CN114443958A - Recommendation method, recommendation system and recommendation system training method - Google Patents
Recommendation method, recommendation system and recommendation system training method Download PDFInfo
- Publication number
- CN114443958A CN114443958A CN202210093523.XA CN202210093523A CN114443958A CN 114443958 A CN114443958 A CN 114443958A CN 202210093523 A CN202210093523 A CN 202210093523A CN 114443958 A CN114443958 A CN 114443958A
- Authority
- CN
- China
- Prior art keywords
- matrix
- network
- entity
- recommendation
- knowledge graph
- 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
Images
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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/3331—Query processing
- G06F16/334—Query execution
- G06F16/3344—Query execution using natural language analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/35—Clustering; Classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/36—Creation of semantic tools, e.g. ontology or thesauri
- G06F16/367—Ontology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/205—Parsing
- G06F40/211—Syntactic parsing, e.g. based on context-free grammar [CFG] or unification grammars
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/279—Recognition of textual entities
- G06F40/289—Phrasal analysis, e.g. finite state techniques or chunking
- G06F40/295—Named entity recognition
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/06—Buying, selling or leasing transactions
- G06Q30/0601—Electronic shopping [e-shopping]
- G06Q30/0631—Item recommendations
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/048—Activation functions
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Databases & Information Systems (AREA)
- General Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Computational Linguistics (AREA)
- Business, Economics & Management (AREA)
- Artificial Intelligence (AREA)
- General Health & Medical Sciences (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Health & Medical Sciences (AREA)
- Accounting & Taxation (AREA)
- Finance (AREA)
- Development Economics (AREA)
- Economics (AREA)
- Marketing (AREA)
- Strategic Management (AREA)
- General Business, Economics & Management (AREA)
- Life Sciences & Earth Sciences (AREA)
- Animal Behavior & Ethology (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a recommendation method, a recommendation system and a recommendation system training method, wherein the recommendation method comprises the following steps: s1, inputting the input information into the knowledge graph network to generate knowledge graph vectors; s2, inputting the knowledge map vector to a recommendation network to obtain a knowledge map user matrix and a commodity matrix; and S3, returning a recommendation result according to the knowledge graph user matrix and the commodity matrix. According to the method and the device, the knowledge graph representation model of the attention mechanism is constructed, so that the semantic relevance of the extracted features is further enhanced, information can be deeply mined, and the obtained recommendation result is richer and more accurate.
Description
Technical Field
The invention relates to the technical field of information retrieval, in particular to a recommendation method, a recommendation system and a recommendation system training method.
Background
The user browses, collects, evaluates or inputs keywords and the like on the network platform to generate various data, and based on the behaviors of the user and the generated information, after the information or the behaviors of the user are subjected to relevant analysis, the characteristics, the interests or the hobbies of the user can be identified, and then the relevant information, products or services and the like can be recommended to the user by combining with a specific purpose.
Therefore, various recommendation technologies are developed, and currently, it is common to recommend items to a user by using a collaborative filtering recommendation algorithm, where collaborative filtering refers to determining a plurality of items through a collaborative filtering model, and giving a prediction score to the plurality of items to form a similarity matrix, and recommending the products with higher prediction scores to the user.
There is another technique of recommending users by clicking an estimated binary algorithm.
However, the above schemes cannot effectively mine rich user images and product images, and the correlation attributes between the user images and the product images are difficult to mine deeply, so that the recommendation effect is not accurate enough.
Disclosure of Invention
The invention provides a recommendation method for solving the technical problems in the prior art, which comprises the following steps:
s1, inputting the input information into the knowledge graph network to generate knowledge graph vectors;
s2, inputting the knowledge map vector to a recommendation network to obtain a knowledge map user matrix and a commodity matrix;
and S3, returning a recommendation result according to the knowledge graph user matrix and the commodity matrix.
Further, step S2 specifically includes:
s21, inputting the knowledge map vector into the neural network to obtain attention expression;
s22, processing the head entity, the relation and the tail entity to obtain self-attention expression;
s23, aggregating the head entity and the self-attention expression to obtain aggregated information;
and S24, splicing the user and the commodity with the aggregation information to obtain a knowledge graph user matrix and a commodity matrix.
Further, S21, inputting the knowledge-graph vector into the neural network, and obtaining the attention expression comprises:
converting the entity of the commodity knowledge graph into a degree matrix through a graph neural network, and performing normalization processing;
converting the relationship of the commodity knowledge graph into an adjacency matrix;
the degree matrix obtained by the normalization process is multiplied by the adjacency matrix to obtain the attention expression.
Further, S22, the processing the head entity, the relationship and the tail entity, and obtaining the self-attention expression includes:
inputting the head entity, the tail entity and the relationship into a first linear network;
and adding the head entity and the relation passing through the first linear network, accessing the first activation function, and multiplying the head entity and the relation passing through the first linear network by the tail entity passing through the first linear network to obtain the self-attention expression.
Further, S23, aggregating the head entity and the self-attention expression to obtain the aggregation information includes:
adding the head entity and the self-attention expression, and accessing a second linear network and a second activation function to obtain first information;
multiplying the head reality by the self-attention expression, and accessing a third linear network and a third activation function to obtain second information;
and adding the first information and the second information to obtain the aggregation information.
The invention also provides a recommendation system, comprising a knowledge graph network and a recommendation network, wherein:
the knowledge graph network is used for inputting input information into the knowledge graph network to generate knowledge graph vectors; the recommendation network is used for inputting the knowledge map vector into the recommendation network to obtain a knowledge map user matrix and a commodity matrix; and returns the recommendation result.
The invention also provides a training method of the recommendation system, which comprises the following steps:
training a knowledge graph network;
and recommending a training step of the network.
Further, the training step of the knowledge-graph network comprises the following steps:
establishing a positive sample and a negative sample of the knowledge graph;
randomly initializing a head entity, a relation and a tail entity into 64-dimensional vectors;
representing the tail entity feature as the sum of the head entity feature and the relationship feature;
respectively multiplying the positive samples and the negative samples of the head entity, the relation entity and the tail entity by a same escape matrix to obtain a head entity matrix, a relation matrix, a tail entity positive sample matrix and a tail entity negative sample matrix;
calculating the matrix, and calculating a loss function by using the calculation result, wherein the formula is as follows:
x=(head_pos_e+relation_e-tail_pos_e)^2-(head_pos_e+relation_e-tail_neg_e)^2
setting a loss function, wherein the formula of the loss function is as follows:
L(x)=log(1+exp(x))
regularize the loss function using L2 regularization;
and training the knowledge graph network according to the loss value.
Further, the training step of the recommendation network comprises:
the user id and the commodity id are respectively represented by 64-dimensional vectors initialized by normal distribution;
constructing a commodity positive sample and a commodity negative sample;
recommending network extraction characteristics to obtain a user matrix and a commodity matrix;
and setting a loss function, and training the recommended network according to the loss value.
The present invention also provides an electronic device comprising: the recommendation system comprises a processor, a storage medium and a bus, wherein the storage medium stores machine-readable instructions executable by the processor, when the electronic device runs, the processor is communicated with the storage medium through the bus, and the processor executes the machine-readable instructions to execute the recommendation method.
In practical applications, the modules in the method and system disclosed by the invention can be deployed on one target server, or each module can be deployed on different target servers independently, and particularly, the modules can be deployed on cluster target servers according to needs in order to provide stronger computing processing capacity.
Therefore, the technical effect obtained by the technical scheme adopted by the invention is as follows:
according to the method and the device, the knowledge graph representation model of the attention mechanism is constructed, so that the semantic relevance of the extracted features is further enhanced, information can be deeply mined, and the obtained recommendation result is richer and more accurate.
In order that the invention may be more clearly and completely understood, specific embodiments thereof are described in detail below with reference to the accompanying drawings.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings used in the description of the embodiments will be briefly introduced below. It is obvious that the drawings in the following description are only some embodiments of the application, and that for a person skilled in the art, other drawings can be derived from them without inventive effort.
Fig. 1 is a schematic flow chart of a recommendation method according to an embodiment of the present application.
Fig. 2 is a schematic diagram of neural network training according to an embodiment of the present application.
Detailed Description
In the traditional two-classification method of collaborative filtering or user click estimation, under the E-commerce recommendation scene, rich user portraits and commodity portraits between users and commodities and associated attributes between the users and the commodities cannot be well mined to make recommendations.
Referring to fig. 1, the present application provides a recommendation method for addressing the defects of the prior art, which includes the steps of:
s1, inputting the input information into the knowledge graph network to generate knowledge graph vectors;
s2, inputting the knowledge map vector to a recommendation network to obtain a knowledge map user matrix and a commodity matrix;
and S3, returning a recommendation result according to the knowledge graph user matrix and the commodity matrix.
The commodity knowledge graph information is added, the representation capability of the enhanced knowledge graph is built, and the accuracy is improved.
The technical solution of the present application will be further described below with reference to various preferred embodiments.
S1, inputting the input information into the knowledge-graph network to generate knowledge-graph vectors:
in this embodiment, the input information may be a keyword or may refer to a sentence, and in this embodiment, the input information is a sentence for example, and by inputting a sentence, the knowledge graph network structure may perform information extraction to obtain information such as a head entity, a tail entity, and a relationship, so as to obtain a knowledge graph vector.
S2, inputting the knowledge map vector to a recommendation network to obtain a knowledge map user matrix and a commodity matrix:
referring to fig. 2, the recommendation network of the present application performs feature extraction on a knowledge graph vector to obtain a knowledge graph user matrix and a commodity matrix, and as a preferred embodiment, the specific method includes steps S21-S24:
and S21, inputting the knowledge map vector into the neural network to obtain the attention expression.
In this embodiment, as a preferred implementation, the neural network is a graph neural network, and the process thereof is as follows:
through the neural network, the entity of the commodity knowledge graph can be converted into a degree matrix, the relation is converted into an adjacent network, the degree matrix is subjected to normalization processing, and then the normalized degree matrix is multiplied by the adjacent network to obtain the attention expression.
S22, processing the head entity, the relation and the tail entity to obtain self-attention expression;
in this step, the head entity, the tail entity and the relationship are input into the first linear network, the head entity and the relationship passing through the first linear network are added, and then the first activation function is accessed.
In this embodiment, the activation function used is a tanh activation function, and as a modified implementation, a ReLu activation function, a sigmoid activation function, a softmax activation function, and the like may be used, and although different activation functions are used, the effects thereof are different, based on the purpose of the present application, accessing the activation function after the first linear network, and combining with other technical units to form a component of the technical solution of the present application do not depart from the essence of the present application.
After the tanh activation function is passed, the self-attention representation is obtained by multiplying the tail entity passing through the first linear network.
S23, aggregating the head entity and the self-attention expression to obtain aggregated information;
a preferred embodiment of the information aggregation step comprises the steps of:
adding the head entity and the self-attention expression, and accessing a second linear network and a second activation function to obtain first information;
multiplying the head reality by the self-attention expression, and accessing a third linear network and a third activation function to obtain second information;
and finally, adding the first information and the second information to obtain the aggregation information.
In this embodiment, both the second activation function and the third activation function select ReLU.
And S24, splicing the user and the commodity with the aggregation information to obtain a knowledge graph user matrix and a commodity matrix.
The knowledge graph user matrix and the commodity matrix are obtained, the recommendation result can be returned according to the knowledge graph user matrix and the commodity matrix, the specific process is that the user prefers scores to commodities, the user expression matrix and the commodity expression matrix are multiplied to obtain scores of the commodities by the user, and the first few commodities with high scores are selected to be returned to the user.
Based on the above embodiment, the present application further discloses a recommendation system, including a knowledge graph network and a recommendation network, wherein:
the knowledge graph network is used for inputting input information into the knowledge graph network to generate knowledge graph vectors; the recommendation network is used for inputting the knowledge map vector into the recommendation network to obtain a knowledge map user matrix and a commodity matrix; and returns the recommendation result.
Based on the recommendation method and the recommendation system of the above embodiment, the application further discloses a training method of the recommendation system, which includes a training step of the knowledge graph network and a training step of the recommendation network, that is, the training method trains the knowledge graph network and the recommendation network respectively, and as a variant implementation, the training method may also train the knowledge graph network first based on a test sample, then is used for testing the recommendation network, trains the knowledge graph network according to feedback of the recommendation network, and trains the recommendation network again by using the knowledge graph network after training the knowledge graph network.
As a preferred embodiment, the training step of the application for training the knowledge-graph independently comprises the following steps:
establishing a knowledge graph positive sample and a knowledge graph negative sample:
a knowledge graph positive sample, real head and tail entities and corresponding relations; such as: nova mobile phone (head) -brand (relation) -Hua is (tail)
Knowledge graph negative samples: the tail entity is randomly replaced with other entities. Such as: nova mobile phone (head) -brand (relation) -samsung (tail).
Then, randomly initializing the head entity, the relation and the tail entity into 64-dimensional vectors;
and representing the tail entity feature as the sum of the head entity feature and the relation feature, namely, the head entity feature + the relation feature is in a tail entity feature mode.
And respectively multiplying the positive samples and the negative samples of the head entity, the relation and the tail entity by a same escape matrix to obtain a head entity matrix head _ pos _ e, a relation matrix relation _ e, a tail entity positive sample matrix tail _ pos _ e and a tail entity negative sample matrix tail _ neg _ e.
And then, calculating the four matrixes, wherein the calculation formula is as follows:
x=(head_pos_e+relation_e-tail_pos_e)^2-(head_pos_e+relation_e-tail_neg_e)^2
after the calculation is completed, the value x is used to calculate a loss function, which has the following formula:
L(x)=log(1+exp(x))
and regularizing the loss function by using an L2 regularization, wherein the L2 regularization (comprising parameters such as head entities, relations, tail entities and the like) prevents overfitting of the model.
And finally, updating the knowledge graph spectrum vector through loss back propagation, thereby realizing the training of the knowledge graph network.
As a preferred embodiment, the training step of the recommendation network includes:
the user id and the commodity id are respectively represented by 64-dimensional vectors initialized by normal distribution and respectively represented by uid _ e and eid _ e.
Constructing a commodity positive sample and a commodity negative sample: a quality sample: the goods purchased by the user; commercial negative examples: the commodities purchased by the user are randomly replaced by other commodities which are not purchased;
recommending network extraction features to obtain a user matrix and a commodity matrix, and extracting the features, wherein the method mainly comprises the following steps:
information dissemination: converting commodity knowledge graph entities and relations into degree matrixes and adjacency matrixes by using a graph neural network; multiplying the normalized degree matrix and the adjacent matrix to obtain attention expression; (S21)
From the attention: adding the head entity and the relation of the first linear network, activating tanh, and multiplying the tail entity of the first linear network by the head entity to obtain self-attention expression;
the information aggregation comprises the following steps:
adding the head entity to the self-attention representation, via a second linear network and a ReLU activation function; multiplying the first information by the head entity to obtain the characteristic of the first information;
multiplying the head entity with the self-attention representation, via the third linear layer and the ReLU, results in a characterization of the second information.
And finally, adding the first information and the second information to obtain aggregation information, and finishing aggregation.
And finally, splicing the uid _ e and eid _ e with the knowledge map vector to obtain a knowledge map enhanced user matrix and commodity matrices uid _ ee and eid _ ee.
Setting a loss function, and training a recommended network according to the loss value:
loss of user preference score for goods: the user subtracts the user's negative sample score from the positive sample score and makes the positive sample score greater than the negative sample score, i.e., the BPR loss.
The embodiment of the present application further provides a computer-readable storage medium, where instructions or a program are stored in the storage medium, and the instructions or the program are loaded by a processor and execute any of the recommendation methods described above.
An embodiment of the present application further provides an electronic device, including: the recommendation system comprises a processor, a storage medium and a bus, wherein the storage medium stores machine-readable instructions executable by the processor, when an electronic device runs, the processor is communicated with the storage medium through the bus, and the processor executes the machine-readable instructions to execute the recommendation method according to any one of the above.
It should be noted that, all or part of the steps in the methods of the above embodiments may be implemented by hardware related to instructions of a computer program, which may be stored in a computer-readable storage medium, which may include, but is not limited to: read Only Memory (ROM), Random Access Memory (RAM), magnetic or optical disks, and the like.
The previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the invention. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims (10)
1. A recommendation method, comprising the steps of:
s1, inputting the input information into the knowledge graph network to generate knowledge graph vectors;
s2, inputting the knowledge map vector to a recommendation network to obtain a knowledge map user matrix and a commodity matrix;
and S3, returning a recommendation result according to the knowledge graph user matrix and the commodity matrix.
2. The recommendation method according to claim 1, wherein the step S2 specifically comprises:
s21, inputting the knowledge map vector into the neural network to obtain attention expression;
s22, processing the head entity, the relation and the tail entity to obtain self-attention expression;
s23, aggregating the head entity and the self-attention expression to obtain aggregated information;
and S24, splicing the user and the commodity with the aggregation information to obtain a knowledge graph user matrix and a commodity matrix.
3. The recommendation method as set forth in claim 2, wherein the entering of the knowledge-graph vector into the neural network, obtaining the attention expression comprises, at S21:
converting the entity of the commodity knowledge graph into a degree matrix through a graph neural network, and performing normalization processing;
converting the relationship of the commodity knowledge graph into an adjacency matrix;
the degree matrix obtained by the normalization process is multiplied by the adjacency matrix to obtain the attention expression.
4. The recommendation method of claim 2, wherein the processing the head entity, the relationship, and the tail entity to obtain the self-attention representation comprises S22:
inputting the head entity, the tail entity and the relationship into a first linear network;
and adding the head entity and the relation passing through the first linear network, accessing the first activation function, and multiplying the head entity and the relation passing through the first linear network by the tail entity passing through the first linear network to obtain the self-attention expression.
5. The recommendation method according to claim 2, wherein the step S23 of aggregating the head entity and the self-attention expression to obtain the aggregated information comprises:
adding the head entity and the self-attention expression, and accessing a second linear network and a second activation function to obtain first information;
multiplying the head reality by the self-attention expression, and accessing a third linear network and a third activation function to obtain second information;
and adding the first information and the second information to obtain the aggregation information.
6. A recommendation system is characterized by comprising a knowledge graph network and a recommendation network, wherein:
the knowledge graph network is used for inputting input information into the knowledge graph network to generate knowledge graph vectors; the recommendation network is used for inputting the knowledge map vector into the recommendation network to obtain a knowledge map user matrix and a commodity matrix; and returns the recommendation result.
7. A training method of a recommendation system is characterized by comprising the following steps:
training a knowledge graph network;
and recommending a training step of the network.
8. The training method of claim 7, wherein the training of the knowledge-graph network comprises:
establishing a positive sample and a negative sample of the knowledge graph;
randomly initializing a head entity, a relation and a tail entity into 64-dimensional vectors;
representing the tail entity feature as the sum of the head entity feature and the relationship feature;
respectively multiplying the positive samples and the negative samples of the head entity, the relation entity and the tail entity by a same escape matrix to obtain a head entity matrix, a relation matrix, a tail entity positive sample matrix and a tail entity negative sample matrix;
calculating the matrix, and calculating a loss function by using the calculation result, wherein the formula is as follows:
x=(head_pos_e+relation_e-tail_pos_e)^2-(head_pos_e+relation_e-tail_neg_e)^2
wherein: a head entity matrix head _ pos _ e, a relation matrix relation _ e, a tail entity positive sample matrix tail _ pos _ e, and a tail entity negative sample matrix tail _ neg _ e;
setting a loss function, wherein the formula of the loss function is as follows:
L(x)=log(1+exp(x))
regularize the loss function using L2 regularization;
and training the knowledge graph network according to the loss value.
9. The training method of claim 7, wherein the training of the recommendation network comprises:
the user id and the commodity id are respectively represented by 64-dimensional vectors initialized by normal distribution;
constructing a commodity positive sample and a commodity negative sample;
recommending network extraction characteristics to obtain a user matrix and a commodity matrix;
and setting a loss function, and training the recommended network according to the loss value.
10. An electronic device, comprising: a processor, a storage medium and a bus, the storage medium storing machine-readable instructions executable by the processor, the processor and the storage medium communicating via the bus when the electronic device is running, the processor executing the machine-readable instructions to perform the recommendation method of claims 1-5.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210093523.XA CN114443958A (en) | 2022-01-26 | 2022-01-26 | Recommendation method, recommendation system and recommendation system training method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210093523.XA CN114443958A (en) | 2022-01-26 | 2022-01-26 | Recommendation method, recommendation system and recommendation system training method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN114443958A true CN114443958A (en) | 2022-05-06 |
Family
ID=81369468
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202210093523.XA Pending CN114443958A (en) | 2022-01-26 | 2022-01-26 | Recommendation method, recommendation system and recommendation system training method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN114443958A (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115033718A (en) * | 2022-08-15 | 2022-09-09 | 浙江大学 | Service application deployment method, device and equipment |
CN115497465A (en) * | 2022-09-06 | 2022-12-20 | 平安银行股份有限公司 | Voice interaction method and device, electronic equipment and storage medium |
CN116308683A (en) * | 2023-05-17 | 2023-06-23 | 武汉纺织大学 | Knowledge-graph-based clothing brand positioning recommendation method, equipment and storage medium |
-
2022
- 2022-01-26 CN CN202210093523.XA patent/CN114443958A/en active Pending
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115033718A (en) * | 2022-08-15 | 2022-09-09 | 浙江大学 | Service application deployment method, device and equipment |
CN115033718B (en) * | 2022-08-15 | 2022-10-25 | 浙江大学 | Service application deployment method, device and equipment |
CN115497465A (en) * | 2022-09-06 | 2022-12-20 | 平安银行股份有限公司 | Voice interaction method and device, electronic equipment and storage medium |
CN116308683A (en) * | 2023-05-17 | 2023-06-23 | 武汉纺织大学 | Knowledge-graph-based clothing brand positioning recommendation method, equipment and storage medium |
CN116308683B (en) * | 2023-05-17 | 2023-08-04 | 武汉纺织大学 | Knowledge-graph-based clothing brand positioning recommendation method, equipment and storage medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111460130B (en) | Information recommendation method, device, equipment and readable storage medium | |
CN111931062B (en) | Training method and related device of information recommendation model | |
CN110717098B (en) | Meta-path-based context-aware user modeling method and sequence recommendation method | |
CN110503531B (en) | Dynamic social scene recommendation method based on time sequence perception | |
CN114443958A (en) | Recommendation method, recommendation system and recommendation system training method | |
CN111798273A (en) | Training method of purchase probability prediction model of product and purchase probability prediction method | |
EP4181026A1 (en) | Recommendation model training method and apparatus, recommendation method and apparatus, and computer-readable medium | |
CN110008397B (en) | Recommendation model training method and device | |
CN112380449B (en) | Information recommendation method, model training method and related device | |
CN111506820A (en) | Recommendation model, method, device, equipment and storage medium | |
CN114298851A (en) | Network user social behavior analysis method and device based on graph sign learning and storage medium | |
CN113656699B (en) | User feature vector determining method, related equipment and medium | |
CN113836390A (en) | Resource recommendation method and device, computer equipment and storage medium | |
Cao et al. | Implicit user relationships across sessions enhanced graph for session-based recommendation | |
CN116910357A (en) | Data processing method and related device | |
CN116541592A (en) | Vector generation method, information recommendation method, device, equipment and medium | |
CN114223012A (en) | Push object determination method and device, terminal equipment and storage medium | |
CN114155388B (en) | Image recognition method and device, computer equipment and storage medium | |
CN115344794A (en) | Scenic spot recommendation method based on knowledge map semantic embedding | |
Hartmann et al. | Studying the evolution of neural activation patterns during training of feed-forward relu networks | |
CN114417161A (en) | Virtual article time sequence recommendation method, device, medium and equipment based on special-purpose map | |
CN116764236A (en) | Game prop recommending method, game prop recommending device, computer equipment and storage medium | |
CN114529399A (en) | User data processing method, device, computer equipment and storage medium | |
CN112150206A (en) | Method and equipment for predicting user interested article | |
CN111860870A (en) | Training method, device, equipment and medium for interactive behavior determination model |
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 | ||
TA01 | Transfer of patent application right | ||
TA01 | Transfer of patent application right |
Effective date of registration: 20221219 Address after: Room 2703, No. 277, Xingang East Road, Haizhu District, Guangzhou, Guangdong 510220 Applicant after: Guangzhou Tianchen Health Technology Co.,Ltd. Address before: Building 10, No. 860, Xinyang Road, Lingang New District, China (Shanghai) pilot Free Trade Zone, Pudong New Area, Shanghai, 200120 Applicant before: Shanghai Yikangyuan Medical Health Technology Co.,Ltd. |