recommendation results, and avoiding users' dependence on books of specific categories.
[0012] Optionally, the cognitive bias identification module specifically includes:
[0013] Cluster analysis submodule: users are divided into multiple groups according to their borrowing behavior data. Each group represents a major type of book interest. These groups reflect different reading preferences of users. For each group, the average borrowing frequency of all users on a specific book type is calculated to identify the group's overall preference for a certain type of book.
[0014] In cluster analysis, users are divided into K different clusters according to their borrowing behavior, and each cluster C k Represents a major type of book interest, expressed as:
[0015] Among them, U k For class C k The average borrowing frequency of all users in B j Belongs to class C k The borrowing frequency of user j for a specific book type, n k For class C k The number of users in
[0016] Anomaly detection submodule: After determining the group to which each user belongs, analyze the user's borrowing behavior in the group, calculate the difference between the user's borrowing frequency of a specific book type and the average borrowing frequency of the group, and identify the degree of deviation of the user in a specific book type through the Zscore normalization method. For example, if the user's borrowing frequency is significantly higher or lower than the average level of the group, it indicates that the user's preference for books of this type may be abnormal.
[0017] Optionally, the cognitive bias identification module also considers the user's deviation in specific book types. If the user's behavior deviates significantly from the group's standard, that is, the deviation is too large, the deviation is amplified to ensure that the system can better reflect the user's unique needs and interests when making recommendations.
[0018] Optionally, the anomaly detection is based on Zscore standardization, which determines the degree of deviation of the user in the cluster to which he belongs. For each user X i In its class Ck The borrowing behavior in , calculate its Zscore: The calculation is expressed as: Among them, B i，k For user i in class C k The borrowing frequency in U k For class C k The average borrowing frequency of k For class C k The standard deviation of .
[0019] Optionally, the diversified recommendation generation module actively recommends diversified books to the user based on the deviation degree analysis result of the cognitive deviation identification module, and the diversified recommendation generation module specifically includes:
[0020] Content-based recommendation calculation submodule: Each book is converted into a book feature vector V through its features (author, subject, keywords) b ; Construct user preference feature vector V based on deviation degree analysis u , use cosine similarity to calculate the similarity between the book feature vector and the user preference feature vector. If the user has a high degree of deviation for a certain type of book, the contribution of this category will be reduced when calculating the similarity, thereby guiding the recommendation system to lean towards other types of books; sort according to the cosine similarity calculation results to generate the first recommendation list;
[0021] Collaborative filtering algorithm submodule: calculate the similarity between users, select neighboring users who are similar to the target user, calculate the recommendation score for the target user's unborrowed books based on the similarity calculation results between users, and generate a second recommendation list based on the recommendation score. The similarity calculation in the neighboring user selection is used to determine the user group similar to the target user. The borrowing behavior and scores of these similar users will be used to generate the recommendation score. In the recommendation score calculation, the score data of these similar users are used to predict the target user's preference for unborrowed books to form the final recommendation list.
[0022] Optionally, when constructing the user preference feature vector, a weight factor W is introduced c ,If the user has a high degree of deviation for the current book, the current weight factor is reduced.
[0023] Optionally, when calculating the similarity between users, the deviation of neighboring users in book types is taken into consideration, and neighboring users who are similar to the target user and have a low degree of deviation are selected.
[0024] Optionally, the user preference feature vector V u Introducing W c Later expressed as V u ′:
[0025] V u ′=Vu -D c ·W c , where W c is the weight factor associated with category c, D c is the degree of deviation (the degree of deviation of book category c). If the user has a high deviation for this category, D c Increase, reduce the influence of this category in the user's interest vector.
[0026] Use cosine similarity to calculate the similarity between the adjusted user interest vector and the book feature vector: By reducing the impact of the bias, the recommendation system is guided to lean towards other book types.
[0027] Optionally, the personalized recommendation list selects duplicate books in the first recommendation list and the second recommendation list;
[0028] Create a collection: Extract the book IDs from the first recommendation list and the second recommendation list respectively to form two collections:
[0029] Set 1 : The book ID set of the first recommendation list;
[0030] Set 2 : The book ID set of the second recommendation list;
[0031] Find duplicate books: Find the intersection of the two sets, which means the books that match the two lists: Personalized recommendation list = Set 1 ∩Set 2 ;
[0032] Output the final recommendation list, including book ID and title.
[0033] Beneficial effects of the present invention:
[0034] The present invention, through a cognitive bias identification module, deeply analyzes the user's borrowing history and behavioral characteristics, and effectively identifies the user's cognitive bias in book selection. It can not only accurately identify the user's excessive dependence on books of specific categories (such as science fiction, history, etc.), but also reduce the impact of bias by dynamically adjusting the feature weights in the user's preference vector, promote users to be exposed to a wider range of book categories, and thus enrich the user's reading experience.
[0035] The present invention adopts a method that combines content-based recommendation and collaborative filtering, and generates a final personalized recommendation list through a comprehensive analysis of two different recommendation strategies. When calculating the recommendation score, the contribution of each book feature can be dynamically adjusted according to the user's degree of deviation, which not only improves the relevance of the recommendation results, but also effectively guides users to explore book types that they have not paid attention to, reduces users' dependence on specific categories of books, and improves the diversity and novelty of recommendations.
[0036] The present invention calculates the similarity between users and, when recommending and scoring books that the target user has not borrowed, takes into account the deviation of neighboring users in book types. By selecting neighboring users who are similar to the target user and have a low degree of deviation, it effectively avoids recommending books with strong bias or limitations to the user. This not only improves the accuracy of the recommendation, but also ensures the diversity of the recommendation results, allowing users to be exposed to a wider range of book types, thereby enriching their reading experience and enhancing overall satisfaction. BRIEF DESCRIPTION OF THE DRAWINGS
[0037] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only for the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0038] Figure 1 A schematic diagram of a recommendation system module according to an embodiment of the present invention;
[0039] Figure 2 A schematic diagram of a diversified recommendation generation module according to an embodiment of the present invention;
[0040] Figure 3 Schematic diagram of a cognitive bias identification module according to an embodiment of the present invention. DETAILED DESCRIPTION
[0041] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. At the same time, it is explained here that in order to make the embodiments more detailed, the following embodiments are the best and preferred embodiments, and those skilled in the art may also adopt other alternatives to implement some known technologies; and the accompanying drawings are only for more specific description of the embodiments, and are not intended to specifically limit the present invention.
[0042] It should be noted that the references to "one embodiment", "an embodiment", "an exemplary embodiment", "some embodiments" and the like in the specification indicate that the embodiments described may include specific features, structures or characteristics, but not every embodiment may include the specific features, structures or characteristics. In addition, when a specific feature, structure or characteristic is described in conjunction with an embodiment, it should be within the knowledge of a person skilled in the art to implement such feature, structure or characteristic in conjunction with other embodiments (whether or not explicitly described).
[0043] In general, a term can be understood, at least in part, from its use in context. For example, depending, at least in part, on the context, the term "one or more" as used herein can be used to describe any feature, structure, or characteristic in the singular sense, or can be used to describe a combination of features, structures, or characteristics in the plural sense. Additionally, the term "based on" can be understood as not necessarily intended to convey an exclusive set of factors, but can instead, depending, at least in part, on the context, allow for the presence of other factors that are not necessarily explicitly described.
[0044] like Figure 1-Figure 3 As shown, a book recommendation system based on artificial intelligence includes:
[0045] User identification module: authenticates the user through the user ID and extracts the user's basic information and borrowing history data after the user logs in;
[0046] Borrowing history data extraction module: Analyzes the borrowing history data obtained from the user identification module, records the borrowing time, borrowing frequency and book type of each book, and forms a user behavior feature data set;
[0047] Cognitive bias identification module: Analyzes user behavior feature data sets through machine learning, including using cluster analysis to classify user borrowing behaviors into different categories to identify user selection trends in specific book types, and then uses anomaly detection to analyze the degree of bias in user preferences (i.e., preference level), such as over-reliance on a certain type of book or neglect of other types;
[0048] Diversified recommendation generation module: Based on the potential bias pattern analysis results of the cognitive bias identification module, it actively recommends diversified books to users. The recommendation strategy adopts a combination of content-based recommendation and collaborative filtering to ensure that the recommendation results meet the user's potential interests while increasing books of different categories and themes, generating personalized recommendation lists, ensuring the relevance and diversity of the recommendation results, and avoiding users' dependence on books of specific categories.
[0049] The cognitive bias identification module specifically includes:
[0050] Cluster analysis submodule: users are divided into multiple groups according to their borrowing behavior data. Each group represents a major type of book interest. These groups reflect the different reading preferences of users. For each group, the average borrowing frequency of all users on a specific book type is calculated to identify the group's overall preference for a certain type of book.
[0051] In cluster analysis, users are divided into K different clusters according to their borrowing behavior, and each cluster C k Represents a major type of book interest, expressed as:
[0052] Among them, U k For class C k The average borrowing frequency of all users in B j Belongs to class C k The borrowing frequency of user j for a specific book type, n k For class C k The number of users in
[0053] Anomaly detection submodule: After determining the group to which each user belongs, analyze the user's borrowing behavior in the group, calculate the difference between the user's borrowing frequency of a specific book type and the average borrowing frequency of the group, and identify the degree of deviation of the user in a specific book type through the Zscore normalization method. For example, if the user's borrowing frequency is significantly higher or lower than the average level of the group, it indicates that the user's preference for books of this type may be abnormal.
[0054] The cognitive bias identification module also considers the user's deviation in specific book types. If the user's behavior deviates significantly from the group's standard, that is, the deviation is too large, the deviation will be amplified to ensure that the system can better reflect the user's unique needs and interests when making recommendations.
[0055] Anomaly detection is based on Zscore normalization, which determines the degree of deviation of users in their clusters. For each user X i In its class C k The borrowing behavior in , calculate its Zscore: The calculation is expressed as: Among them, B i，k For user i in class C k The borrowing frequency in U k For class C k The average borrowing frequency of k For class C k The standard deviation of .
[0056] The diversified recommendation generation module actively recommends diversified books to users based on the bias degree analysis results of the cognitive bias identification module. The diversified recommendation generation module specifically includes:
[0057] Content-based recommendation calculation submodule: Each book is converted into a book feature vector V through its features (author, subject, keywords) b ; Construct user preference feature vector V based on deviation degree analysis u , use cosine similarity to calculate the similarity between the book feature vector and the user preference feature vector. If the user has a high degree of deviation for a certain type of book, the contribution of this category will be reduced when calculating the similarity, thereby guiding the recommendation system to lean towards other types of books; sort according to the cosine similarity calculation results to generate the first recommendation list;
[0058] Book feature vector V b The construction is as follows:
[0059] Feature selection: Select relevant features of books, including authors, topics, and keywords, and extract keywords through natural language processing technology.
[0060] Feature encoding: Convert each feature into a numerical representation:
[0061] Authors: Use one-hot encoding to map each author to a binary vector.
[0062] Topic: Use word embedding (Word2Vec) to represent topics.
[0063] Keywords: Similarly, one-hot encoding is used to indicate the importance of keywords.
[0064] Vector concatenation: concatenate all feature vectors together to form the book feature vector V b .
[0065] Construct user preference vector V u :
[0066] Feature extraction: Select books with high deviation as user preference books, use user preference book features (author, subject, keyword) for statistics, and construct user preference vector V in the same way as above. u .
[0067] Collaborative filtering algorithm submodule: calculate the similarity between users, select neighboring users who are similar to the target user, calculate the recommendation score for the target user's unborrowed books based on the similarity calculation results between users, and generate a second recommendation list based on the recommendation score. The similarity calculation in the neighboring user selection is used to determine the user group similar to the target user. The borrowing behavior and scores of these similar users will be used to generate the recommendation score. In the recommendation score calculation, the score data of these similar users are used to predict the target user's preference for unborrowed books to form the final recommendation list.
[0068] When constructing the user preference feature vector, the weight factor W is introduced c ,If the user has a high degree of deviation for the current book, the current weight factor is reduced.
[0069] When calculating the similarity between users, the deviation of neighboring users in book types is taken into account, and neighboring users who are similar to the target user and have a low degree of deviation are selected.
[0070] Neighboring user selection: Calculating similarity between users Consider the deviation when calculating similarity
[0071] The specific method is:
[0072] in;
[0073] User i With user u j The similarity between
[0074] B: The collection of all books;
[0075] User i Rating of book b;
[0076] User i The average rating is calculated as:
[0077] D c : The degree of deviation of the user in book category c;
[0078] |B|: the number of books in set B;
[0079] Recommendation score calculation: Calculate the recommendation score for each unborrowed book b′:
[0080] If the user's deviation degree on book category c is high, the corresponding score will be reduced, where;
[0081] R u，b ′: user u’s rating of the unborrowed book b′;
[0082] The average rating of user u is calculated using the same formula as above;
[0083] U k : A set of neighboring users similar to user u;
[0084] User u and neighboring user u j The similarity between
[0085] Neighboring user u j Rating of book b′;
[0086] Neighboring user u j The average score is calculated using the same formula as above.
[0087] User preference feature vector V u Introducing W c Later it is expressed as V u ′:
[0088] V u ′=V u -D c ·W c , where W c is the weight factor associated with category c, D c is the degree of deviation (the degree of deviation of book category c). If the user has a high deviation for this category, D c Increase, reduce the influence of this category in the user's interest vector.
[0089] Use cosine similarity to calculate the similarity between the adjusted user interest vector and the book feature vector: By reducing the impact of the bias, the recommendation system is guided to lean towards other book types.
[0090] The personalized recommendation list selects duplicate books in the first recommendation list and the second recommendation list;
[0091] Create a collection: Extract the book IDs from the first recommendation list and the second recommendation list respectively to form two collections:
[0092] Set 1 : The book ID set of the first recommendation list;
[0093] Set 2 : The book ID set of the second recommendation list;
[0094] Find duplicate books: Find the intersection of the two sets, which means the books that match the two lists: Personalized recommendation list = Set 1 ∩Set 2 ;
[0095] Output the final recommendation list, including book ID and title.
[0096] The present invention covers any substitution, modification, equivalent method and scheme made on the essence and scope of the present invention. In order to make the public have a thorough understanding of the present invention, specific details are described in detail in the following preferred embodiments of the present invention, but those skilled in the art can fully understand the present invention without the description of these details. In addition, in order to avoid unnecessary confusion about the essence of the present invention, well-known methods, processes, procedures, components and circuits are not described in detail.
[0097] 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 system based on artificial intelligence, characterized in that: include: User identification module: authenticates the user through the user ID and extracts the user's basic information and borrowing history data after the user logs in; Borrowing history data extraction module: Analyzes the borrowing history data obtained from the user identification module, records the borrowing time, borrowing frequency and book type of each book, and forms a user behavior feature data set; Cognitive bias identification module: Analyzes user behavior feature data sets through machine learning, including using cluster analysis to classify user borrowing behaviors into different categories to identify user selection trends in specific book types, and then analyzes the degree of bias in user preferences through anomaly detection; Diversified recommendation generation module: Based on the potential bias pattern analysis results of the cognitive bias identification module, it actively recommends diversified books to users. The recommendation strategy adopts a combination of content-based recommendation and collaborative filtering to ensure that the recommendation results meet the user's potential interests while increasing books of different categories and themes to generate a personalized recommendation list.
2. The book recommendation system based on artificial intelligence according to claim 1, characterized in that: The cognitive bias identification module specifically includes: Cluster analysis submodule: users are divided into multiple groups based on their borrowing behavior data. Each group represents a major type of book interest. For each group, the average borrowing frequency of all users on a specific book type is calculated to identify the group's overall preference for a certain type of book. Anomaly detection submodule: After determining the group to which each user belongs, analyze the user's borrowing behavior in the group, calculate the difference between the user's borrowing frequency of specific book types and the average borrowing frequency of the group, and identify the degree of deviation of the user in specific book types through the Zscore normalization method.
3. The book recommendation system based on artificial intelligence according to claim 2, characterized in that: The cognitive deviation identification module also considers the user's deviation in specific book types. If the user's behavior obviously deviates too much from the group's standard, that is, the deviation is too large, the deviation will be amplified.
4. The book recommendation system based on artificial intelligence according to claim 3, characterized in that: The anomaly detection is based on Zscore standardization, which determines the degree of deviation of users in their clusters. i In its class C k The borrowing behavior in is used to calculate its Zscore.
5. The book recommendation system based on artificial intelligence according to claim 1, characterized in that: The diversified recommendation generation module actively recommends diversified books to the user based on the deviation degree analysis result of the cognitive deviation identification module. The diversified recommendation generation module specifically includes: Content-based recommendation calculation submodule: Each book is converted into a book feature vector V through its features b ; Construct user preference feature vector V based on deviation degree analysis u , using cosine similarity to calculate the similarity between the book feature vector and the user preference feature vector, sorting according to the cosine similarity calculation results to generate a first recommendation list; Collaborative filtering algorithm submodule: calculates the similarity between users, selects neighboring users who are similar to the target user, calculates recommendation scores for the target user's unborrowed books based on the similarity calculation results between users, and generates a second recommendation list based on the recommendation scores.
6. The book recommendation system based on artificial intelligence according to claim 5, characterized in that: When constructing the user preference feature vector, the weight factor W is introduced c ,If the user has a high degree of deviation for the current book, then reduce the current weight factor.
7. The book recommendation system based on artificial intelligence according to claim 5, characterized in that: When calculating the similarity between users, the deviation of neighboring users in book types is taken into consideration, and neighboring users who are similar to the target user and have a low degree of deviation are selected.
8. The book recommendation system based on artificial intelligence according to claim 6, characterized in that: The user preference feature vector V u Introducing W c Later it is expressed as V u ′: V u ′=V u -D c ·W c , where W c is the weight factor associated with category c, D c is the degree of deviation. If the user has a high deviation for this category, D c Increase.
9. The book recommendation system based on artificial intelligence according to claim 6, characterized in that: The personalized recommendation list selects duplicate books in the first recommendation list and the second recommendation list; Create a collection: Extract the book IDs from the first recommendation list and the second recommendation list respectively to form two collections: Set1: the book ID set of the first recommendation list; Set2: the book ID set of the second recommendation list; Find duplicate books: Find the intersection of the two sets, which represents the books that meet the requirements in both lists: Personalized recommendation list = Set1∩Set2.