CN109918563A - A method of the book recommendation based on public data - Google Patents
A method of the book recommendation based on public data Download PDFInfo
- Publication number
- CN109918563A CN109918563A CN201910066005.7A CN201910066005A CN109918563A CN 109918563 A CN109918563 A CN 109918563A CN 201910066005 A CN201910066005 A CN 201910066005A CN 109918563 A CN109918563 A CN 109918563A
- Authority
- CN
- China
- Prior art keywords
- books
- book
- reader
- value
- recommendation
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a kind of methods of book recommendation based on public data, more particularly to digital library field, specifically include following recommended method: internet book store's book data acquisition, internet book store can provide the reading value assessment of the books of information, the acquisition of Feature of Library Collection book data, the reading value assessment of Feature of Library Collection books, book recommendation strategy and externally provide interface totally six steps of books reading value and recommendation service, the present invention by group wisdom book recommendation technology, the Books Marketing and recommendation information of major network audience publication are constantly acquired by network technology, according to these public datas as the thinking for recommending reference standard, the reading experience and wisdom of the mass users of the user in the library collective number Yuan Chaoyige, to complete the analysis being worth to books reading, to recommend most suitable books to reader, And targeted suggested design can be provided according to the different characteristics of all kinds of Library Books of books and the individual demand of user.
Description
Technical field
The present invention relates to Digital Library Technique field, it is more particularly related to a kind of based on public data
The method of book recommendation.
Background technique
Currently, flourishing for publishing provides more and more books to the people, and demand of the reader to reading
Higher and higher, since the background of reader is different, multifarious feature is presented in the requirement to book recommendation.By taking library as an example, remove
Collected books are continuously increased outer Main change and also focus on the service quality of internal environment and staff, for user
Personalization, activeization, high efficiency information service etc. it is more more and more intense demand response it is insufficient, especially provide it is good
Books counseling services in terms of.Reader it is generally desirable to library can according to consult request recommend some pertinent texts data so as into
Row selection.Current Libraries system can only provide very limited service mostly in this regard, or rely on Librarians
Irregular professional standards, or record Recommended Books are borrowed according to the limited user in the library.Regardless of single books
Still the data in multiple libraries are realized that shared is still the lesser data set of scale in shop, it is difficult to realize it is very accurate and
Comprehensive book recommendation service;
The developing direction of library information services is to provide individual info service, and reader is helped to find books, improves figure
Book utilization rate and book-loaning ratio solve the problems, such as " the book information overload " increasingly expanded.Currently, the work in this field mainly stresses
Digital Library-Oriented field.The Individual book information service of book borrowing and reading towards traditional libraries is not yet in practical application
Apparent progress is obtained in field.The external relatively early information services research work for having carried out library's information recommender system,
Such as the FAB recommender system of Stanford University and the My Library system of Cornell University.It is domestic relatively early successfully to develop and apply
Be library, Zhejiang University " my library " system, correlation is researched and developed in main research library or in internet book store
Correlation rule between books is designed and calculates relevant proposed algorithm, recommends related books list to reader.Base is used mostly
In the two class technology of recommendation and collaborative filtering recommending of content;
Content-based recommendation: the product that it likes in the past according to user is the product phase that user recommends and he likes in the past
As product.Its process generally comprises following three step: 1) extracting some features for each item;2) it is liked in the past using a user
The vigorously characteristic of the item of (and not liking), to learn the hobby feature of this user out;3) by comparing use obtained in the previous step
The feature of feature and candidate item is liked at family, and user recommends the item of one group of correlation maximum thus.The advantages of such method, is: 1) using
Independence between family;2) result is easy to explain;3) new project can be recommended immediately.And disadvantage then includes: 1) project
Permitted to be not easy extraction feature;2) the potential interest characteristics of user are unable to get;3) recommendation can not be made for new user;
Collaborative filtering recommending: what Amazon (amazon.com) internet book store and Facebook advertisement were recommended to use is collaboration
Filtered recommendation (Collaborative Filtering recommendation) algorithm, by analyzing user interest, in user
Similar (interest) user of designated user is found in group, evaluation of these the comprehensive similar users to a certain information forms system pair
The designated user predicts the fancy grade of this information.Collaborative filtering has the advantage that (1) can filter and is difficult to carry out machine certainly
The dynamic information based on content analysis;It (2) can be based on some complexity, it is difficult to which the concept (information quality, grade) of expression carries out
Filtering;(3) novelty recommended.The disadvantage is that: (1) user to the evaluation of commodity very sparse, the evaluation institute in this way based on user
Similitude between obtained user may inaccuracy (i.e. sparsity problem);(2) system performance can be with the increasing of user and commodity
It reduces more;(3) if never there is user to evaluate a certain commodity, this commodity is impossible to be recommended (i.e. most
First evaluation problem).Therefore, present Technologies of Recommendation System in E-Commerce all uses the recommended technology that several technologies combine.
The patent of invention of patent application publication CN104679835A discloses a kind of books based on multiple view Hash and pushes away
Method is recommended, is included the following steps: 1) from filtering out behavioral data of the user on two views in result collection system, including figure
Book click data and search data;2) building user is clicking and is searching for the user characteristics vector on view;3) two views are utilized
The behavioral data of figure obtains the weight of user's Hash coding, hash function and two views by multiple view hash algorithm;4)
Target user, which is encoded to, using obtained user's Hash finds similar users;5) the books set of similar users click is obtained, is made
For recommended candidate list, target user is calculated to the preference of books, returns to maximum preceding this figure of N of target user's preference
Book.User can be integrated into Hash coding by the present invention in the behavioral data of two views, improve book recommendation accuracy;Separately
On the one hand, quickly, the efficiency of book recommendation can be improved in the Hamming distance calculating speed of Hash coding;
The patent of invention of patent application publication CN103886067A disclose it is a kind of using label imply theme carry out figure
The method that book is recommended, it is using books as document, and book labels are as the word in document, using LDA-Gibbs algorithm to figure
Book label carries out theme modeling, obtains label-topic model, then records to obtain user and label according to the books reading of user
Corresponding relationship, user-topic model is obtained using LDA-inference algorithm, finally according to user on theme distribution
Similarity finds the similar user of interest, carries out collaborative filtering recommending to books.The present invention has sufficiently excavated in book labels
Semantic information, dimension needed for reducing expression user by using theme, reduces calculation amount, helps to improve recommendation results
Quality, have certain practical value;
The patent of invention of patent application publication CN103886048A discloses a kind of incremental digital books based on cluster
Recommended method, it is the following steps are included: then (1) gives birth to from the information of the web log of user acquisition user's read books
Vector is indicated at user;(2) calculative gathering is selected using dimension array, is then calculated remaining between user and gathering
String similarity forms Candidate Set;(3) found out from Candidate Set with the highest cluster of target user's similarity, then according to amalgamation result
It is clustered, and incrementally updates the cluster heart, cluster diameter;(4) cluster center value is used to arrange project in cluster as ranking functions
Sequence, the high project that will sort is as recommendation results.The present invention can be by excavating user to book from the books access log of user
The preference information of nationality, then recommends for user, improves the scalability and real-time of recommended method, enhances digital book
The reading experience of resource utilization and user;
The patent of invention of patent application publication CN103488714A discloses a kind of book recommendation based on social networks
Method and system, method include: step 1 extracts the interactive information of user and other users in social networks, is user's structure
Several interaction style buddy groups are built, then there will be successfully the other users of interactive relationship to be divided according to its interaction style with user
Into different interaction style buddy groups, the successfully interaction is that user carries out oneself interactive relationship between other users
It responds;Step 2, the successful interaction number for calculating separately user and each good friend in each interaction style buddy group, then from every
Several maximum preceding good friends of successfully interaction number are picked out in a interaction style buddy group, finally readding several good friends
Most books are read to recommend to user.The invention belongs to network communication technology fields, can be mutual in social networks according to user
Dynamic behavior carries out the personalized recommendations of books;
Foregoing invention scheme uses conventional based on commending contents or collaborative filtering recommending method, their master mostly
Problem is wanted to be:
(1) all without providing the method for being easy to get a large amount of data good enough, some even lacks in actual use
The means of data needed for obtaining.Such as: it relies on the data such as personal information and the usage behavior of user and analyzes, and in fact,
Personal account can be seldom logged in when user merely desires to using library indexing system to search books, this is to say system can not adopt
Collecting the use process data of user, this personalized service has just lacked its availability, and even if reader has logged in a acknowledge a debt
Family, this searching system still lack the recommending module of books correlation;Other schemes need the label about book content,
The quality of label directly affects the accuracy of recommendation, and rely on the method for the contents such as books keyword progress automated tag merely
Quality not can guarantee;
(2) it is all based on what subject key words etc. were retrieved.This search modes are obtained based on Keywords matching, once
The result of retrieval may reach hundreds of, and it is often sub-fraction therein that user is interested, and related due to lacking
The recommended technology of books, thus can not retrieve content it is related but and do not contain the books of the keyword, search result seems not
It is reasonable to the greatest extent;
(3) can not overcome based on commending contents or the shortcomings that collaborative filtering recommending method, including item characteristic is not easy to extract
Feature;The potential interest characteristics of user can not be analyzed;Recommendation can not be made for new user;User is sparse to the evaluation of commodity to be led
It causes between the analysis misalignment of similitude user;User and increasing for books cause the performance of system to decline;The commodity of evaluation are not obtained
It would not be recommended.
Summary of the invention
In order to overcome the drawbacks described above of the prior art, the embodiment of the present invention provides a kind of books based on public data and pushes away
The method recommended directly acquires the information of these magnanimity by web crawlers from major internet book store, has well solved analysis number
The problem of according to source;The internal association between books analyzed using the real behavior of magnanimity reader is not rely on shared
Keyword or other manually or automatically labels, be associated with reader and approve that content is related but be free of the books of keyword, this
The reader that one scheme makes book recommendation eliminate the reliance on limited quantity borrows record, and can directly the magnanimity of bookstore is used from network
It analyzes and obtains in the comment at family, thus more rationally;The books arranged mode that books press similar library is constituted one by the present invention
The reading value information of these books is packaged into service interface, to readers and users and public library etc. by a simulation library
Mechanism provides, and according to keyword or specified book that user provides, the pertinent texts of value are most read by system recommendation,
Make the available better Book Review service of user.
To achieve the above object, the invention provides the following technical scheme: a kind of side of the book recommendation based on public data
Method specifically includes following recommended method:
Step 1: internet book store's book data acquires, firstly, developing network crawler, from brilliance-Amazon, when
Large-scale internet book store acquires book data, the net of incidence relation between one books of information structuring according to provided by internet book store
Network, associated information includes " people for having bought the book also buys ... book ", " often buying ... book together ", " browsed between books
The evaluation of the star of the people of the book also bought ... book " and user, but do not include specific evaluation of the reader to books, these are associated with
Be regarded as a kind of mutual recommendation relationship, these recommendation relationships can form a huge related network, except master data it
Outside, associated data between acquisition books, data are saved into database after over cleaning arranges;
Step 2: internet book store can provide the reading value assessment of the books of information, based on the data acquired in step 1,
It is proposed a kind of assessment algorithm (hereinafter referred to as BookRank algorithm), according to the chain of a books enter with chain go out quality and quantity come
It measures it and reads value, specifically calculate the reading value (hereinafter referred to as BookRank) of every books, the core of BookRank algorithm
Thought road are as follows: the link for each arriving the books is primary approval ballot to it, 1) the higher books of degree of recognition are more worth pushes away
Recommend and be recognized the more worth recommendation of more books, 2) approved source book quality is higher, is recognized by these high quality books
Can books also more worth recommendation, linked of book mean to be voted by other books much more, read value
It is just higher, its higher BookRank value is assigned at this time, for measuring the i.e. recommendable degree of its significance level;
Step 3: Feature of Library Collection book data acquires, according to the identity of reader, to provide information service as give-and-take conditions
Related data of the Feature of Library Collection books after desensitizing is obtained from library, books are then excavated by analysis reader conduct log sheet
Internal association between reader and between books is specially reserved, borrows and is renewed, by calculating between books and reader
Contribution weight, books between disturbance degree and books between similar support measure the level of intimate of this relationship, will
The characteristic books of acquisition borrow data and are saved into database after over cleaning arranges;
Step 4: the reading value assessment of Feature of Library Collection books is measured based on the data in step 3 according to following principles
The recommendation value of one Feature of Library Collection books: 1) books and authoritative higher reader push away there are the number of correlation behavior more more
It is higher to recommend value;2) there are the more more then more worth recommendations of the number of correlation behavior with more readers for books;3) two books simultaneously with
The recommendation weight that more readers exist simultaneously between the more more then books of reader's quantity of correlation behavior is higher;
The reading of books is worth and is carried out according to the identity of reader, the other books many aspects borrowing quantity, borrowing
Assessment: different readers are calculated from initial data first to the disturbance degree of different books classifications, reader conduct of the reader A at oneself
In log, the number of behavior incidence relation existing for certain class C to books is more, then reader A compares this books classification C
The influence power of other books classifications is bigger;If the identity of reader A is more professional, more authoritative, then reader A is to this books classification
C is bigger than to influence power of other readers to this books classification C;In order to measure different readers to belonging to its related books
The influence degree and readership technorati authority of books classification introduce " reader-classification " weight to the disturbance degree of books classification;
Then calculate books between recommendation weight, on the basis of previous step, if books X and books Y with it is multiple
Reader exists simultaneously incidence relation, then books X is that these readers " read classification belonging to books Y to the contribution weight of books Y
The weighted sum of person-classification " weight;Similarity between books X and books Y is that there are the readings of incidence relation with this books simultaneously
Person's quantity accounts for that there are the ratios of reader's quantity union of incidence relation with books X, books Y;So recommendation of the books X to books Y
Weight is product of the books X to the contribution weight of books Y and the similarity of books X and books Y;Present invention introduces between books
Recommend weight w (y → x), wherein only consider simultaneously with the related readership of books x and y;
Step 5: book recommendation strategy designs two kinds of book recommendation plans on the basis of obtaining the BookRank value of books
Slightly, the Generalization bounds respectively for the Generalization bounds of keyword retrieval and for certain this books, and the books recommended are pressed
BookRank value sorts be supplied to user from high to low;
For the Generalization bounds of search key: when user inputs keyword and retrieves, system is to being retrieved
As a result statistic of classification is carried out by the different books categorical measures to appearance of books classification, then according to each classification number in result
The difference of amount, the Recommended Books number of each classification of reasonable distribution in proportion, and selected from these classifications according to BookRank value
Books are selected, are sorted after summarizing to these books by BookRank value;
For the Generalization bounds of specific books: when user selects a certain books, being pushed away according to the classification information of the books
A certain number of related books are recommended, thinking is: when user selects certain this books, it is meant that the interested books of the user may
In the books classification for being same classification with selected books, so recommendable books should concentrate on the books of current class
In, herein using the category as selecting bibliography to record;In book database, all books are all sorted out by tree structure, catalogue
Record is total classification, and the directory name of catalogue is total classification number, and as the root node of tree structure, subdirectory is intermediate knot
Point, books are as leaf node;In this way, select bibliography record leaf lazy weight even without when, bibliography can be selected to record this
Parent directory be set as currently bibliography being selected to record, step by step expand select bibliography to record, recommend number as defined in system until the books selected reach
Until mesh N;
So recommendable range can be extended to upper level catalogue when current class Books recommend insufficient, when
When user will search a books, which can search books according to the search key that user inputs first, and retrieve
Result have content relevance books be presented to the user again according to specific sort algorithm;
The interface of books reading value and recommendation service is externally provided, by books reading value and associated recommendation information preservation
In background data base, an external group polling interface is then provided by network, provides relevant information service to each library.
In a preferred embodiment, the iterative formula of BookRank is as follows in the step 2:
Wherein, BRn(A) the BookRank value of books A is represented;(BRn-1(Ti) represent books TiIn upper primary iteration
BookRank value, when books A appears in books TiRelated books list in, i.e. TiIt is linked to A;C(Ti) represent books Ti's
The total quantity of all books inside related books list;D represents damped coefficient;In above-mentioned formulaIt indicates in related network by T during access chain random access booksiBy itself
The part that proportion is d in BookRank value is averagely allocated to each books in its related list, therefore books A is obtained
Come from books Ti1/C (Ti) BookRank value.
In a preferred embodiment, in the step 2, in calculating process, algorithm after successive ignition,
The BookRank value of every books is finally converged in a more stable value.
In a preferred embodiment, in the step 2, in algorithm realization, according to the meter of PageRank power method
Principle is calculated, BookRank is also calculated in this way, to be converted into solution for what above-mentioned formula was abstracted's
Value, wherein x is the initial BookRank value of every books, and initial value can arbitrarily be set, it is set as 1 by the present invention;Matrix A
Are as follows:
A=(1-d) × eeT+d×PT
Wherein, d is damped coefficient, and e is n dimensional vector, eTRow vector, P are tieed up for nTFor probability transfer matrix [S
Krishnaswamy];In iterative calculation each time, matrix A is all changing, the following institute of dominant eigenvalue calculating process
Show:
(1) the initial BookRank value of x ← init//tax
(2) r ← Ax//r is BookRank value, and A is linking relationship matrix (as shown in formula 5.5)
(3) the BookRank value of if f (| | x-r | |) < ε, return r//compare before and after iteration
(4) else x ← r, goto (2) // and it is unsatisfactory for ε, continue iteration
Wherein, init is the initial BookRank value of setting, and x indicates current BookRank value, after r represents iteration
BookRank value, ε indicate tolerance.
In a preferred embodiment, in the step 2, for construct probability transfer matrix A construction, by books
Incidence relation linked network is expressed as the linking relationship matrix of n row (n+1) column, convenient for the realization of system program.
In a preferred embodiment, " reader-classification " weight is as follows in the step 4:
Wherein i is indicated with books there are the reader of incidence relation, c indicate with belonging to books of the reader i there are incidence relation
Classification, h (i, c) indicate reader i incidence relation books classification be c books behavior number, f (i) indicate reader i closing
The behavior number of the books of connection relationship, k indicate the weight value coefficient of readership authority, and value range is to go beyond one's commission between 0 to 1
The reader of prestige, rights relating the person authority weight values k is bigger, and λ (i, c) indicates " reader-classification " weight, i.e. reader i, which has it, to close
The affiliated books classification of the books of connection relationship is the disturbance degree of c class.
In a preferred embodiment, in the step 4, the recommendation weight between books is as follows:
Wherein i is with books x there are the reader of incidence relation, and j is reader's quantity with books x there are incidence relation, c
It (x) is books classification belonging to books x, λ (j, c) indicates reader j to the contribution weight of books classification c, and k is indicated while and books
For x and books y there are reader's quantity of incidence relation, m indicates that with books x, n indicates that books y exists there are associated reader's quantity
Associated reader's quantity.
In a preferred embodiment, book recommendation formula is as follows in the step 4:
Wherein, RMD (W) indicates the total quantity recommended for search key W;N is indicated appeared in search result
Different books classifications quantity;The total quantity of N expression search result;NiIndicate the figure for belonging to i-th of classification in search result
Book quantity;AiIndicate i-th of total quantity being sorted in entire book database;It indicates from i-th categorizing selection
Books quantitative range for recommendation;U is quantity coefficient, and value is between 0 to 1, to guarantee to be elected to be the books of recommendation at such
Number is in the reasonable range of a comparison, and the recommendation that M is then every class books sets a upper limit.
Technical effect and advantage of the invention:
1, the present invention constantly acquires major network audience hair by network technology by the book recommendation technology of group wisdom
The Books Marketing and recommendation information of cloth, according to these public datas as the thinking for recommending reference standard, collective number far surpasses one
The reading experience and wisdom of the mass users of the user in a library, to complete the analysis being worth to books reading, so as to reading
Person recommends most suitable books, so that library is no longer limited to rely on locally limited data of borrowing and is recommended, it can be significant
Improve its reader service level, and can be provided according to the different characteristics of all kinds of Library Books of books and the individual demand of user
Targeted suggested design, and provided with the method for service of service interface to library;
2, the present invention directly acquires the information of these magnanimity by web crawlers from major internet book store, well solves
The problem of analyzing data source;The internal association between books analyzed using the real behavior of magnanimity reader is not relying on
In shared keyword or other manually or automatically labels, it is associated with reader and approves that content is related but is free of the figure of keyword
Book, the reader that this scheme makes book recommendation eliminate the reliance on limited quantity borrow record, and can direct bookstore from network
It analyzes and obtains in the comment of mass users, thus more rationally;
3, the books arranged mode that books press similar library is constituted a simulation library by the present invention, these books
Reading value information be packaged into service interface, to the mechanisms such as readers and users and public library provide, according to user provide
Keyword or specified book, the pertinent texts of value are most read by system recommendation, make the available better figure of user
Book information service.
Detailed description of the invention
Fig. 1 is overall flow figure of the invention.
Fig. 2 is to carry out book recommendation flow chart of steps according to the search key that user provides in the embodiment of the present invention 2.
Fig. 3 is the recommendation step flow chart that specific books are directed in the embodiment of the present invention 3.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation description, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts every other
Embodiment shall fall within the protection scope of the present invention.
Embodiment 1:
The present invention provides a kind of methods of book recommendation based on public data as shown in Figure 1, specifically include following
Recommended method:
Step 1: internet book store's book data acquires, firstly, developing network crawler, from brilliance-Amazon, when
Large-scale internet book store acquires book data, the net of incidence relation between one books of information structuring according to provided by internet book store
Network, associated information includes " people for having bought the book also buys ... book ", " often buying ... book together ", " browsed between books
The evaluation of the star of the people of the book also bought ... book " and user, but do not include specific evaluation of the reader to books, these are associated with
Be regarded as a kind of mutual recommendation relationship, these recommendation relationships can form a huge related network, except master data it
Outside, associated data between acquisition books, data are saved into database after over cleaning arranges;
Step 2: internet book store can provide the reading value assessment of the books of information, based on the data acquired in step 1,
It is proposed a kind of assessment algorithm (hereinafter referred to as BookRank algorithm), according to the chain of a books enter with chain go out quality and quantity come
It measures it and reads value, specifically calculate the reading value (hereinafter referred to as BookRank) of every books, the core of BookRank algorithm
Thought road are as follows: the link for each arriving the books is primary approval ballot to it, 1) the higher books of degree of recognition are more worth pushes away
Recommend and be recognized the more worth recommendation of more books, 2) approved source book quality is higher, is recognized by these high quality books
Can books also more worth recommendation, linked of book mean to be voted by other books much more, read value
It is just higher, its higher BookRank value is assigned at this time, for measuring the i.e. recommendable degree of its significance level;
The iterative formula of BookRank is as follows:
Wherein, BRn(A) the BookRank value of books A is represented;(BRn-1(Ti) represent books TiIn upper primary iteration
BookRank value, when books A appears in books TiRelated books list in, i.e. TiIt is linked to A;C(Ti) represent books Ti's
The total quantity of all books inside related books list;D represents damped coefficient;In above-mentioned formulaIt indicates in related network by T during access chain random access booksiBy itself
The part that proportion is d in BookRank value is averagely allocated to each books in its related list, therefore books A is obtained
Come from books Ti1/C (Ti) BookRank value;
In calculating process, algorithm after successive ignition, the BookRank value of every books be finally converged in one compared with
Stable value;
In algorithm realization, according to the Computing Principle of PageRank power method, BookRank is also counted in this way
It calculates, to be converted into solution for what above-mentioned formula was abstractedValue, wherein x is the initial BookRank value of every books,
Initial value can arbitrarily be set, it is set as 1 by the present invention;Matrix A are as follows:
A=(1-d) × eeT+d×PT
Wherein, d is damped coefficient, and e is n dimensional vector, eTRow vector, P are tieed up for nTFor probability transfer matrix [S
Krishnaswamy];In iterative calculation each time, matrix A is all changing, the following institute of dominant eigenvalue calculating process
Show:
(1) the initial BookRank value of x ← init//tax
(2) r ← Ax//r is BookRank value, and A is linking relationship matrix (as shown in formula 5.5)
(3) the BookRank value of if f (| | x-r | |) < ε, return r//compare before and after iteration
(4) else x ← r, goto (2) // and it is unsatisfactory for ε, continue iteration
Wherein, init is the initial BookRank value of setting, and x indicates current BookRank value, after r represents iteration
BookRank value, ε indicate tolerance;
For the construction for constructing probability transfer matrix A, books incidence relation linked network is expressed as to the chain of n row (n+1) column
Relational matrix is connect, convenient for the realization of system program, BookRank algorithm is (BookRank program circuit pseudocode) as follows:
Step 3: Feature of Library Collection book data acquire, due to the rare library preservation characteristic books of internet book store (ancient books,
Secondhand book, publication etc.) information, these books reading value must be assessed by the way of different from the above method, therefore, root
According to the identity of reader, using provide information service as give-and-take conditions from library obtain Feature of Library Collection books desensitized (remove or
Obscure may cause leakage personal information data data) after related data, then by analysis reader conduct log sheet dig
(reserve, borrow and renew) between pick books and reader and books between internal association, by calculate books and reader it
Between contribution weight, books between disturbance degree and books between similar support measure the level of intimate of this relationship,
The characteristic books that will acquire borrow data and are saved into database after over cleaning arranges;
Step 4: the reading value assessment of Feature of Library Collection books is measured based on the data in step 3 according to following principles
The recommendation value of one Feature of Library Collection books: 1) books and authoritative higher reader push away there are the number of correlation behavior more more
It is higher to recommend value;2) there are the more more then more worth recommendations of the number of correlation behavior with more readers for books;3) two books simultaneously with
The recommendation weight that more readers exist simultaneously between the more more then books of reader's quantity of correlation behavior is higher;
According to the identity (related to its professional authority) of reader, the other books many aspects borrowing quantity, borrowing
The readings of books value is assessed: calculating different readers from initial data first to the disturbance degree of different books classifications,
Reader A is in the reader conduct log of oneself, and the number of behavior incidence relation existing for certain class C to books is more, then reading
Person A is bigger than the influence power to other books classifications to this books classification C;If the identity of reader A is more professional, more authoritative,
So reader A is bigger than to influence power of other readers to this books classification C to this books classification C;In order to measure difference
Influence of the reader to the influence degree and readership technorati authority of its related affiliated books classification of books to books classification
It is as follows to introduce " reader-classification " weight for degree:
Wherein i is indicated with books there are the reader of incidence relation, c indicate with belonging to books of the reader i there are incidence relation
Classification, h (i, c) indicate reader i incidence relation books classification be c books behavior number, f (i) indicate reader i closing
The behavior number of the books of connection relationship, k indicate the weight value coefficient of readership authority, and value range is to go beyond one's commission between 0 to 1
The reader of prestige, rights relating the person authority weight values k is bigger, and λ (i, c) indicates " reader-classification " weight, i.e. reader i, which has it, to close
The affiliated books classification of the books of connection relationship is the disturbance degree of c class;
Then calculate books between recommendation weight, on the basis of previous step, if books X and books Y with it is multiple
Reader exists simultaneously incidence relation, then books X is that these readers " read classification belonging to books Y to the contribution weight of books Y
The weighted sum of person-classification " weight;Similarity between books X and books Y is that there are the readings of incidence relation with this books simultaneously
Person's quantity accounts for that there are the ratios of reader's quantity union of incidence relation with books X, books Y;So recommendation of the books X to books Y
Weight is product of the books X to the contribution weight of books Y and the similarity of books X and books Y;Present invention introduces between books
Recommend weight w (y → x), wherein only consider simultaneously with the related readership of books x and y:
Wherein i is with books x there are the reader of incidence relation, and j is reader's quantity with books x there are incidence relation, c
It (x) is books classification belonging to books x, λ (j, c) indicates reader j to the contribution weight of books classification c, and k is indicated while and books
For x and books y there are reader's quantity of incidence relation, m indicates that with books x, n indicates that books y exists there are associated reader's quantity
Associated reader's quantity;
Step 5: book recommendation strategy designs two kinds of book recommendation plans on the basis of obtaining the BookRank value of books
Slightly, the Generalization bounds respectively for the Generalization bounds of keyword retrieval and for certain this books, and the books recommended are pressed
BookRank value sorts be supplied to user from high to low;
For the Generalization bounds of search key: when user inputs keyword and retrieves, system is to being retrieved
As a result statistic of classification is carried out by the different books categorical measures to appearance of books classification, then according to each classification number in result
The difference of amount, the Recommended Books number of each classification of reasonable distribution in proportion, and selected from these classifications according to BookRank value
Books are selected, are sorted after summarizing to these books by BookRank value;
Wherein, RMD (W) indicates the total quantity recommended for search key W;N is indicated appeared in search result
Different books classifications quantity;The total quantity of N expression search result;NiIndicate the figure for belonging to i-th of classification in search result
Book quantity;AiIndicate i-th of total quantity being sorted in entire book database;It indicates from i-th categorizing selection
Books quantitative range for recommendation;U is quantity coefficient, and value is between 0 to 1, to guarantee to be elected to be the books of recommendation at such
Number is in the reasonable range of a comparison, and the recommendation that M is then every class books sets a upper limit;
For the Generalization bounds of specific books: when user selects a certain books, being pushed away according to the classification information of the books
A certain number of related books are recommended, thinking is: when user selects certain this books, it is meant that the interested books of the user may
In the books classification for being same classification with selected books, so recommendable books should concentrate on the books of current class
In, herein using the category as selecting bibliography to record;In book database, all books are all sorted out by tree structure, catalogue
Record is total classification, and the directory name of catalogue is total classification number, and as the root node of tree structure, subdirectory is intermediate knot
Point, books are as leaf node;In this way, select bibliography record leaf lazy weight even without when, bibliography can be selected to record this
Parent directory be set as currently bibliography being selected to record, step by step expand select bibliography to record, recommend number as defined in system until the books selected reach
Until mesh N;
So recommendable range can be extended to upper level catalogue when current class Books recommend insufficient, when
When user will search a books, which can search books according to the search key that user inputs first, and retrieve
Result have content relevance books be presented to the user again according to specific sort algorithm;
The interface of books reading value and recommendation service is externally provided, by books reading value and associated recommendation information preservation
In background data base, an external group polling interface is then provided by network, provides relevant information service to each library.
Core of the invention thinking is to be associated with to close first between one books of information structuring according to provided by internet book store
The network of system, associated information includes " people for having bought the book also buys ... book ", " often buys together ... between these books
The evaluation of the star of book ", " people for browsing the book also bought ... book " etc. and user, but do not include reader to the specific of books
Evaluation;Then using the algorithm of the PageRank of a similar Google, from analysis meter in incidence relation network between this books
Calculation obtains the reading value information of books.For Feature of Library Collection books a large amount of in the collected books in library (such as ancient books and old edition
Books publication etc.), the present invention by according to the identity (related to its professional authority) of reader, borrow quantity, borrow
The many aspects such as other books assess the reading value of books;
On this basis, the present invention provides two kinds of inquiry modes based on keyword and title, according to the inquiry of reader and
Books read value, recommend the related books for being suitble to its demand to reader.
Embodiment 2:
Carrying out book recommendation according to the search key that user provides, steps are as follows (referring in particular to Figure of description 2):
1) start;
2) keyword is inputted;
3) with the presence or absence of the book information to match with term, and if it exists, then obtained to exist according to correlated results and be recommended
Associated books, if not in the presence of then exporting related prompt, terminating;
4) it is obtained to exist according to correlated results and recommends associated books, if getting, filtered recommendation book information, if not
It gets, then directly terminates;
5) after filtered recommendation book information, the books that output recommends weight high finally terminate.
Embodiment 3:
It is following (referring in particular to Figure of description 3) for the recommendation step of specific books:
1) search books when work as upper recommendation list, if finding, filter out with the consistent books of current book classification,
If not finding, bibliography is selected to record the affiliated directory device of current books;
2) it filters out with after the consistent books of current book classification, when enough this whens of N, is directly recommended;
3) after selecting bibliography to record the affiliated directory device of current books, book is selected in selecting bibliography record, complements to N sheet;
If 4) enough N sheets, are directly recommended, if not enough N sheet, checks whether parent directory is root node;
If 5) root node is then directly recommended, if not root node, then set parent directory to that bibliography is selected to record, so
Book is selected back in selecting bibliography record afterwards, complements in this step of of N, continues to recommend.
The several points that should finally illustrate are: firstly, in the description of the present application, it should be noted that unless otherwise prescribed and
It limits, term " installation ", " connected ", " connection " shall be understood in a broad sense, can be mechanical connection or electrical connection, be also possible to two
Connection inside element, can be directly connected, and "upper", "lower", "left", "right" etc. are only used for indicating relative positional relationship, when
The absolute position for being described object changes, then relative positional relationship may change;
Secondly: the present invention discloses in embodiment attached drawing, relates only to the structure being related to the embodiment of the present disclosure, other knots
Structure, which can refer to, to be commonly designed, and under not conflict situations, the same embodiment of the present invention and different embodiments be can be combined with each other;
Last: the foregoing is only a preferred embodiment of the present invention, is not intended to restrict the invention, all in the present invention
Spirit and principle within, any modification, equivalent replacement, improvement and so on, should be included in protection scope of the present invention it
It is interior.
Claims (8)
1. a kind of method of the book recommendation based on public data, it is characterised in that: specifically include following recommended method:
Step 1: internet book store's book data acquires, firstly, developing network crawler, from brilliance-Amazon, when some large sizes
Internet book store acquires book data, the network of incidence relation between one books of information structuring according to provided by internet book store,
Associated information includes " people for having bought the book also buys ... book ", " often buying ... book together ", " browsed the book between books
People also bought ... book " and user star evaluation, but do not include specific evaluation of the reader to books, these associations be regarded as
A kind of mutual recommendation relationship, these recommendation relationships can form a huge related network and adopt in addition to master data
Associated data between collection books, data are saved into database after over cleaning arranges;
Step 2: internet book store can provide the reading value assessment of the books of information, based on the data acquired in step 1, propose
A kind of assessment algorithm (hereinafter referred to as BookRank algorithm) enters the quality and quantity gone out with chain according to the chain of a books and measures
It reads value, specifically calculates the reading value (hereinafter referred to as BookRank) of every books, and the core of BookRank algorithm is thought
Road are as follows: each arrive the books link be to it primary approval ballot, 1) the higher books of degree of recognition it is more worth recommendation be
Be recognized the more worth recommendation of more books, 2) approved source book quality is higher, is approved by these high quality books
Books also more worth recommendation reads value more as soon as linked of book mean to be voted by other books much more
Height assigns its higher BookRank value at this time, for measuring the i.e. recommendable degree of its significance level;
It is give-and-take conditions from figure to provide information service according to the identity of reader Step 3: Feature of Library Collection book data acquires
Book shop obtains related data of the Feature of Library Collection books after desensitizing, and then excavates books by analysis reader conduct log sheet and reads
Internal association between person and between books is specially reserved, borrows and is renewed, by calculating the tribute between books and reader
The similar support between disturbance degree and books between weight, books is offered to measure the level of intimate of this relationship, will acquire
Characteristic books borrow data and be saved into database after over cleaning arranges;
Step 4: the reading value assessment of Feature of Library Collection books measures one according to following principles based on the data in step 3
The recommendation of Feature of Library Collection books is worth: 1) there are more more then its recommendation valences of the number of correlation behavior by books and authoritative higher reader
It is worth higher;2) there are the more more then more worth recommendations of the number of correlation behavior with more readers for books;3) two books simultaneously with it is more
The recommendation weight that reader exists simultaneously between the more more then books of reader's quantity of correlation behavior is higher;
The reading value of books is assessed according to the identity of reader, the other books many aspects borrowing quantity, borrowing:
Different readers are calculated from initial data first to the disturbance degree of different books classifications, reader conduct log of the reader A at oneself
In, the number of behavior incidence relation existing for certain class C to books is more, then reader A other to this books classification C comparison
The influence power of books classification is bigger;If the identity of reader A is more professional, more authoritative, then reader A is to this books classification C ratio
It is bigger to influence power of other readers to this books classification C;In order to measure different readers to figure belonging to its related books
The influence degree and readership technorati authority of book classification introduce " reader-classification " weight to the disturbance degree of books classification;
Then the recommendation weight between books is calculated, on the basis of previous step, if books X and books Y and multiple readers
Incidence relation is existed simultaneously, then books X is " reader-of these readers to classification belonging to books Y to the contribution weight of books Y
The weighted sum of classification " weight;Similarity between books X and books Y is that there are the readers of incidence relation with this books simultaneously
Quantity accounts for that there are the ratios of reader's quantity union of incidence relation with books X, books Y;So advowson of the books X to books Y
Value is product of the books X to the contribution weight of books Y and the similarity of books X and books Y;Present invention introduces pushing away between books
Recommend weight w (y → x), wherein only consider simultaneously with the related readership of books x and y;
Step 5: book recommendation strategy designs two kinds of book recommendation strategies on the basis of obtaining the BookRank value of books,
Generalization bounds respectively for the Generalization bounds of keyword retrieval and for certain this books, and the books recommended are pressed
BookRank value sorts be supplied to user from high to low;
For the Generalization bounds of search key: when user inputs keyword and retrieves, system is to the result being retrieved
Statistic of classification is carried out by the different books categorical measures to appearance of books classification, then according to each categorical measure in result
Difference, the Recommended Books number of each classification of reasonable distribution in proportion, and figure is selected from these classifications according to BookRank value
Book sorts after summarizing to these books by BookRank value;
For the Generalization bounds of specific books: when user selects a certain books, recommending one according to the classification information of the books
The related books of fixed number amount, thinking is: when user selects certain this books, it is meant that the interested books of the user may with
Selected books are in the books classification of same classification, so recommendable books should concentrate in the books of current class,
Herein using the category as selecting bibliography to record;In book database, all books are all sorted out by tree structure, and catalogue is
Always to classify, the directory name of catalogue is total classification number, and as the root node of tree structure, subdirectory is intermediate node, figure
Book is as leaf node;In this way, select bibliography record leaf lazy weight even without when, this can be selected bibliography record father's mesh
Record is set as currently bibliography being selected to record, and expands bibliography is selected to record step by step, is until the books selected reach recommendation number N as defined in system
Only;
So recommendable range can be extended to upper level catalogue when current class Books recommend insufficient, work as user
When searching a books, which can search books according to the search key that user inputs first, with the knot retrieved
There are fruit the books of content relevance to be presented to the user again according to specific sort algorithm;
The interface of books reading value and recommendation service is externally provided, by books reading value and associated recommendation information preservation rear
In platform database, an external group polling interface is then provided by network, provides relevant information service to each library.
2. a kind of method of book recommendation based on public data according to claim 1, it is characterised in that: the step
The iterative formula of BookRank is as follows in two:
BRn(A)=(1-d)+d × ∑n I=1(BRn-1(Ti)/C(Ti))
Wherein, BRn(A) the BookRank value of books A is represented;(BRn-1(Ti) represent books TiIn upper primary iteration
BookRank value, when books A appears in books TiRelated books list in, i.e. TiIt is linked to A;C(Ti) represent books Ti's
The total quantity of all books inside related books list;D represents damped coefficient;D × ∑ in above-mentioned formulan I=1(BRn-1(Ti)/
C(Ti)) indicate in related network by T during access chain random access booksiBy proportion in itself BookRank value
Each books in its related list are averagely allocated to for the part of d, therefore books A is obtained from books Ti1/C (Ti)
BookRank value.
3. a kind of method of book recommendation based on public data according to claim 2, it is characterised in that: the step
In two, in calculating process, algorithm after successive ignition, the BookRank value of every books be finally converged in one it is more stable
Value.
4. a kind of method of book recommendation based on public data according to claim 3, it is characterised in that: the step
In two, in algorithm realization, according to the Computing Principle of PageRank power method, BookRank is also calculated in this way,
To which be abstracted above-mentioned formula is converted into solutionValue, wherein x is the initial BookRank value of every books, just
Initial value can arbitrarily be set, it is set as 1 by the present invention;Matrix A are as follows:
A=(1-d) × eeT+d×PT
Wherein, d is damped coefficient, and e is n dimensional vector, eTRow vector, P are tieed up for nTFor probability transfer matrix [S
Krishnaswamy];In iterative calculation each time, matrix A is all changing, the following institute of dominant eigenvalue calculating process
Show:
(1) the initial BookRank value of x ← init//tax
(2) r ← Ax//r is BookRank value, and A is linking relationship matrix (as shown in formula 5.5)
(3) the BookRank value of if f (| | x-r | |) < ε, return r//compare before and after iteration
(4) else x ← r, goto (2) // and it is unsatisfactory for ε, continue iteration
Wherein, init is the initial BookRank value of setting, and x indicates current BookRank value, and r represents the BookRank after iteration
Value, ε indicate tolerance.
5. a kind of method of book recommendation based on public data according to claim 4, it is characterised in that: the step
In two, for the construction for constructing probability transfer matrix A, the link that books incidence relation linked network is expressed as n row (n+1) column is closed
It is matrix, convenient for the realization of system program.
6. a kind of method of book recommendation based on public data according to claim 1, it is characterised in that: the step
" reader-classification " weight is as follows in four:
Wherein i indicates that with books, c is indicated and class belonging to books of the reader i there are incidence relation there are the reader of incidence relation
Not, h (i, c) indicates that the books behavior number that reader i is c in the books classification of incidence relation, f (i) indicate that reader i is closed in association
The behavior number of the books of system, k indicate the weight value coefficient of readership authority, and value range is more authoritative between 0 to 1
Reader, rights relating the person authority weight values k is bigger, and λ (i, c) indicates that " reader-classification " weight, i.e. reader i have association to it and close
The affiliated books classification of the books of system is the disturbance degree of c class.
7. a kind of method of book recommendation based on public data according to claim 1, it is characterised in that: the step
In four, the recommendation weight between books is as follows:
Wherein i is with books x there are the reader of incidence relation, and j is reader's quantity with books x there are incidence relation, and c (x) is
Books classification belonging to books x, λ (j, c) indicate reader j to the contribution weight of books classification c, k indicate simultaneously with books x and figure
For book y there are reader's quantity of incidence relation, m indicates that with books x, n indicates books y, and there are associated there are associated reader's quantity
Reader's quantity.
8. a kind of method of book recommendation based on public data according to claim 1, it is characterised in that: the step
Book recommendation formula is as follows in four:
Wherein, RMD (W) indicates the total quantity recommended for search key W;N is indicated appeared in search result not
With the quantity of books classification;The total quantity of N expression search result;NiIndicate the books number for belonging to i-th of classification in search result
Amount;AiIndicate i-th of total quantity being sorted in entire book database;It indicates to be used for from i-th categorizing selection
The books quantitative range of recommendation;U is quantity coefficient, and value is between 0 to 1, to guarantee to be elected to be the books number of recommendation at such
In the reasonable range of a comparison, and the recommendation that M is then every class books sets a upper limit.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910066005.7A CN109918563B (en) | 2019-01-24 | 2019-01-24 | Book recommendation method based on public data |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910066005.7A CN109918563B (en) | 2019-01-24 | 2019-01-24 | Book recommendation method based on public data |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109918563A true CN109918563A (en) | 2019-06-21 |
CN109918563B CN109918563B (en) | 2023-10-20 |
Family
ID=66960668
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910066005.7A Active CN109918563B (en) | 2019-01-24 | 2019-01-24 | Book recommendation method based on public data |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109918563B (en) |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110674448A (en) * | 2019-09-27 | 2020-01-10 | 浙江树联智能科技有限公司 | Book pushing method and device |
CN112800341A (en) * | 2021-04-15 | 2021-05-14 | 广州华赛数据服务有限责任公司 | Education resource transmission system based on big data |
CN113505154A (en) * | 2021-05-31 | 2021-10-15 | 南京分布文化发展有限公司 | Digital reading statistical analysis method and system based on big data |
CN114398540A (en) * | 2021-12-17 | 2022-04-26 | 广州诺图计算机科技有限公司 | Intelligent book list recommendation method for book management |
CN115098803A (en) * | 2022-08-24 | 2022-09-23 | 深圳市华图测控系统有限公司 | Book recommendation algorithm and system based on mobile library |
CN116578726A (en) * | 2023-07-10 | 2023-08-11 | 悦读天下(北京)国际教育科技有限公司 | Personalized book recommendation system |
CN116720927A (en) * | 2023-08-08 | 2023-09-08 | 北京人天书店集团股份有限公司 | Book recommendation method, system and storage medium |
CN116911926A (en) * | 2023-06-26 | 2023-10-20 | 杭州火奴数据科技有限公司 | Advertisement marketing recommendation method based on data analysis |
CN116992093A (en) * | 2023-09-14 | 2023-11-03 | 东北农业大学 | Library intelligent indexing method, device and storage medium based on reader borrowing behaviors |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105183727A (en) * | 2014-05-29 | 2015-12-23 | 上海研深信息科技有限公司 | Method and system for recommending book |
CN106067134A (en) * | 2016-06-03 | 2016-11-02 | 朱志伟 | A kind of network self-service type books are recommended and are purchased and borrow method |
CN106202184A (en) * | 2016-06-27 | 2016-12-07 | 华中科技大学 | A kind of books personalized recommendation method towards libraries of the universities and system |
CN106897430A (en) * | 2017-02-27 | 2017-06-27 | 温州职业技术学院 | The implementation method of digital library |
-
2019
- 2019-01-24 CN CN201910066005.7A patent/CN109918563B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105183727A (en) * | 2014-05-29 | 2015-12-23 | 上海研深信息科技有限公司 | Method and system for recommending book |
CN106067134A (en) * | 2016-06-03 | 2016-11-02 | 朱志伟 | A kind of network self-service type books are recommended and are purchased and borrow method |
CN106202184A (en) * | 2016-06-27 | 2016-12-07 | 华中科技大学 | A kind of books personalized recommendation method towards libraries of the universities and system |
CN106897430A (en) * | 2017-02-27 | 2017-06-27 | 温州职业技术学院 | The implementation method of digital library |
Cited By (14)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110674448A (en) * | 2019-09-27 | 2020-01-10 | 浙江树联智能科技有限公司 | Book pushing method and device |
CN112800341A (en) * | 2021-04-15 | 2021-05-14 | 广州华赛数据服务有限责任公司 | Education resource transmission system based on big data |
CN112800341B (en) * | 2021-04-15 | 2021-06-22 | 广州华赛数据服务有限责任公司 | Education resource transmission system based on big data |
CN113505154A (en) * | 2021-05-31 | 2021-10-15 | 南京分布文化发展有限公司 | Digital reading statistical analysis method and system based on big data |
CN114398540A (en) * | 2021-12-17 | 2022-04-26 | 广州诺图计算机科技有限公司 | Intelligent book list recommendation method for book management |
CN114398540B (en) * | 2021-12-17 | 2024-04-26 | 广州诺图计算机科技有限公司 | Intelligent book order recommending method for book management |
CN115098803A (en) * | 2022-08-24 | 2022-09-23 | 深圳市华图测控系统有限公司 | Book recommendation algorithm and system based on mobile library |
CN116911926A (en) * | 2023-06-26 | 2023-10-20 | 杭州火奴数据科技有限公司 | Advertisement marketing recommendation method based on data analysis |
CN116578726B (en) * | 2023-07-10 | 2023-09-29 | 悦读天下(北京)国际教育科技有限公司 | Personalized book recommendation system |
CN116578726A (en) * | 2023-07-10 | 2023-08-11 | 悦读天下(北京)国际教育科技有限公司 | Personalized book recommendation system |
CN116720927A (en) * | 2023-08-08 | 2023-09-08 | 北京人天书店集团股份有限公司 | Book recommendation method, system and storage medium |
CN116720927B (en) * | 2023-08-08 | 2023-11-03 | 北京人天书店集团股份有限公司 | Book recommendation method, system and storage medium |
CN116992093A (en) * | 2023-09-14 | 2023-11-03 | 东北农业大学 | Library intelligent indexing method, device and storage medium based on reader borrowing behaviors |
CN116992093B (en) * | 2023-09-14 | 2024-05-28 | 东北农业大学 | Library intelligent indexing method, device and storage medium based on reader borrowing behaviors |
Also Published As
Publication number | Publication date |
---|---|
CN109918563B (en) | 2023-10-20 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109918563A (en) | A method of the book recommendation based on public data | |
CN102982042B (en) | A kind of personalization content recommendation method, platform and system | |
Wan et al. | Aminer: Search and mining of academic social networks | |
CN105224699B (en) | News recommendation method and device | |
KR102075833B1 (en) | Curation method and system for recommending of art contents | |
CN109934721A (en) | Finance product recommended method, device, equipment and storage medium | |
CN100401292C (en) | Systems and methods for search query processing using trend analysis | |
CN1882943B (en) | Systems and methods for search processing using superunits | |
CN110532479A (en) | A kind of information recommendation method, device and equipment | |
CN111708740A (en) | Mass search query log calculation analysis system based on cloud platform | |
CN106339502A (en) | Modeling recommendation method based on user behavior data fragmentation cluster | |
CN105787068B (en) | The academic recommended method and system analyzed based on citation network and user's proficiency | |
CN101111837A (en) | Search processing with automatic categorization of queries | |
CN103020164A (en) | Semantic search method based on multi-semantic analysis and personalized sequencing | |
CN109165367B (en) | News recommendation method based on RSS subscription | |
KR101088710B1 (en) | Method and Apparatus for Online Community Post Searching Based on Interactions between Online Community User and Computer Readable Recording Medium Storing Program thereof | |
CN109584003A (en) | Intelligent recommendation system | |
CN115408618B (en) | Point-of-interest recommendation method based on social relation fusion position dynamic popularity and geographic features | |
CN116595246A (en) | Book recommendation retrieval system based on knowledge graph and reader portrait | |
Yang | An active recommendation approach to improve book-acquisition process | |
Ding et al. | Clustering Merchants and Accurate Marketing of Products Using the Segmentation Tree Vector Space Model | |
Liu et al. | Understanding Consumer Preferences---Eliciting Topics from Online Q&A Community | |
Li et al. | A Method of Interest Degree Mining Based on Behavior Data Analysis | |
CN102495867B (en) | Online social network webpage searching method and webpage searching system | |
Wu et al. | Clustering technology application in e-commerce recommendation system |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |