CN103955535A - Individualized recommending method and system based on element path - Google Patents
Individualized recommending method and system based on element path Download PDFInfo
- Publication number
- CN103955535A CN103955535A CN201410204492.6A CN201410204492A CN103955535A CN 103955535 A CN103955535 A CN 103955535A CN 201410204492 A CN201410204492 A CN 201410204492A CN 103955535 A CN103955535 A CN 103955535A
- Authority
- CN
- China
- Prior art keywords
- user
- meta
- path
- information
- node
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 52
- 230000006399 behavior Effects 0.000 claims abstract description 26
- 238000005295 random walk Methods 0.000 claims abstract description 15
- 238000001914 filtration Methods 0.000 claims abstract description 10
- 238000010801 machine learning Methods 0.000 claims abstract description 5
- 230000002457 bidirectional effect Effects 0.000 claims description 20
- 238000004364 calculation method Methods 0.000 claims description 14
- 238000010586 diagram Methods 0.000 claims description 14
- 238000004422 calculation algorithm Methods 0.000 claims description 12
- 238000007781 pre-processing Methods 0.000 claims description 6
- 238000013507 mapping Methods 0.000 claims description 5
- 230000008569 process Effects 0.000 claims description 4
- 230000000694 effects Effects 0.000 abstract description 4
- 239000000654 additive Substances 0.000 abstract 1
- 230000000996 additive effect Effects 0.000 abstract 1
- 230000002349 favourable effect Effects 0.000 abstract 1
- 238000000691 measurement method Methods 0.000 abstract 1
- 239000011159 matrix material Substances 0.000 description 5
- 230000006870 function Effects 0.000 description 3
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000006855 networking Effects 0.000 description 2
- 230000008859 change Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000007418 data mining Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
Landscapes
- Engineering & Computer Science (AREA)
- Databases & Information Systems (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses an individualized recommending method and system based on an element path. The recommending method comprises the following steps: building a heterogeneous network graph through collecting article content information, social relationship among users and user historical behavior information; utilizing a relevance measurement method based on heterogeneous bi-directional random walk to measure the preference degrees of a user on articles based on semantics; utilizing machine learning to synthesize the preference degrees of the user under different semantics and predict the preference degrees of the user on unselected articles so as to recommend favorable articles to the user. According to the system, the accuracy, novelty and diversity of the recommending effect are effectively improved through comprehensively utilizing the additive information of other article types to a target article type based on the content, cooperation filtering and the social network relationship and digging the semantic features of the information.
Description
Technical Field
The invention relates to a meta-path-based personalized recommendation method and system, and belongs to the technical field of data mining and machine learning.
Background
With the rapid development of the Web2.0 technology and the social network, more and more information enters the daily life of people, and people are more and more difficult to obtain valuable information from the massive information. For this reason, recommendation systems aiming at helping users easily obtain useful information from massive information play an increasingly important role in people's daily life. The recommendation system mainly analyzes the user preference and tracks the interest change of the user according to the personal information of the user and the historical behavior of the object, thereby realizing the recommendation of the interested related information to the user and avoiding the trouble of searching information by the user. Mainstream recommendation systems mainly employ content-based and collaborative filtering-based methods.
The content-based recommendation method is the earliest adopted recommendation method, and recommendation is performed according to the similarity between document content and user interest. Because the similarity is calculated according to the content of the document, the generated recommendation result is often very specific and lacks novelty and diversity. Furthermore, the content-based recommendation method is only applicable to item recommendation in a document format, and cannot be applied to item recommendation requirements such as video and music.
The recommendation method based on collaborative filtering utilizes the relationship between users and articles to calculate the similarity between users and users or between articles and articles, so as to recommend the users according to the users or articles of the most similar neighbors. Because the method only considers the relation between the user and the article and does not consider the content of the article, the recommendation precision is reduced.
In the face of the shortcomings of the traditional recommendation system, how to improve the recommendation algorithm by comprehensively utilizing the content-based and collaborative filtering information becomes a problem to be solved. With the continuous socialization of networks, social relationships exist between users, and how to fully utilize the social relationships among the users to improve recommendation precision also becomes a technical problem. The traditional recommendation system only focuses on the target item type information, but ignores the improvement of the recommendation effect by other item type information, for example, a recommendation system of a movie, a user may like a movie adapted by a novel because of the novel that the user likes; it is also possible to play movies by a singer who likes it. Therefore, considering other non-target item type information can greatly increase the novelty and diversity of recommendation results. In addition, the traditional recommendation method rarely pays attention to the influence of semantics on the recommendation effect, and the accuracy of the recommendation system can be improved by introducing semantic information.
Therefore, there is a need to provide a recommendation method and system with semantic recognition by comprehensively utilizing additional information of other item types to target item types based on content, collaborative filtering, social networking relationships, so as to effectively improve the accuracy, novelty and diversity of recommendation.
Disclosure of Invention
The invention aims to provide a personalized recommendation method and system based on a meta path. The invention provides a framework containing information based on content, collaborative filtering, social relationship and the like based on the concept of the meta-path, introduces the concept of semantic features on the basis and fuses additional information of other articles to the target article, so that the recommendation precision is higher, and the article recommendation is more novel and diverse.
In order to achieve the above object, according to an aspect of the present invention, there is provided a meta path-based personalized recommendation method, including the steps of:
1) acquiring historical behavior information of a user on an article, social relation between the user and content information of the article;
2) constructing a heterogeneous network directed graph: constructing a heterogeneous network directed graph containing node types based on all the acquired information, wherein each node has a node type consistent with an actual physical type;
3) constructing a schematic diagram: the method comprises the steps of abstracting a heterogeneous network directed graph, only preserving type relations among nodes, and constructing a summary graph of the heterogeneous network directed graph based on node types and relations among the node types in the heterogeneous network directed graph;
4) acquiring a meta path: based on the sketch map, obtaining a meta-path containing different semantics by using a breadth-first traversal algorithm, wherein the meta-path is a one-way path, a source node of the meta-path is a user type, a tail node of the meta-path is an article type, the path specifies the behavior and the direction of node traversal, and different meta-paths correspond to different semantic information semantically;
5) and (3) correlation calculation: calculating the relevance between the user and the article under the meta-path by using a relevance calculation method based on asynchronous bidirectional random walk based on the meta-path;
6) generating a recommendation model: the importance weight of each meta path is learned through a machine, a recommendation model is generated by combining the meta path, and the preference degree of the user to the articles is calculated;
7) generating a recommended item list: and according to the preference degree of the user on the unselected items predicted by the recommendation model, sorting in a descending manner, and recommending one or more items ranked at the top to the user.
In the recommendation method, the information obtaining step includes: if the social relationship between the users exists in the application system, the information is acquired, and if the social relationship does not exist, the information does not need to be acquired; the historical behavior information of the user on the resource is not limited to the scoring information, and can be any other useful information, such as the number of clicks of the user on a certain webpage/the number of purchases of a certain article.
In the above recommendation method, between the information obtaining step and the step of constructing the digraph of the heterogeneous network, the method further includes: the method comprises the following steps of preprocessing historical behavior data of a user: and information which is negative (disliked) to the resource by the user is filtered, and only positive (favorite) information is reserved. In the user scoring system, only the user scores above the middle score in the scoring system are kept as positive information. In a non-scoring system, both a user's click on a web page and a purchase of an item are positive information.
In the above recommendation method, the step of constructing the heterogeneous network directed graph specifically includes: firstly, mapping an object entity and an attribute value thereof to corresponding nodes in a graph, wherein if a movie is an entity, and an actor/director/style is the attribute of the movie; and using the type of the entity or attribute as the type of the corresponding node; adding bidirectional edges between corresponding nodes according to attribute values owned by the entities; secondly, if the user has positive behavior information on the object entity, adding a bidirectional edge between the user and the entity; and finally, if the user is interested by other users or fans of the users, namely the user has social relations, adding a one-way edge pointing to other user nodes from the user node. Each node in the heterogeneous network directed graph has a node type consistent with an actual physical type.
In the recommendation method, the step of constructing the sketch specifically comprises: abstracting the heterogeneous network directed graph, and only preserving the type relation among the nodes. In the summary graph, a node is a general name of a series of nodes of the same type, and indicates a topological structure relationship of the heterogeneous network directed graph.
In the recommendation method, the meta path obtaining step specifically includes: and acquiring a plurality of meta-paths from the user type to the item entity type on the overview chart by using a breadth-first chart searching method. Wherein the meta-path is a one-way path whose source node is the user type and end node is the item type, the path specifying the behavior and direction of traversal of the node. Semantically, different meta-paths correspond to different semantic information. The meta-path includes the remaining item type nodes in addition to the target item type node. In addition, an upper limit of the required meta-path length needs to be specified, and this upper limit is used as an end condition of the meta-path acquisition method.
In the recommendation method, the step of calculating the correlation degree based on the asynchronous bidirectional random walk specifically includes:
1) and respectively starting from a source node and a tail node of the path according to the meta-path, wherein the source node randomly walks along the appointed meta-path, the tail node randomly walks along the meta-path in a reverse direction, and the probability of reaching the corresponding node of each position on the meta-path is calculated.
2) The probability of reaching the same node is normalized by the cosine law and is taken as the meeting probability of two nodes.
3) And adopting the encountering probability of the source node and the tail node at each position of the meta-path as the correlation degree of the source node and the tail node by arithmetic mean.
In the recommendation method, in the step of generating the model, a ranking learning method is adopted to learn the importance weights of the various paths, so as to generate the recommendation model.
In addition, the present invention provides a meta path-based personalized recommendation system, comprising: the information acquisition module is used for acquiring historical behavior information of the user on the article, social relations between the user and attribute information of the article; the preprocessing module is used for filtering the information that the user rejects (dislikes) the resources and only keeping the information that the user affirms (likes); the graph establishing module is used for establishing a heterogeneous network directed graph containing node types and a schematic graph thereof based on historical behavior information of users on articles, social relations among the users and attribute information of the articles; the recommendation model generation module is used for predicting the preference degree of the user to the articles by utilizing a correlation calculation method based on asynchronous bidirectional random walk based on the heterogeneous network directed graph and the schematic graph thereof; and the recommending module is used for sorting in a descending mode according to the preference degree of the user on the unselected items predicted by the recommending model and recommending one or more items ranked most front to the user.
In the recommendation system, the graph building module includes: the heterogeneous network directed graph establishing module is used for establishing a heterogeneous network directed graph containing node types based on historical behavior information of users on articles, social relations among the users and attribute information of the articles; and the overview establishing module is used for establishing an overview based on the node types in the heterogeneous network directed graph and the relationship among the node types.
In the recommendation system, the recommendation model generation module includes: the meta-path acquisition module acquires meta-paths containing different semantics on the schematic diagram by using a breadth-first traversal algorithm; the relevancy calculation module is used for calculating the relevancy of the user and the article under the meta-path by utilizing the relevancy algorithm based on the asynchronous bidirectional random walk according to the meta-path; and the weight learning module learns the importance weight of the meta path based on the meta path and the correlation thereof.
The invention provides a meta-path-based personalized recommendation method and system. The relevance calculating method based on asynchronous bidirectional random walk is used for the first time, the preference of a user to unselected articles is calculated more accurately based on content, collaborative filtering and social network relation information, and the recommendation accuracy is improved more effectively. Meanwhile, by utilizing the concept of meta-path, the characteristic information based on semantics is introduced, and the item preference of the user can be dug more deeply by adding the related information of the non-recommended target item, so that the novelty and diversity of the recommendation result are effectively increased.
Drawings
FIG. 1 is a flowchart illustrating steps of a meta-path-based personalized recommendation method according to the present invention;
FIG. 2 is a schematic diagram illustrating a scenario description according to an embodiment of the present invention;
FIG. 3 is a schematic diagram of an embodiment of the present invention;
FIG. 4 is a meta-path diagram according to an embodiment of the present invention;
fig. 5 is a schematic structural diagram of the meta-path-based personalized recommendation system of the present invention.
Detailed Description
The invention provides a meta-path-based personalized recommendation method and system. The following detailed description is made with reference to the accompanying drawings.
Referring to fig. 1, fig. 1 is a flowchart illustrating steps of a meta-path-based personalized recommendation method according to the present invention, including the following steps:
an information acquisition step 101: acquiring historical behavior information of a user on an article, social relations between the user and attribute information of the article;
constructing a heterogeneous network directed graph step 102: constructing a heterogeneous network directed graph containing node types based on all the acquired information;
step 103 of constructing a summary chart: constructing a summary graph of the heterogeneous network directed graph based on the node types and the relations among the node types in the heterogeneous network directed graph;
meta-path acquisition step 104: based on the schematic diagram, acquiring meta-paths containing different semantics by using a breadth-first traversal algorithm;
correlation calculation step 105: based on the meta path, calculating the relevance between the user and the article under the meta path by using the relevance calculation method based on the asynchronous bidirectional random walk provided by the invention;
generating a recommendation model step 106: the importance weight of each meta path is learned through a machine, a recommendation model is generated by combining the meta path, and the preference degree of the user to the articles is calculated;
generating a list of recommended items step 107: and according to the preference degree of the user on the unselected items predicted by the recommendation model, sorting in a descending manner, and recommending one or more items ranked at the top to the user.
In practice, a preferable processing mode is that a preprocessing step is further included before the step of constructing the heterogeneous network directed graph, the user historical behavior information is processed, information that the user rejects (dislikes) the article is filtered, and only positive (favorite) information is kept. In the user scoring system, only the user scores above the middle score in the scoring system are kept as positive information. In a non-scoring system, both a user's click on a web page and a purchase of an item are positive information.
The following describes the embodiments of the method in detail with reference to an example of a social networking platform sharing media information. Referring to fig. 2, fig. 2 is a scene description diagram of the present embodiment, which describes a social network in which a user shares three types of articles, i.e., movies, music, and books, each type of article has tag information provided by the user and content information of the article, such as a lead actor of a movie, a singer of music, and an author of a book, and the users can pay attention to each other, and the users can score the viewed movies, music, and books by 1-5 points. In the present embodiment, movie recommendation is targeted.
Firstly, before constructing a heterogeneous network directed graph, the historical behavior data of a user needs to be preprocessed, in this embodiment, scores of movie scores of the user are specifically 1, 1.5, 2, 2.5, 3, 3.5, 4, 4.5, and 5, a score (3.5, 4, 4.5, and 5) greater than 3 is selected as an item liked by the user, and score information less than or equal to 3 is filtered. If in the system without scoring, all the behavior information of the users and the articles, such as shopping and webpage clicking, is considered as favorite by the users. According to the preference of the user to the item, a user-item relation matrix R can be constructed, and is specifically described as formula (1).
Where R is a matrix of dimensions m x n, m being the number of users in the system and n being the number of items in the system.
And then constructing a heterogeneous network directed graph according to the user-article relation matrix and the social relation between the content information of the article and the user. A heterogeneous network directed graph refers to a directed graphWhere the node V ∈ V represents an object (e.g., movie, user, label), there is a directed edge < u, and V > ∈ E indicates that there is a relationship from node u to node V. A represents a set of object types, and the mapping function φ V → A represents that each object V ∈ V belongs to a particular object type φ (V) ∈ A. R represents a set of link relation types, mapping functionIndicates that each directed edge < u, v ∈ E belongs to a special characterSpecified link relation typeIf the number of object types | A | > 1 or the number of link relationship types | R | > 1, the network type is called a heterogeneous network directed graph, otherwise, the network type is called a homogeneous network directed graph. Note that if a relationship R exists from type A to type B, it is denoted asNaturally, there will also be an inverse relationship R-Exist inMost often, R is inversely related to R-Are not equivalent.
According to the definition of the heterogeneous network directed graph, the specific construction process is as follows: firstly, mapping an object entity and an attribute value thereof to corresponding nodes in a graph, wherein if a movie is an entity, and an actor/director/style is the attribute of the movie; and using the type of the entity or attribute as the type of the corresponding node; adding bidirectional edges between corresponding nodes according to attribute values owned by the entities; secondly, if the user has positive behavior information on the object entity, adding a bidirectional edge between the user and the entity; finally, if the user is interested in or fans of other users, adding a one-way edge pointing to other user nodes from the user node.
And then, according to the object relationship of the heterogeneous network directed graph, the object type is used as a node, the link relationship type is used as an edge, and the schematic graph is constructed. Fig. 3 is a schematic diagram of fig. 2 depicting the relationship of these object types, here with participants as the collective term for the actor, singer, author, as in fig. 3.
In the heterogeneous network directed graph, two nodes can be connected through different types of paths, for example, a user and a movie can be connected through 'user-movie-actor-movie', 'user-movie', 'user-movie-tag-movie', and the like. Obviously, these different path types imply different semantic information, and for this reason, these path types are represented by meta-paths, which are defined as:
a meta path P is defined in the summary diagram TGPath on (A, R), in the form ofIf there is no redundant type relationship between the same object types, the representation of the meta-path can be simplified by using only the object types, i.e. P ═ a (a)0A1...Al+1). Referring to FIG. 4, FIG. 4 is a meta pathMay be abbreviated as "user-movie" and "user-movie-actor-movie".
The meta-path obtaining method is that for two given object types, such as user and movie, a breadth-first traversal algorithm can be adopted to start from a user type node to end from a movie type node on the sketch. Since the traversal algorithm does not end naturally due to the existence of the bidirectional edge in the sketch, the path length needs to be specified as a constraint condition, and the traversal is stopped when the traversed path exceeds the length.
The correlation calculation method based on asynchronous bidirectional random walk is described below. For a given meta-path 'user-movie-actor-movie', starting from a source node user node and a tail node movie node of the path respectively, the source node randomly walks along the specified meta-path, the tail node randomly walks along the meta-path in the reverse direction, and the probability of reaching the corresponding node of each position on the meta-path is calculated. First, the probability that the source node and the end node meet at the same point is introduced, as shown in formula (2):
wherein,for the meta path, the source node u belongs to the A0And the end node v ∈ AlThe type of the encountered node is AiProbability of a node. The function Prob in equation (2) refers to a random walk along the meta-path in a specified direction, specifically: prob (u, v | P) refers to the probability of arriving at node v along meta-path P, starting from source node u; prob (u, v | P)-1) Refers to the probability of arriving at node v starting from the last node u and following the meta-path P in reverse. Specifically, the formula (3) and (4):
wherein, O (u | A)0A1) Is u is based on a relationThe out-of-range neighbor node set; i (u | A)t-1At) Is u is based on a relationThe set of in-degree neighbor nodes.
When randomly walking to the target node, Prob is defined as formula (5):
the nodes u and v are of the same type A, the delta (u and v) is an indication function, the value is 1 when the nodes u and v are of the same node, and the value is 0 otherwise.
Since the number of nodes of different object types is different, the probability of meeting different object types is also different, and for this reason, formula (1) is normalized by using the cosine theorem, as specifically described in formula (6):
equation (6) is to compute the source node and the last node at a certain node A in the meta-pathiThe probability of meeting. By using the formula (6), the total probability of the source node and the end node meeting at each node of the meta-path can be calculated by adopting an arithmetic mean method, and the total probability is taken as the correlation degree of the source node and the end node, which is specifically described as the formula (7):
wherein l is the meta pathLength of (d).
The steps of generating the recommendation model are described in detail below. Assuming that meta-paths between L users and articles are obtained through the meta-path obtaining step, L meta-paths based on the meta-paths can be calculated according to the meta-paths to obtain L correlation degree matrixes S of the users and the articles based on the meta-paths1,S2,...,SL. Since different meta-paths contain different semantics, the parameter θ is usediTo indicate the importance of each meta path. Thus, a recommendation model as described in equation (8) is obtained:
wherein,is a matrix of the relevance of the user to the item, each element of whichAnd the relevance of the user i and the item j is shown, namely the preference degree of the user i for the item j.
The importance weight parameter θ i in equation (8) can be obtained by machine learning. Since the user-item relationship matrix (shown in equation 1) contains only the values 1 and 0, a machine learning method of rank learning is preferably employed.
According to the recommendation model of the formula (8), the predicted preference degrees of the user on the unselected items are sorted in a descending manner, and one or more items which are arranged at the top are recommended to the user to generate a list.
Preferably, although movie recommendations are targeted, the computed meta-path also contains non-recommended object types and their related information, such as music and books. In the conventional recommendation system, only information related to the recommendation target item is focused, and information of the non-recommendation target item is ignored, however, the information of the non-recommendation target item can also improve the recommendation effect, for example, a user may like a movie adapted by the novel because of the novel that the user likes; it may also be that the film is liked by the singer who is liked by himself. Therefore, the invention can improve the recommendation accuracy by utilizing the non-target item information and also can increase the novelty and diversity of recommendation.
Next, a meta path-based personalized recommendation system is described with reference to fig. 5. As shown in fig. 5, the recommendation system specifically includes: the information acquisition module 501 is configured to acquire historical behavior information of a user on an article, social relationships between the user and the user, and attribute information of the article; a preprocessing module 502, configured to filter out negative (disliked) information of the resource by the user, and only retain positive (liked) information; the graph establishing module 503 is configured to establish a heterogeneous network directed graph containing node types and a schematic graph thereof based on historical behavior information of the users on the articles, social relationships among the users and attribute information of the articles themselves; a recommendation model generation module 504, configured to predict, based on the heterogeneous network directed graph and a schematic diagram thereof, a preference degree of a user for an item by using a correlation calculation method based on asynchronous bidirectional random walk; and the recommending module 505 is used for sorting in a descending manner according to the preference degree of the user on the unselected items predicted by the recommending model, and recommending one or more items ranked most front to the user.
As shown in fig. 5, the graph establishing module 503 includes: the heterogeneous network directed graph establishing module 503a is configured to establish a heterogeneous network directed graph containing node types based on historical behavior information of users on articles, social relationships among the users and attribute information of the articles; the profile establishing module 503b constructs a profile based on the node types and the relationships between the node types in the heterogeneous network directed graph.
The recommendation model generation module 504 includes: a meta-path obtaining module 504a, which obtains meta-paths containing different semantics on the schematic diagram by using a breadth-first traversal algorithm; the relevance calculating module 504b is used for calculating the relevance between the user and the article under the meta-path by utilizing the asynchronous bidirectional random walk algorithm according to the meta-path; the weight learning module 504c learns the importance weight of the meta path based on the meta path and the correlation thereof.
Therefore, the personalized recommendation system based on the meta-path is realized.
It is noted that the principles of the system embodiments are the same as the principles of the method embodiments, and may be referred to one another. And will not be described in detail herein.
Although the embodiments of the present invention have been described in connection with the accompanying drawings, it should not be construed as limiting the invention. It should be noted that, for those skilled in the art, several modifications and variations can be made without departing from the principle and scope of the present invention, and these modifications and variations should be considered to be within the protection scope of the present invention.
Claims (10)
1. A personalized recommendation method based on meta-path is characterized by comprising the following steps:
1) information acquisition: acquiring historical behavior information of a user on an article, social relations between the user and content information of the article;
2) constructing a heterogeneous network directed graph: constructing a heterogeneous network directed graph containing node types based on all the acquired information, wherein each node has a node type consistent with an actual physical type;
3) constructing a schematic diagram: the method comprises the steps of abstracting a heterogeneous network directed graph, only preserving type relations among nodes, and constructing a summary graph of the heterogeneous network directed graph based on node types and relations among the node types in the heterogeneous network directed graph;
4) acquiring a meta path: based on the sketch map, obtaining a meta-path containing different semantics by using a breadth-first traversal algorithm, wherein the meta-path is a one-way path, a source node of the meta-path is a user type, a tail node of the meta-path is an article type, the path specifies the behavior and the direction of node traversal, and different meta-paths correspond to different semantic information semantically;
5) and (3) correlation calculation: calculating the relevance between the user and the article under the meta-path by using a relevance calculation method based on asynchronous bidirectional random walk based on the meta-path;
6) generating a recommendation model: the importance weight of each meta path is learned through a machine, a recommendation model is generated by combining the meta path, and the preference degree of the user to the articles is calculated;
7) generating a recommended item list: and according to the preference degree of the user on the unselected items predicted by the recommendation model, sorting in a descending manner, and recommending one or more items ranked at the top to the user.
2. The personalized recommendation method according to claim 1, wherein in the information acquisition process of step 1), for the social relationship between the user and the user, if the social relationship exists in the application system, the information is acquired; if not, acquisition is not needed.
3. The personalized recommendation method according to claim 1, wherein between the step 1) of obtaining information and the step 2) of constructing a heterogeneous network directed graph, the method further comprises a step of preprocessing user historical behavior data, specifically: filtering out negative information of the user on the article, and only keeping positive information; for the user scoring system, the user scores above the middle score are positive information; in a non-scoring system, both a user's click on a web page and a purchase of an item are positive information.
4. The personalized recommendation method according to claim 1, 2 or 3, wherein the specific process of constructing the heterogeneous network directed graph in step 2) is as follows:
the method comprises the steps of firstly mapping an object entity and a content attribute value thereof to corresponding nodes in a graph, using the entity or the type of the attribute value as the type of the corresponding node, adding bidirectional edges between the corresponding nodes according to the attribute value owned by the entity, secondly adding the bidirectional edges between a user and the entity if the user has positive behavior information on the object entity, and finally adding unidirectional edges pointing to other user nodes from the user node if the user is interested in other users or fans of the users, namely social relations exist.
5. The personalized recommendation method according to claim 4, wherein in the step 4) meta-path acquisition, an upper limit of the required meta-path length needs to be specified, and the upper limit is used as an end condition of meta-path acquisition.
6. The personalized recommendation method according to claim 5, wherein in the correlation calculation in step 5), the specific process of the correlation calculation method based on asynchronous bidirectional random walk is as follows:
1) respectively starting from a source node and a tail node of a path according to a specified meta-path, wherein the source node randomly walks along the specified meta-path, the tail node randomly walks along the meta-path in a reverse direction, and the probability of reaching the corresponding node of each position on the meta-path is calculated;
2) normalizing the probability of reaching the same node by using a cosine law, and taking the normalized probability as the meeting probability of two nodes;
3) and adopting the encountering probability of the source node and the tail node at each position of the meta-path as the correlation degree of the source node and the tail node by arithmetic mean.
7. The personalized recommendation method according to claim 1, wherein the machine learning method in the step of generating the recommendation model in step 6) is a ranking learning method.
8. A meta-path based personalized recommendation system, the system comprising:
the information acquisition module is used for acquiring historical behavior information of the user on the article, social relations between the user and attribute information of the article;
the preprocessing module is used for filtering out negative information of a user on an article and only keeping positive information;
the graph establishing module is used for establishing a heterogeneous network directed graph containing node types and a schematic graph thereof based on historical behavior information of users on articles, social relations among the users and attribute information of the articles;
the recommendation model generation module is used for predicting the preference degree of the user to the articles by utilizing a correlation calculation method based on asynchronous bidirectional random walk based on the heterogeneous network directed graph and the schematic graph thereof;
and the recommending module is used for sorting in a descending mode according to the preference degree of the user on the unselected items predicted by the recommending model and recommending one or more items ranked most front to the user.
9. The personalized recommendation system of claim 8, wherein the graph building module comprises:
1) the heterogeneous network directed graph establishing module is used for establishing a heterogeneous network directed graph containing node types based on historical behavior information of users on articles, social relations among the users and attribute information of the articles;
2) and the overview establishing module is used for establishing an overview based on the node types in the heterogeneous network directed graph and the relationship among the node types.
10. The personalized recommendation system according to claim 8 or 9, wherein the recommendation model generation module comprises:
1) the meta-path acquisition module acquires meta-paths containing different semantics on the schematic diagram by using a breadth-first traversal algorithm;
2) the relevancy calculation module is used for calculating the relevancy of the user and the article under the meta-path by utilizing the relevancy calculation algorithm based on the asynchronous bidirectional random walk according to the meta-path;
3) and the weight learning module learns the importance weight of the meta path based on the meta path and the correlation thereof.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410204492.6A CN103955535A (en) | 2014-05-14 | 2014-05-14 | Individualized recommending method and system based on element path |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410204492.6A CN103955535A (en) | 2014-05-14 | 2014-05-14 | Individualized recommending method and system based on element path |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN103955535A true CN103955535A (en) | 2014-07-30 |
Family
ID=51332810
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410204492.6A Pending CN103955535A (en) | 2014-05-14 | 2014-05-14 | Individualized recommending method and system based on element path |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103955535A (en) |
Cited By (38)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105913125A (en) * | 2016-04-12 | 2016-08-31 | 北京邮电大学 | Heterogeneous information network element determining method, link prediction method, heterogeneous information network element determining device and link prediction device |
| CN105913296A (en) * | 2016-04-01 | 2016-08-31 | 北京理工大学 | Customized recommendation method based on graphs |
| CN106156483A (en) * | 2016-01-18 | 2016-11-23 | 李雪 | A kind of risk evaluating method based on data in literature, device and server |
| CN106202502A (en) * | 2016-07-20 | 2016-12-07 | 福州大学 | In music information network, user interest finds method |
| WO2017041541A1 (en) * | 2015-09-08 | 2017-03-16 | 北京邮电大学 | Method for pushing recommendation information, and server and storage medium |
| CN106708938A (en) * | 2016-11-18 | 2017-05-24 | 北京大米科技有限公司 | Method and device for assisting recommendation |
| CN106802956A (en) * | 2017-01-19 | 2017-06-06 | 山东大学 | A kind of film based on weighting Heterogeneous Information network recommends method |
| CN106934071A (en) * | 2017-04-27 | 2017-07-07 | 北京大学 | Recommendation method and device based on Heterogeneous Information network and Bayes's personalized ordering |
| CN107391577A (en) * | 2017-06-20 | 2017-11-24 | 中国科学院计算技术研究所 | A kind of works label recommendation method and system based on expression vector |
| CN107515943A (en) * | 2017-09-01 | 2017-12-26 | 安徽教育网络出版有限公司 | User preferences modeling construction method based on collaborative filtering |
| CN109034960A (en) * | 2018-07-12 | 2018-12-18 | 电子科技大学 | A method of more inferred from attributes based on user node insertion |
| CN109299373A (en) * | 2018-10-20 | 2019-02-01 | 上海交通大学 | Recommender system based on figure convolution technique |
| CN109801077A (en) * | 2019-01-21 | 2019-05-24 | 北京邮电大学 | A kind of arbitrage user detection method, device and equipment |
| CN109947870A (en) * | 2019-03-26 | 2019-06-28 | 第四范式（北京）技术有限公司 | The prediction meanss and method of specific type personnel calculate equipment and storage medium |
| CN110020214A (en) * | 2019-04-08 | 2019-07-16 | 北京航空航天大学 | A kind of social networks streaming events detection system merging knowledge |
| CN110083766A (en) * | 2019-04-26 | 2019-08-02 | 北京邮电大学 | A kind of inquiry recommended method and device based on the insertion of first Route guiding |
| CN110119478A (en) * | 2019-05-15 | 2019-08-13 | 南京大学 | A kind of item recommendation method based on similarity of a variety of user feedback datas of combination |
| CN110309404A (en) * | 2018-03-08 | 2019-10-08 | 优酷网络技术（北京）有限公司 | Content recommendation method and device |
| CN110334134A (en) * | 2019-07-17 | 2019-10-15 | 中国人民解放军国防科技大学 | Heterogeneous information network capability node importance degree evaluation method based on meta-path |
| CN110555050A (en) * | 2018-03-30 | 2019-12-10 | 华东师范大学 | heterogeneous network node representation learning method based on meta-path |
| CN110659382A (en) * | 2019-09-04 | 2020-01-07 | 杭州电子科技大学 | Mixed music recommendation method based on heterogeneous information network representation learning technology |
| CN110750598A (en) * | 2019-09-18 | 2020-02-04 | 精锐视觉智能科技（深圳）有限公司 | Method and device for predicting article label, terminal equipment and storage medium |
| CN110929046A (en) * | 2019-12-10 | 2020-03-27 | 华中师范大学 | Knowledge entity recommendation method and system based on heterogeneous network embedding |
| CN111222053A (en) * | 2019-11-27 | 2020-06-02 | 腾讯音乐娱乐科技（深圳）有限公司 | Object recommendation method and device and related equipment |
| CN111462897A (en) * | 2020-04-01 | 2020-07-28 | 山东大学 | Patient similarity analysis method and system based on improved heterogeneous information network |
| CN111507788A (en) * | 2019-01-31 | 2020-08-07 | 阿里巴巴集团控股有限公司 | Data recommendation method and device, storage medium and processor |
| CN111522889A (en) * | 2020-04-24 | 2020-08-11 | 腾讯科技（深圳）有限公司 | User interest tag expansion method and device, electronic equipment and storage medium |
| CN111538906A (en) * | 2020-05-29 | 2020-08-14 | 支付宝(杭州)信息技术有限公司 | Information pushing method and device based on privacy protection |
| CN112016701A (en) * | 2020-09-09 | 2020-12-01 | 四川大学 | Abnormal change detection method and system integrating time sequence and attribute behaviors |
| CN112221159A (en) * | 2020-10-21 | 2021-01-15 | 腾讯科技（深圳）有限公司 | Virtual item recommendation method and device and computer readable storage medium |
| CN112541806A (en) * | 2020-12-18 | 2021-03-23 | 北京航天云路有限公司 | Recommendation method and device based on heterogeneous information network |
| CN113239271A (en) * | 2021-05-11 | 2021-08-10 | 四川大学 | Recommendation method based on interest drift |
| WO2021208583A1 (en) * | 2020-04-16 | 2021-10-21 | 北京三快在线科技有限公司 | Recommendation information generation method and apparatus, electronic device and readable storage medium |
| CN113935804A (en) * | 2021-10-15 | 2022-01-14 | 燕山大学 | Semantic recommendation method based on reinforcement learning and weighted element path |
| CN113961693A (en) * | 2021-09-15 | 2022-01-21 | 北京三快在线科技有限公司 | Search result recommendation method and device, electronic device and readable storage medium |
| US11250347B2 (en) | 2018-06-27 | 2022-02-15 | Microsoft Technology Licensing, Llc | Personalization enhanced recommendation models |
| CN114637888A (en) * | 2022-05-18 | 2022-06-17 | 深圳市华曦达科技股份有限公司 | Video pushing method and device |
| CN116089730A (en) * | 2023-04-07 | 2023-05-09 | 深圳市活力天汇科技股份有限公司 | Travel route recommending method |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20050193002A1 (en) * | 2004-02-26 | 2005-09-01 | Yahoo! Inc. | Method and system for generating recommendations |
| US20080077574A1 (en) * | 2006-09-22 | 2008-03-27 | John Nicholas Gross | Topic Based Recommender System & Methods |
| CN101196904A (en) * | 2007-11-09 | 2008-06-11 | 清华大学 | News keyword abstraction method based on word frequency and multi-component grammar |
| CN101321190A (en) * | 2008-07-04 | 2008-12-10 | 清华大学 | Recommend method and recommend system of heterogeneous network |
-
2014
- 2014-05-14 CN CN201410204492.6A patent/CN103955535A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20050193002A1 (en) * | 2004-02-26 | 2005-09-01 | Yahoo! Inc. | Method and system for generating recommendations |
| US20080077574A1 (en) * | 2006-09-22 | 2008-03-27 | John Nicholas Gross | Topic Based Recommender System & Methods |
| CN101196904A (en) * | 2007-11-09 | 2008-06-11 | 清华大学 | News keyword abstraction method based on word frequency and multi-component grammar |
| CN101321190A (en) * | 2008-07-04 | 2008-12-10 | 清华大学 | Recommend method and recommend system of heterogeneous network |
Non-Patent Citations (1)
| Title |
|---|
| YIZHOU SUN ET AL: "PathSim: Meta Path-Based Top-K Similarity Search in Heterogeneous Information Networks", 《PROCEEDINGS OF THE VIDB ENDOWMENT》 * |
Cited By (62)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10609433B2 (en) | 2015-09-08 | 2020-03-31 | Tencent Technology (Shenzhen) Company Limited | Recommendation information pushing method, server, and storage medium |
| WO2017041541A1 (en) * | 2015-09-08 | 2017-03-16 | 北京邮电大学 | Method for pushing recommendation information, and server and storage medium |
| CN106156483B (en) * | 2016-01-18 | 2018-10-02 | 李雪 | A kind of risk evaluating method, device and server based on data in literature |
| CN106156483A (en) * | 2016-01-18 | 2016-11-23 | 李雪 | A kind of risk evaluating method based on data in literature, device and server |
| CN105913296A (en) * | 2016-04-01 | 2016-08-31 | 北京理工大学 | Customized recommendation method based on graphs |
| CN105913125A (en) * | 2016-04-12 | 2016-08-31 | 北京邮电大学 | Heterogeneous information network element determining method, link prediction method, heterogeneous information network element determining device and link prediction device |
| CN105913125B (en) * | 2016-04-12 | 2018-05-25 | 北京邮电大学 | Heterogeneous information network element path determines, link prediction method and device |
| CN106202502B (en) * | 2016-07-20 | 2020-02-07 | 福州大学 | User interest discovery method in music information network |
| CN106202502A (en) * | 2016-07-20 | 2016-12-07 | 福州大学 | In music information network, user interest finds method |
| CN106708938A (en) * | 2016-11-18 | 2017-05-24 | 北京大米科技有限公司 | Method and device for assisting recommendation |
| CN106802956A (en) * | 2017-01-19 | 2017-06-06 | 山东大学 | A kind of film based on weighting Heterogeneous Information network recommends method |
| CN106802956B (en) * | 2017-01-19 | 2020-06-05 | 山东大学 | Movie recommendation method based on weighted heterogeneous information network |
| CN106934071A (en) * | 2017-04-27 | 2017-07-07 | 北京大学 | Recommendation method and device based on Heterogeneous Information network and Bayes's personalized ordering |
| CN107391577A (en) * | 2017-06-20 | 2017-11-24 | 中国科学院计算技术研究所 | A kind of works label recommendation method and system based on expression vector |
| CN107391577B (en) * | 2017-06-20 | 2020-04-03 | 中国科学院计算技术研究所 | Work label recommendation method and system based on expression vector |
| CN107515943A (en) * | 2017-09-01 | 2017-12-26 | 安徽教育网络出版有限公司 | User preferences modeling construction method based on collaborative filtering |
| CN110309404A (en) * | 2018-03-08 | 2019-10-08 | 优酷网络技术（北京）有限公司 | Content recommendation method and device |
| CN110555050B (en) * | 2018-03-30 | 2023-03-31 | 华东师范大学 | Heterogeneous network node representation learning method based on meta-path |
| CN110555050A (en) * | 2018-03-30 | 2019-12-10 | 华东师范大学 | heterogeneous network node representation learning method based on meta-path |
| US11250347B2 (en) | 2018-06-27 | 2022-02-15 | Microsoft Technology Licensing, Llc | Personalization enhanced recommendation models |
| CN109034960A (en) * | 2018-07-12 | 2018-12-18 | 电子科技大学 | A method of more inferred from attributes based on user node insertion |
| CN109034960B (en) * | 2018-07-12 | 2022-09-06 | 电子科技大学 | Multi-attribute inference method based on user node embedding |
| CN109299373B (en) * | 2018-10-20 | 2021-10-29 | 上海交通大学 | Recommendation system based on graph convolution technology |
| CN109299373A (en) * | 2018-10-20 | 2019-02-01 | 上海交通大学 | Recommender system based on figure convolution technique |
| CN109801077A (en) * | 2019-01-21 | 2019-05-24 | 北京邮电大学 | A kind of arbitrage user detection method, device and equipment |
| CN111507788A (en) * | 2019-01-31 | 2020-08-07 | 阿里巴巴集团控股有限公司 | Data recommendation method and device, storage medium and processor |
| CN111507788B (en) * | 2019-01-31 | 2023-07-14 | 阿里巴巴华北技术有限公司 | Data recommendation method, device, storage medium and processor |
| CN109947870A (en) * | 2019-03-26 | 2019-06-28 | 第四范式（北京）技术有限公司 | The prediction meanss and method of specific type personnel calculate equipment and storage medium |
| CN110020214B (en) * | 2019-04-08 | 2021-05-18 | 北京航空航天大学 | Knowledge-fused social network streaming event detection system |
| CN110020214A (en) * | 2019-04-08 | 2019-07-16 | 北京航空航天大学 | A kind of social networks streaming events detection system merging knowledge |
| CN110083766B (en) * | 2019-04-26 | 2021-05-14 | 北京邮电大学 | Query recommendation method and device based on meta-path guiding embedding |
| CN110083766A (en) * | 2019-04-26 | 2019-08-02 | 北京邮电大学 | A kind of inquiry recommended method and device based on the insertion of first Route guiding |
| CN110119478A (en) * | 2019-05-15 | 2019-08-13 | 南京大学 | A kind of item recommendation method based on similarity of a variety of user feedback datas of combination |
| CN110334134B (en) * | 2019-07-17 | 2022-05-13 | 中国人民解放军国防科技大学 | Heterogeneous information network capability node importance degree evaluation method based on meta-path |
| CN110334134A (en) * | 2019-07-17 | 2019-10-15 | 中国人民解放军国防科技大学 | Heterogeneous information network capability node importance degree evaluation method based on meta-path |
| CN110659382B (en) * | 2019-09-04 | 2021-10-15 | 杭州电子科技大学 | Mixed music recommendation method based on heterogeneous information network representation learning technology |
| CN110659382A (en) * | 2019-09-04 | 2020-01-07 | 杭州电子科技大学 | Mixed music recommendation method based on heterogeneous information network representation learning technology |
| CN110750598A (en) * | 2019-09-18 | 2020-02-04 | 精锐视觉智能科技（深圳）有限公司 | Method and device for predicting article label, terminal equipment and storage medium |
| CN110750598B (en) * | 2019-09-18 | 2022-07-01 | 精锐视觉智能科技（深圳）有限公司 | Method and device for predicting article label, terminal equipment and storage medium |
| CN111222053A (en) * | 2019-11-27 | 2020-06-02 | 腾讯音乐娱乐科技（深圳）有限公司 | Object recommendation method and device and related equipment |
| CN110929046B (en) * | 2019-12-10 | 2022-09-30 | 华中师范大学 | Knowledge entity recommendation method and system based on heterogeneous network embedding |
| CN110929046A (en) * | 2019-12-10 | 2020-03-27 | 华中师范大学 | Knowledge entity recommendation method and system based on heterogeneous network embedding |
| CN111462897B (en) * | 2020-04-01 | 2021-05-11 | 山东大学 | Patient similarity analysis method and system based on improved heterogeneous information network |
| CN111462897A (en) * | 2020-04-01 | 2020-07-28 | 山东大学 | Patient similarity analysis method and system based on improved heterogeneous information network |
| WO2021208583A1 (en) * | 2020-04-16 | 2021-10-21 | 北京三快在线科技有限公司 | Recommendation information generation method and apparatus, electronic device and readable storage medium |
| CN111522889A (en) * | 2020-04-24 | 2020-08-11 | 腾讯科技（深圳）有限公司 | User interest tag expansion method and device, electronic equipment and storage medium |
| CN111538906B (en) * | 2020-05-29 | 2023-06-20 | 支付宝(杭州)信息技术有限公司 | Information pushing method and device based on privacy protection |
| CN111538906A (en) * | 2020-05-29 | 2020-08-14 | 支付宝(杭州)信息技术有限公司 | Information pushing method and device based on privacy protection |
| CN112016701B (en) * | 2020-09-09 | 2023-09-15 | 四川大学 | Abnormal change detection method and system integrating time sequence and attribute behaviors |
| CN112016701A (en) * | 2020-09-09 | 2020-12-01 | 四川大学 | Abnormal change detection method and system integrating time sequence and attribute behaviors |
| CN112221159B (en) * | 2020-10-21 | 2022-04-08 | 腾讯科技（深圳）有限公司 | Virtual item recommendation method and device and computer readable storage medium |
| CN112221159A (en) * | 2020-10-21 | 2021-01-15 | 腾讯科技（深圳）有限公司 | Virtual item recommendation method and device and computer readable storage medium |
| CN112541806A (en) * | 2020-12-18 | 2021-03-23 | 北京航天云路有限公司 | Recommendation method and device based on heterogeneous information network |
| CN113239271B (en) * | 2021-05-11 | 2022-03-15 | 四川大学 | Recommendation method based on interest drift |
| CN113239271A (en) * | 2021-05-11 | 2021-08-10 | 四川大学 | Recommendation method based on interest drift |
| CN113961693A (en) * | 2021-09-15 | 2022-01-21 | 北京三快在线科技有限公司 | Search result recommendation method and device, electronic device and readable storage medium |
| CN113935804A (en) * | 2021-10-15 | 2022-01-14 | 燕山大学 | Semantic recommendation method based on reinforcement learning and weighted element path |
| CN113935804B (en) * | 2021-10-15 | 2024-05-14 | 燕山大学 | Semantic recommendation method based on reinforcement learning and weighting element path |
| CN114637888B (en) * | 2022-05-18 | 2022-08-02 | 深圳市华曦达科技股份有限公司 | Video pushing method and device |
| CN114637888A (en) * | 2022-05-18 | 2022-06-17 | 深圳市华曦达科技股份有限公司 | Video pushing method and device |
| CN116089730A (en) * | 2023-04-07 | 2023-05-09 | 深圳市活力天汇科技股份有限公司 | Travel route recommending method |
| CN116089730B (en) * | 2023-04-07 | 2023-06-30 | 深圳市活力天汇科技股份有限公司 | Travel route recommending method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103955535A (en) | Individualized recommending method and system based on element path | |
| Shokeen et al. | A study on features of social recommender systems | |
| Anandhan et al. | Social media recommender systems: review and open research issues | |
| Tang et al. | Social recommendation: a review | |
| TWI636416B (en) | Method and system for multi-phase ranking for content personalization | |
| Huang et al. | Utilizing user tag-based interests in recommender systems for social resource sharing websites | |
| Panda et al. | Approaches and algorithms to mitigate cold start problems in recommender systems: a systematic literature review | |
| Shang et al. | A randomwalk based model incorporating social information for recommendations | |
| Kuanr et al. | Recent challenges in recommender systems: a survey | |
| Li et al. | Social recommendation based on trust and influence in SNS environments | |
| Wu et al. | Friend recommendation by user similarity graph based on interest in social tagging systems | |
| Revathy et al. | Cold start problem in social recommender systems: State-of-the-art review | |
| Vahedian et al. | Weighted random walk sampling for multi-relational recommendation | |
| Mao et al. | A fuzzy content matching-based e-commerce recommendation approach | |
| Zhao et al. | Combining meta-graph and attention for recommendation over heterogenous information network | |
| Fakhfakh et al. | Personalizing information retrieval: A new model for user preferences elicitation | |
| Zeng et al. | Context-aware social media recommendation based on potential group | |
| Pongnumkul et al. | Random walk-based recommendation with restart using social information and bayesian transition matrices | |
| Lu et al. | Profile Inference from Heterogeneous Data: Fundamentals and New Trends | |
| Ben Kharrat et al. | Empirical study of social collaborative filtering algorithm | |
| Movahedian et al. | A tag-based recommender system using rule-based collaborative profile enrichment | |
| Burke et al. | Recommendation in the social web | |
| Mustapha et al. | User recommendation algorithm in social tagging system based on hybrid user trust | |
| Zhang et al. | An efficient personalized video recommendation algorithm based on mixed mode | |
| Elahi | Cold Start Solutions For Recommendation Systems |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20140730 |