CN114780863A - Project recommendation method and device based on artificial intelligence, computer equipment and medium - Google Patents
Project recommendation method and device based on artificial intelligence, computer equipment and medium Download PDFInfo
- Publication number
- CN114780863A CN114780863A CN202210702599.8A CN202210702599A CN114780863A CN 114780863 A CN114780863 A CN 114780863A CN 202210702599 A CN202210702599 A CN 202210702599A CN 114780863 A CN114780863 A CN 114780863A
- Authority
- CN
- China
- Prior art keywords
- node
- nodes
- augmentation
- edge
- probability
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
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
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/901—Indexing; Data structures therefor; Storage structures
- G06F16/9024—Graphs; Linked lists
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- 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
- G06Q10/00—Administration; Management
- G06Q10/10—Office automation; Time management
- G06Q10/103—Workflow collaboration or project management
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- Databases & Information Systems (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Strategic Management (AREA)
- Human Resources & Organizations (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Entrepreneurship & Innovation (AREA)
- Software Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Economics (AREA)
- Marketing (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention relates to the technical field of artificial intelligence, in particular to an artificial intelligence-based project recommendation method, an artificial intelligence-based project recommendation device, computer equipment and a medium. The method includes the steps of constructing interactive graph data, calculating the augmentation probability of nodes according to each edge connected with the nodes in the interactive graph data, carrying out augmentation processing on the interactive graph data according to the augmentation probability of each node, forming a training data set by the obtained augmented interactive graph data, training a recommendation model to obtain a trained recommendation model, inputting the interactive graph data into the trained recommendation model to obtain a project recommendation result of each user, avoiding the situation that the interaction relation of hot users or hot projects generates large preference influence on the recommendation model to cause low generalization capability of the recommendation model and cannot provide accurate project recommendation for the users, and training the recommendation model by the augmented interactive graph data to fully learn the potential interaction characteristics of the users and the projects so as to improve the project recommendation accuracy.
Description
Technical Field
The invention relates to the technical field of artificial intelligence, in particular to a project recommendation method and device based on artificial intelligence, computer equipment and a medium.
Background
Currently, with the development of big data technology, the interaction relationship between the items and the users is stored in the form of interaction records, so that the potential preference of the users can be mined according to the interaction records of the items and the users to realize item recommendation, the item recommendation generally ranks the items to be recommended according to the similarity between each item to be recommended and the interacted item and the interaction heat of the items to be recommended, and the items to be recommended ranked at the front are determined as recommended items.
However, since the habits of users for selecting items are different, it is difficult to provide diversified selections for the users by recommending items according to the similarity degree, and recommending items according to the interactive heat of the items may cause that non-popular items are difficult to be recommended, so that the most suitable recommended items cannot be provided for the users, and the accuracy of item recommendation is low. Therefore, how to improve the precision of item recommendation becomes an urgent problem to be solved.
Disclosure of Invention
In view of this, embodiments of the present invention provide an artificial intelligence-based item recommendation method, an artificial intelligence-based item recommendation apparatus, a computer device, and a medium, so as to solve a problem of low accuracy of item recommendation.
In a first aspect, an embodiment of the present invention provides an artificial intelligence-based item recommendation method, where the item recommendation method includes:
constructing interactive graph data, wherein the interactive graph data comprises nodes and edges, the nodes comprise project nodes and user nodes, the node information of the project nodes comprises project information of a project, the node information of the user nodes comprises user information of a user, the edges are connected with the project nodes and the user nodes, and the edges are used for representing the interactive degree between the project nodes and the user nodes;
aiming at any node, calculating the augmentation probability of the node according to the edge connected with the node;
according to the augmentation probabilities of all the nodes, carrying out augmentation processing on the interactive graph data to obtain preset augmented interactive graph data;
constructing a training data set by using the preset augmented interactive map data, and updating parameters of a recommendation model by adopting a back propagation algorithm based on a comparison loss function until the comparison loss function is converged to obtain a trained recommendation model;
and inputting the interaction diagram data into the trained recommendation model to obtain a project recommendation result of each user in the interaction diagram data, wherein the project recommendation result is at least one item information in the interaction diagram data.
In one embodiment, the calculating, for any node, an augmented probability of the node according to an edge connected to the node includes:
counting the number of nodes except the nodes in the interactive graph data aiming at any node to obtain the number of the nodes;
and calculating the ratio of the number of edges connected with the nodes to the number of the nodes to obtain the augmentation probability of the nodes.
In an embodiment, after the building the interaction graph data, the method further includes:
aiming at any edge, determining the number of times of interaction between a user node connected with the edge and a project node connected with the edge as the weight corresponding to the edge;
updating the corresponding edges in the interactive map data according to the weight corresponding to each edge;
accordingly, calculating the augmented probability of the node according to the edges connected with the node comprises:
and calculating the augmentation probability of the node according to the weight corresponding to the edge connected with the node.
In one embodiment, the calculating the augmented probability of the node according to the weight corresponding to the edge connected to the node includes:
for any node, determining edges connected with the node as target edges, and calculating the sum of weights corresponding to the edges except the target edges in the interactive graph data to obtain an overall weight;
and calculating the ratio of the sum of the weights corresponding to the target edges to the overall weight to obtain the augmentation probability of the node.
In an embodiment, performing any one-time augmentation processing on the interaction graph data according to the augmentation probabilities of all the nodes includes:
aiming at any node, taking the augmentation probability of the node as the node rejection probability of the node;
sampling whether the node is discarded or not according to the node discarding probability of the node, and if the sampling result is discarded, discarding the node and an edge connected with the node from the interactive graph data;
and determining the discarded interactive map data as the interactive map data after the primary augmentation processing.
In an embodiment, performing any one-time augmentation processing on the interaction graph data according to the augmentation probabilities of all the nodes includes:
regarding any edge, taking the average value of the augmentation probabilities of the nodes connected with the edge as the edge rejection probability of the edge;
according to the edge rejection probability of the edge, sampling whether the edge is rejected, and if the sampling result is rejection, rejecting the edge from the interactive map data;
and determining the discarded interactive map data as the interactive map data after the augmentation processing.
In one embodiment, the recommendation model includes a first graph encoder and a second graph encoder;
the step of forming a training data set by using the preset augmented interaction map data, and updating the parameters of the recommendation model by using a back propagation algorithm based on a comparison loss function comprises the following steps:
combining any two augmented interactive graph data into a training sample to obtain M training samples, wherein M is an integer greater than zero;
inputting two pieces of augmentation data in any training sample into the first image encoder and the second image encoder respectively to obtain a first feature vector and a second feature vector;
and calculating the contrast loss function through the first feature vector and the second feature vector, and updating the parameters of the first image encoder and the second image encoder by adopting a back propagation algorithm based on the contrast loss function.
In a second aspect, an embodiment of the present invention provides an artificial intelligence-based item recommendation apparatus, where the item recommendation apparatus includes:
the graph data construction module is used for constructing interactive graph data, the interactive graph data comprises nodes and edges, the nodes comprise project nodes and user nodes, the node information of the project nodes comprises project information of a project, the node information of the user nodes comprises user information of a user, the edges are connected with the project nodes and the user nodes, and the edges are used for representing the interactive degree between the project nodes and the user nodes;
the probability calculation module is used for calculating the augmentation probability of any node according to the edge connected with the node;
the augmentation processing module is used for carrying out preset augmentation processing on the interactive map data according to the augmentation probabilities of all the nodes to obtain preset augmented interactive map data;
the model training module is used for forming a training data set by using the preset augmented interactive map data, updating parameters of a recommendation model by adopting a back propagation algorithm based on a comparison loss function until the comparison loss function is converged to obtain a trained recommendation model;
and the item recommendation module is used for inputting the interaction diagram data into the trained recommendation model to obtain an item recommendation result of each user in the interaction diagram data, wherein the item recommendation result is at least one item information in the interaction diagram data.
In one embodiment, the probability calculation module comprises:
the quantity counting unit is used for counting the number of the nodes except the nodes in the interactive graph data aiming at any node to obtain the number of the nodes;
and the probability calculation unit is used for calculating the ratio of the number of edges connected with the nodes to the number of the nodes to obtain the augmentation probability of the nodes.
In one embodiment, the item recommendation apparatus further comprises:
the weight determining module is used for determining the number of interaction times between the user node connected with the edge and the project node connected with the edge as the weight corresponding to the edge aiming at any edge;
the edge updating module is used for updating the corresponding edges in the interactive map data according to the weight corresponding to each edge;
accordingly, the probability calculation module comprises:
and the weighted probability calculating unit is used for calculating the augmentation probability of the node according to the weight corresponding to the edge connected with the node.
In one embodiment, the weighted probability calculating unit includes:
the overall weight calculation subunit is used for determining edges connected with the nodes as target edges aiming at any node and calculating the weight sum corresponding to the edges except the target edges in the interactive graph data to obtain the overall weight;
and the weighted probability determining subunit is used for calculating the ratio of the sum of the weights corresponding to the target edges to the overall weight to obtain the augmentation probability of the node.
In one embodiment, the augmentation processing module comprises:
a node probability determination unit configured to determine, as a node rejection probability of any node, an augmented probability of the node;
the node sampling unit is used for sampling whether the node is abandoned according to the node abandon probability of the node, and if the sampling result is abandoned, the node and an edge connected with the node are abandoned from the interactive graph data;
and determining the discarded interactive map data as the interactive map data after the primary augmentation processing.
In one embodiment, the augmentation processing module includes:
an edge probability determination unit configured to determine, for any edge, an edge rejection probability using an average value of the augmented probabilities of the nodes connected to the edge;
the edge sampling unit is used for sampling whether the edge is discarded or not according to the edge discarding probability of the edge, and if the sampling result is that the edge is discarded, the edge is discarded from the interactive map data;
and determining the discarded interactive map data as the interactive map data after the primary augmentation processing.
In one embodiment, the recommendation model includes a first graph encoder and a second graph encoder;
the model training module comprises:
the sample combination unit is used for combining any two augmented interactive graph data into one training sample to obtain M training samples, wherein M is an integer larger than zero;
the feature extraction unit is used for respectively inputting the two pieces of augmentation data in any training sample into the first image encoder and the second image encoder to obtain a first feature vector and a second feature vector;
and the parameter updating unit is used for calculating a contrast loss function through the first feature vector and the second feature vector, and updating the parameters of the first image encoder and the second image encoder by adopting a back propagation algorithm based on the contrast loss function.
In a third aspect, an embodiment of the present invention provides a computer device, where the computer device includes a processor, a memory, and a computer program stored in the memory and executable on the processor, and the processor implements the item recommendation method according to the first aspect when executing the computer program.
In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium, where a computer program is stored, and when the computer program is executed by a processor, the computer program implements the item recommendation method according to the first aspect.
Compared with the prior art, the embodiment of the invention has the following beneficial effects:
constructing interactive graph data, wherein the interactive graph data comprises nodes and edges, the nodes comprise project nodes and user nodes, the node information of the project nodes comprises project information of a project, the node information of the user nodes comprises user information of a user, the edges are connected with the project nodes and the user nodes and are used for representing the interactive degree between the project nodes and the user nodes, aiming at any node, the augmentation probability of the node is calculated according to the edges connected with the nodes, the interactive graph data is subjected to preset times of augmentation processing according to the augmentation probabilities of all the nodes to obtain preset augmented interactive graph data, the preset augmented interactive graph data forms a training data set, the parameters of a recommendation model are updated by adopting a back propagation algorithm according to a comparison loss function until the comparison loss function is converged to obtain the trained recommendation model, the method comprises the steps of inputting interactive map data into a trained recommendation model to obtain a project recommendation result of each user in the interactive map data, wherein the project recommendation result is at least one item information in the interactive map data, and the augmentation probability determined through the interactive relation of the nodes samples the augmentation processing modes of the nodes and the edges, so that the situation that the interactive relation of hot users or hot projects generates large preference influence on the recommendation model, the generalization capability of the recommendation model is low, accurate project recommendation cannot be provided for the users, the recommendation model is compared and learned through the augmentation data, the potential interactive characteristics of the users and the projects can be fully learned, and the project recommendation accuracy is improved.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings needed to be used in the embodiments or the prior art descriptions will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without inventive exercise.
FIG. 1 is a schematic diagram of an application environment of an artificial intelligence-based item recommendation method according to an embodiment of the present invention;
FIG. 2 is a flowchart illustrating an artificial intelligence-based item recommendation method according to an embodiment of the present invention;
FIG. 3 is a flowchart illustrating a second embodiment of an artificial intelligence-based item recommendation method according to the present invention;
fig. 4 is a schematic structural diagram of an artificial intelligence-based item recommendation apparatus according to a third embodiment of the present invention;
fig. 5 is a schematic structural diagram of a computer device according to a fourth embodiment of the present invention.
Detailed Description
In the following description, for purposes of explanation and not limitation, specific details are set forth, such as particular system structures, techniques, etc. in order to provide a thorough understanding of the embodiments of the invention. It will be apparent, however, to one skilled in the art that the present invention may be practiced in other embodiments that depart from these specific details. In other instances, detailed descriptions of well-known systems, devices, circuits, and methods are omitted so as not to obscure the description of the present invention with unnecessary detail.
It should be understood that the terms "comprises" and/or "comprising," when used in this specification and the appended claims, specify the presence of stated features, integers, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and/or groups thereof.
It should also be understood that the term "and/or" as used in this specification and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.
As used in this specification and the appended claims, the term "if" may be interpreted contextually as "when", "upon" or "in response to a determination" or "in response to a detection". Similarly, the phrase "if it is determined" or "if a [ described condition or event ] is detected" may be interpreted contextually to mean "upon determining" or "in response to determining" or "upon detecting [ described condition or event ]" or "in response to detecting [ described condition or event ]".
Furthermore, in the description of the present invention and the appended claims, the terms "first," "second," "third," and the like are used for distinguishing between descriptions and not necessarily for describing or implying relative importance.
Reference throughout this specification to "one embodiment" or "some embodiments," or the like, means that a particular feature, structure, or characteristic described in connection with the embodiment is included in one or more embodiments of the present invention. Thus, appearances of the phrases "in one embodiment," "in some embodiments," "in other embodiments," or the like, in various places throughout this specification are not necessarily all referring to the same embodiment, but rather mean "one or more but not all embodiments" unless specifically stated otherwise. The terms "comprising," "including," "having," and variations thereof mean "including, but not limited to," unless otherwise specifically stated.
The embodiment of the invention can acquire and process related data based on an artificial intelligence technology. Among them, Artificial Intelligence (AI) is a theory, method, technique and application system that simulates, extends and expands human Intelligence using a digital computer or a machine controlled by a digital computer, senses the environment, acquires knowledge and uses the knowledge to obtain the best result.
The artificial intelligence infrastructure generally includes technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technologies, operation/interaction systems, mechatronics, and the like. The artificial intelligence software technology mainly comprises a computer vision technology, a robot technology, a biological recognition technology, a voice processing technology, a natural language processing technology, machine learning/deep learning and the like.
It should be understood that, the sequence numbers of the steps in the following embodiments do not imply an execution sequence, and the execution sequence of each process should be determined by the function and the internal logic thereof, and should not limit the implementation process of the embodiments of the present invention in any way.
In order to explain the technical means of the present invention, the following description will be given by way of specific examples.
The artificial intelligence-based project recommendation method provided by the embodiment of the invention can be applied to an application environment shown in fig. 1, wherein a client communicates with a server. The client includes, but is not limited to, a palm top computer, a desktop computer, a notebook computer, an ultra-mobile personal computer (UMPC), a netbook, a cloud terminal device, a Personal Digital Assistant (PDA), and other computer devices. The server side can be implemented by an independent server or a server cluster formed by a plurality of servers.
Referring to fig. 2, which is a flowchart illustrating a project recommendation method based on artificial intelligence according to an embodiment of the present invention, the project recommendation method may be applied to the server in fig. 1, where after receiving a recommendation request sent by a client, the server obtains project information and user information provided by the corresponding client, and the recommendation request is sent by the client and is used for requesting the server to perform project recommendation. The server is deployed with a recommendation model for accurately recommending items for the user, and model parameters of the recommendation model are in a writable state, namely the model parameters are allowed to be updated in real time. As shown in fig. 2, the item recommendation method may include the steps of:
step S201, interactive graph data is constructed.
The interactive graph data comprises nodes and edges, the nodes comprise project nodes and user nodes, the node information of the project nodes comprises project information of a project, the node information of the user nodes comprises user information of a user, the project information can be information such as project identification and project content, and the user information can be user identification and user personal information.
In this embodiment, the interaction degree is set to be binary data according to whether there is interaction between the project node and the user node, where the binary data includes 0 and 1, and when there is interaction between the project node and the user node, the interaction degree is set to be 1, that is, there is a connected edge between the project node and the user node, and when there is no interaction between the project node and the user node, the interaction degree is set to be 0, that is, there is no connected edge between the project node and the user node.
Specifically, the interaction graph data is a bipartite graph, that is, the interaction graph data includes two node sets, namely, a project node set and a user node set, the project node set includes all project nodes, the user node set includes all user nodes, no edge exists in the same node set, and an edge exists between the two node sets.
In the step of constructing the interactive map data, the interactive relation between the user and the project is expressed in the form of the interactive map data, so that the potential characteristics of the interactive relation can be conveniently mined by a subsequent recommendation model, and the accuracy of subsequent project recommendation is improved.
Step S202, aiming at any node, calculating the augmentation probability of the node according to the edge connected with the node.
The edge may be an edge whose vertex is a node, and the augmentation probability may be a sampling probability of the augmentation operation in the augmentation process.
Specifically, the degree of hotness of the node is evaluated according to the attributes of the edges connected with the node, wherein the attributes of the edges can refer to the number of the edges, the weight of the edges and the like. According to the fact that the popular degrees of the nodes are different, the augmentation probabilities of the nodes are also different, the nodes with the higher popular degrees represent users who frequently interact with the projects or projects which are frequently interacted with the projects, and the popular users or popular nodes can have considerable influence on the overall preference of the recommendation model.
Optionally, for any node, calculating the augmented probability of the node according to the edge connected to the node includes:
counting the number of nodes except the nodes in the interactive graph data aiming at any node to obtain the number of the nodes;
and calculating the ratio of the number of edges connected with the nodes to the number of the nodes to obtain the augmentation probability of the nodes.
The number of nodes other than the nodes may refer to the total number of other nodes, the number of nodes is fixed because the structure of the interaction graph data is fixed after construction, and the number of edges connected to the nodes may refer to the number of edges with the nodes as vertices.
For example, if the number of all nodes is a, the number of nodes is a-1, in the same user graph data, the number of nodes corresponding to each node is a-1, and if the number of edges using the target node as a vertex is B, the augmentation probability of the target node is
。
In the embodiment, the augmentation probability is calculated by using the ratio of the number of edges connected with the nodes to the number of the nodes, so that the importance degree of the nodes can be effectively represented, and the deviation caused by hot users or hot projects can be avoided in subsequent augmentation processing.
Optionally, after the interactive graph data is constructed, the method further includes:
aiming at any edge, determining the number of interaction times between the user node connected with the edge and the project node connected with the edge as the weight corresponding to the edge;
updating the corresponding edges in the interactive graph data according to the weight corresponding to each edge;
accordingly, calculating the augmented probability of a node based on edges connected to the node comprises:
and calculating the augmentation probability of the node according to the weight corresponding to the edge connected with the node.
The edge weight is 1 by default, and in this embodiment, the edge weight is the number of interactions between the connected user node and the project node.
Specifically, for any edge, before the interactive graph data is updated, the edge weight is a default value of 1, and during updating, the number of times of interaction between the user node and the project node connected with the corresponding edge is assigned to the corresponding edge as the weight.
According to the method and the device, richer information is provided for the interaction relationship through the form of weight assignment, so that the extraction of subsequent potential features is facilitated, and the accuracy of project recommendation is improved.
Optionally, calculating the augmented probability of the node according to the weight corresponding to the edge connected to the node includes:
determining edges connected with the nodes as target edges aiming at any node, and calculating the sum of weights corresponding to the edges except the target edges in the interactive graph data to obtain the overall weight;
and calculating the ratio of the sum of the weights corresponding to the target edges to the whole weight to obtain the augmentation probability of the node.
The overall weight refers to a total weight corresponding to an edge that does not use the target node as a vertex.
For example, assuming that the weights of the target edges with the target nodes as the vertices are 1, 2, and 3, respectively, the sum of the weights of all edges in the interaction graph data is 100, the overall weight is 94, the sum of the weights of the target edges is 6, and the augmentation probability of the target nodes is
。
In the embodiment, when the interactive map data is a weighted map, the augmentation probability is calculated by using the ratio of the sum of the weights corresponding to the target edges to the overall weight, so that the importance degree of the node can be effectively represented, and the deviation caused by hot users or hot projects can be avoided in the subsequent augmentation processing.
The step of calculating the augmentation probability of the node according to the edge connected with the node aiming at any node is carried out, and higher augmentation probability is set for the hot user or the hot node, so that the hot node is easier to discard during augmentation processing, model parameter deviation caused by heat is avoided, and generalization capability of the recommended model is improved.
Step S203, according to the augmentation probabilities of all the nodes, preset augmentation processing is carried out on the interactive map data, and interactive map data after preset augmentation processing are obtained.
The preset value is an integer greater than zero, the augmentation processing may refer to processing such as node rejection, node change, edge rejection, edge change and the like on the interactive map data, and the augmented interactive map data may refer to map data obtained after the augmentation processing on the interactive map data.
Specifically, whether the corresponding nodes are subjected to augmentation processing or not is sampled according to the augmentation probability of each node, augmented interaction graph data and interaction graph data are different, differences also exist among the augmented interaction graph data due to sampling uncertainty, and the augmented interaction graph data are used for hiding or changing part of information in the interaction graph data, so that the recommendation model can learn enough features from less information.
Optionally, performing any amplification processing on the interaction graph data according to the amplification probabilities of all the nodes includes:
aiming at any node, taking the augmentation probability of the node as the node rejection probability of the node;
sampling whether the nodes are discarded or not according to the node discarding probability of the nodes, and if the sampling result is that the nodes are discarded, discarding the nodes and edges connected with the nodes from the interactive map data;
and determining the discarded interactive map data as the interactive map data after the primary augmentation processing.
The node discarding probability may refer to a probability that the target node is deleted in the interactive graph data in an augmentation processing process, and discarding may refer to deleting the node from the interactive graph data.
Sampling may refer to sampling events, and in this embodiment, the events include discarding the target node from the interaction graph data and retaining the node in the interaction graph data.
Specifically, the value range of the augmented probability is [0, 1], the value range of the node rejection probability is [0, 1], the value obtained by subtracting the node rejection probability from 1 is determined as the node retention probability, and the value range of the node retention probability is [0, 1 ].
The target node is discarded from the interactive graph data and the target node is retained in the interactive graph data for sampling in a probabilistic sampling mode, specifically, sampling can be performed by generating a random number within a range of [0, 1], a section in which the random number is included is judged, if the random number is included in the node discarding section, the sampling result is to discard the target node from the interactive graph data, and if the random number is included in the node retaining section, the sampling result is to retain the target node in the interactive graph data. For example, if the node rejection probability is 0.3 and the node retention probability is 0.7, the node rejection interval is determined to be [0, 0.3), the node retention interval is [0.3, 1], a number is randomly generated within the range of [0, 1], if the generated number is 0.2, 0.2 is included in the node rejection interval [0, 0.3), and the sampling result is that the target node is rejected from the interaction map data.
It should be noted that, under the condition of a large number of nodes, the node sampling probability will be low, and the requirement of the augmentation process may not be met, and the implementer may set the maximum rejection probability, in this embodiment, the maximum rejection probability is set to 0.3, and the ratio of the maximum rejection probability to the node rejection probability is used as the scaling quantity, and each node rejection probability is multiplied by the scaling quantity to update the corresponding node rejection probability.
In the embodiment, the augmentation processing result of each node is determined node by node in a sampling mode, rich interaction graph data after augmentation processing can be generated, the condition that a training data set is few is avoided, the training effect of the recommendation model is improved, and the recommendation model is beneficial to learning and extracting effective features in the training process.
Optionally, performing any amplification processing on the interaction graph data according to the amplification probabilities of all the nodes includes:
regarding any edge, taking the average value of the augmentation probabilities of the nodes connected with the edge as the edge rejection probability of the edge;
and sampling whether the edges are discarded or not according to the edge discarding probability of the edges, and if the sampling result is discarded, discarding the edges from the interactive map data to obtain the discarded interactive map data as the interactive map data after primary augmentation processing.
The edge discarding probability may refer to a probability that an edge is deleted in the interactive map data in an augmentation process, and the discarding may refer to deleting an edge from the interactive map data.
Sampling may refer to sampling events, and in this embodiment, the events include discarding edges from the interaction graph data and retaining edges in the interaction graph data.
Specifically, the value range of the augmented probability is [0, 1], the value range of the edge rejection probability is [0, 1], the value of the edge rejection probability subtracted by 1 is determined to be the edge retention probability, the value range of the edge retention probability is [0, 1], and the edge to be rejected from the interactive map data and the edge to be retained in the interactive map data are sampled in a probability sampling mode.
In the embodiment, the augmentation processing results of the corresponding edges are determined edge by edge in a sampling mode, rich interaction graph data after augmentation processing can be generated, the condition that a training data set is few is avoided, the effect of training the recommendation model is improved, and the recommendation model can learn to extract effective features in the training process.
According to the step of carrying out the preset amplification treatment on the interactive map data according to the amplification probability of all the nodes to obtain the interactive map data after the preset amplification treatment, the training data set is favorably expanded, overfitting during training of the recommendation model is avoided, meanwhile, the capability of the recommendation model for extracting effective features is improved, and therefore the accuracy of project recommendation is improved.
And S204, forming a training data set by using preset augmented interactive map data, and updating parameters of the recommendation model by using a back propagation algorithm based on the comparison loss function until the comparison loss function is converged to obtain the trained recommendation model.
The training data set comprises a plurality of training samples, the training samples can refer to input quantity of each training batch during model training, and the training samples are determined by interaction diagram data after augmentation processing.
The contrast loss function is:
wherein, L is a contrast loss function, K represents the kth training batch, K is the number of times of training total batches, y represents whether the samples are matched or not, when the y value is 1, the samples are matched, when the y value is 0, the samples are not matched, d represents the Euclidean distance between the samples, and margin is a set distance threshold.
The back propagation algorithm is based on a gradient descent method, and partial derivatives of the loss function to each model parameter are calculated layer by layer from the output layer to the input layer and serve as a basis for modifying the model parameters, wherein the gradient descent method can be a random gradient descent method, a batch gradient descent method and the like.
The convergence of the contrast loss function means that the gradient change of the contrast loss function approaches 0 along with the update of the model parameters.
In this embodiment, the recommendation model may be a graph and volume matrix completion network, which is used to complete the input graph data according to the information of the input graph data, and only an encoder pre-trained by the graph and volume matrix completion network is selected as the recommendation model, so as to ensure that the recommendation model has a capability of basically extracting features.
Specifically, in this embodiment, the augmented interaction graph data forms a training data set, a difference exists between the augmented interaction graph data, and the expected recommendation model learns different feature information according to the interaction graph data after different augmentation processes, so as to more effectively mine potential features of user-item interaction, and therefore, when sample labels y are all set to 0, the comparison loss function can be expressed as:
that is, if the euclidean distance between samples is smaller than the distance threshold margin, the contrast loss function is greater than 0, and if the euclidean distance between samples is greater than the distance threshold margin, the contrast loss function is equal to 0, so that the feature vectors corresponding to different samples are as far as possible.
In an embodiment, the recommendation model may also refer to a neural system filtering model, a linear residual map convolution collaborative filtering model, or the like.
And the step of constructing a training data set by using the preset augmented interactive map data, updating parameters of the recommendation model by adopting a back propagation algorithm based on the comparison loss function until the comparison loss function is converged to obtain the trained recommendation model, and extracting different feature information from the recommendation model by using the different augmented interactive map data, so that the recommendation model can more effectively explore potential features of interaction between users and projects, and the accuracy of the recommendation model is improved.
And S205, inputting the interactive map data into the trained recommendation model to obtain the item recommendation result of each user in the interactive map data.
And the item recommendation result is at least one item information in the interactive map data. And the decoder is used for reconstructing the characteristics into recommended items corresponding to each user in the interaction diagram data.
Specifically, the pre-training of the encoder adopts an effective feature vector as a pre-training sample, a reference recommended item corresponding to each user is used as a pre-training label, cross entropy loss is used as a pre-training loss function, the pre-training loss function is used as a basis, and a back propagation algorithm is adopted to update parameters of the decoder until the pre-training loss function converges, so that the trained decoder is obtained.
The effective feature vector may refer to a feature vector output by a trained recommendation model of the interactive map data, the reference recommendation item may refer to a recommendation item of a corresponding user marked by a human, and the cross entropy loss is:
wherein, l represents the cross entropy loss,
indicates the label corresponding to the ith sample,
indicates the probability of the label corresponding to the ith sample,
the output probability corresponding to the ith sample is shown, and I represents the total number of samples.
The step of inputting the interactive map data into the trained recommendation model to obtain the item recommendation result of each user in the interactive map data can output the corresponding recommended item for each user according to the potential interactive characteristics of the users and the items, and the accuracy of item recommendation is effectively improved.
According to the method, the augmentation probability determined by the interactive relation of the nodes is used for sampling the augmentation processing modes of the nodes and the edges, the phenomenon that the interactive relation of hot users or hot projects generates great preference influence on the recommendation model is avoided, the generalization capability of the recommendation model is low, accurate project recommendation cannot be provided for the users, the recommendation model is compared and learned through augmentation data, potential interactive characteristics of the users and the projects can be fully learned, and the project recommendation accuracy is improved.
Referring to fig. 3, a flowchart of an artificial intelligence based item recommendation method according to a second embodiment of the present invention is shown, in which a recommendation model includes a first graph encoder and a second graph encoder;
the training process of the recommendation model is as follows:
step S301, combining any two augmented interactive graph data into one training sample to obtain M training samples.
Step S302, inputting two pieces of augmentation data in any training sample into a first graph encoder and a second graph encoder respectively to obtain a first feature vector and a second feature vector.
Step S303, a contrast loss function is calculated through the first eigenvector and the second eigenvector, and the parameters of the first image encoder and the second image encoder are updated by using a back propagation algorithm based on the contrast loss function.
Where M is an integer greater than zero, and the value of M can be set by the implementer, in this embodiment, 1000.
And parameters of the first graph encoder and the second graph encoder are shared, and the first graph encoder and the second graph encoder adopt the graph convolution matrix completion network pre-trained encoder.
The first feature vector may refer to a feature vector corresponding to first augmented interactive map data in the training sample, and the second feature vector may refer to a feature vector corresponding to second augmented interactive map data in the training sample.
In particular, the maximum value of M is influenced by the above-mentioned preset value N, i.e.
The implementer should not exceed 0 when determining the value of M,
]the range of (1).
According to the embodiment, the recommendation model is trained in a comparison learning mode, so that the distance between the feature vectors corresponding to different input quantities is as far as possible, the recommendation model is guided to extract potential features, and the accuracy of item recommendation output by the trained recommendation model is improved.
Fig. 4 shows a block diagram of a third embodiment of an artificial intelligence-based item recommendation device applied to a server, where the server obtains item information and user information provided by a corresponding client after receiving a recommendation request sent by the client, and the recommendation request is sent by the client and used for requesting the server to recommend an item. The server is deployed with a recommendation model for accurately recommending items for the user, and model parameters of the recommendation model are in a writable state, namely the model parameters are allowed to be updated in real time. For convenience of explanation, only portions related to the embodiments of the present invention are shown.
Referring to fig. 4, the item recommendation apparatus includes:
the graph data building module 41 is configured to build interactive graph data, where the interactive graph data includes nodes and edges, and the nodes include project nodes and user nodes, where the node information of a project node includes project information of a project, the node information of a user node includes user information of a user, and the edge connects the project node and the user node and is used to represent an interaction degree between the project node and the user node;
a probability calculation module 42, configured to calculate, for any node, an augmented probability of the node according to an edge connected to the node;
an augmentation processing module 43, configured to perform preset times of augmentation processing on the interactive map data according to the augmentation probabilities of all nodes to obtain preset augmented interactive map data, where the preset is an integer greater than zero;
the model training module 44 is configured to form a training data set by using preset augmented interaction map data, and update parameters of the recommendation model by using a back propagation algorithm based on a comparison loss function until the comparison loss function converges to obtain a trained recommendation model;
and the item recommendation module 45 is configured to input the interactive map data into the trained recommendation model to obtain an item recommendation result of each user in the interactive map data, where the item recommendation result is at least one item information in the interactive map data.
Optionally, the probability calculating module 42 includes:
the quantity counting unit is used for counting the number of nodes except the nodes in the interactive graph data aiming at any node to obtain the number of the nodes;
and the probability calculation unit is used for calculating the ratio of the number of edges connected with the nodes to the number of the nodes to obtain the augmentation probability of the nodes.
Optionally, the item recommendation apparatus further includes:
the weight determining module is used for determining the number of interaction times between the user node connected with the edge and the project node connected with the edge as the weight corresponding to the edge aiming at any edge;
the edge updating module is used for updating the corresponding edges in the interactive map data according to the weight corresponding to each edge;
accordingly, the probability calculation module 42 includes:
and the weighted probability calculating unit is used for calculating the augmentation probability of the node according to the weight corresponding to the edge connected with the node.
Optionally, the weighted probability calculating unit includes:
the overall weight calculation subunit is used for determining edges connected with the nodes as target edges aiming at any node, and calculating the sum of weights corresponding to the edges except the target edges in the interactive graph data to obtain the overall weight;
and the weighted probability determining subunit is used for calculating the ratio of the sum of the weights corresponding to the target edges to the overall weight to obtain the augmented probability of the node.
Optionally, the amplification processing module 43 includes:
a node probability determination unit configured to use the augmentation probability of the node as a node rejection probability of the node for any node;
the node sampling unit is used for sampling whether the node is discarded or not according to the node discarding probability of the node, and if the sampling result is discarded, the node and the edge connected with the node are discarded from the interactive graph data;
and determining the discarded interactive map data as the interactive map data after the primary augmentation processing.
Optionally, the amplification processing module 43 includes:
an edge probability determination unit configured to use, as an edge discard probability of an edge, an augmented probability mean of nodes connected to the edge for any edge;
the edge sampling unit is used for sampling whether the edge is discarded or not according to the edge discarding probability of the edge, and if the sampling result is discarded, discarding the edge from the interactive map data;
and determining the discarded interactive map data as the interactive map data after the primary augmentation processing.
Optionally, the recommendation model comprises a first graph encoder and a second graph encoder;
the model training module 44 includes:
the sample combination unit is used for combining any two augmented interactive graph data into one training sample to obtain M training samples, wherein M is an integer larger than zero;
the feature extraction unit is used for respectively inputting the two pieces of augmentation data in any training sample into the first image encoder and the second image encoder to obtain a first feature vector and a second feature vector;
and the parameter updating unit is used for calculating a contrast loss function through the first feature vector and the second feature vector, and updating the parameters of the first image encoder and the second image encoder by adopting a back propagation algorithm based on the contrast loss function.
It should be noted that, because the above-mentioned information interaction, execution process and other contents between the modules, units and sub-units are based on the same concept as the method embodiment of the present invention, specific functions and technical effects thereof may be referred to specifically in the method embodiment section, and are not described herein again.
Fig. 5 is a schematic structural diagram of a computer device according to a fourth embodiment of the present invention. As shown in fig. 5, the computer apparatus of this embodiment includes: at least one processor (only one shown in fig. 5), a memory, and a computer program stored in the memory and executable on the at least one processor, the processor when executing the computer program implementing the steps in any of the various item recommendation method embodiments described above.
The computer device may include, but is not limited to, a processor, a memory. It will be appreciated by those skilled in the art that fig. 5 is merely an example of a computer device and is not intended to be limiting, and that a computer device may include more or fewer components than those shown, or some components may be combined, or different components may be included, such as a network interface, a display screen, and input devices, etc.
The Processor may be a CPU, or other general purpose Processor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA) or other Programmable logic device, discrete Gate or transistor logic, discrete hardware components, or the like. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.
The memory includes readable storage medium, internal memory, etc., where the internal memory may be a memory of the computer device, and the internal memory provides an environment for the operating system and the execution of computer-readable instructions in the readable storage medium. The readable storage medium may be a hard disk of the computer device, and in other embodiments may also be an external storage device of the computer device, such as a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) Card, a Flash memory Card (Flash Card), and the like, provided on the computer device. Further, the memory may also include both internal and external storage units of the computer device. The memory is used for storing an operating system, application programs, a BootLoader (BootLoader), data, and other programs, such as program codes of a computer program, and the like. The memory may also be used to temporarily store data that has been output or is to be output.
It will be apparent to those skilled in the art that, for convenience and brevity of description, only the above-mentioned division of the functional units and modules is illustrated, and in practical applications, the above-mentioned function distribution may be performed by different functional units and modules according to needs, that is, the internal structure of the apparatus is divided into different functional units or modules, so as to perform all or part of the functions described above. Each functional unit and module in the embodiments may be integrated in one processing unit, or each unit may exist alone physically, or two or more units are integrated in one unit, and the integrated unit may be implemented in a form of hardware, or in a form of software functional unit. In addition, specific names of the functional units and modules are only used for distinguishing one functional unit from another, and are not used for limiting the protection scope of the present invention. For the specific working processes of the units and modules in the above-mentioned apparatus, reference may be made to the corresponding processes in the foregoing method embodiments, which are not described herein again. The integrated unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, all or part of the flow of the method of the above embodiments may be implemented by a computer program, which may be stored in a computer readable storage medium and used by a processor to implement the steps of the above method embodiments. Wherein the computer program comprises computer program code, which may be in the form of source code, object code, an executable file or some intermediate form, etc. The computer readable medium may include at least: any entity or device capable of carrying computer program code, recording medium, computer Memory, Read-Only Memory (ROM), Random Access Memory (RAM), electrical carrier wave signals, telecommunications signals, and software distribution media. Such as a usb-disk, a removable hard disk, a magnetic or optical disk, etc. In some jurisdictions, computer-readable media may not be an electrical carrier signal or a telecommunications signal in accordance with legislative and proprietary practices.
The present invention may also be implemented by a computer program product, which when executed on a computer device, enables the computer device to implement all or part of the processes in the method according to the above embodiments.
In the above embodiments, the descriptions of the respective embodiments have respective emphasis, and reference may be made to the related descriptions of other embodiments for parts that are not described or illustrated in a certain embodiment.
Those of ordinary skill in the art will appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, or combinations of computer software and electronic hardware. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the technical solution. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present invention.
In the embodiments provided by the present invention, it should be understood that the disclosed apparatus/computer device and method may be implemented in other ways. For example, the above-described apparatus/computer device embodiments are merely illustrative, and for example, a module or a unit may be divided into only one logical function, and may be implemented in other ways, for example, a plurality of units or components may be combined or integrated into another system, or some features may be omitted, or not implemented. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, devices or units, and may be in an electrical, mechanical or other form.
The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one position, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.
The above examples are only intended to illustrate the technical solution of the present invention, and not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; such modifications and substitutions do not depart from the spirit and scope of the embodiments of the present invention, and they should be construed as being included therein.
Claims (10)
1. An artificial intelligence based item recommendation method, characterized in that the method comprises:
constructing interactive graph data, wherein the interactive graph data comprises nodes and edges, the nodes comprise project nodes and user nodes, the node information of the project nodes comprises project information of a project, the node information of the user nodes comprises user information of a user, the edges are connected with the project nodes and the user nodes, and the edges are used for representing the interactive degree between the project nodes and the user nodes;
aiming at any node, calculating the augmentation probability of the node according to the edge connected with the node;
performing preset-time augmentation processing on the interactive graph data according to the augmentation probabilities of all the nodes to obtain preset augmented interactive graph data;
constructing a training data set by using the preset augmented interactive map data, and updating parameters of a recommendation model by adopting a back propagation algorithm based on a comparison loss function until the comparison loss function is converged to obtain a trained recommendation model;
and inputting the interaction diagram data into the trained recommendation model to obtain a project recommendation result of each user in the interaction diagram data, wherein the project recommendation result is at least one item information in the interaction diagram data.
2. The item recommendation method according to claim 1, wherein said calculating, for any node, an augmented probability of the node from edges connected to the node comprises:
counting the number of nodes except the nodes in the interactive graph data aiming at any node to obtain the number of the nodes;
and calculating the ratio of the number of edges connected with the nodes to the number of the nodes to obtain the augmentation probability of the nodes.
3. The item recommendation method according to claim 1, further comprising, after said constructing the interaction graph data:
aiming at any edge, determining the number of times of interaction between a user node connected with the edge and a project node connected with the edge as the weight corresponding to the edge;
updating the corresponding edges in the interactive graph data according to the weight corresponding to each edge;
correspondingly, calculating the augmented probability of the node according to the edges connected with the node comprises:
and calculating the augmentation probability of the node according to the weight corresponding to the edge connected with the node.
4. The item recommendation method according to claim 3, wherein the calculating the augmented probability of the node according to the weight corresponding to the edge connected to the node comprises:
for any node, determining edges connected with the node as target edges, and calculating the sum of weights corresponding to the edges except the target edges in the interactive graph data to obtain an overall weight;
and calculating the ratio of the sum of the weights corresponding to the target edges to the overall weight to obtain the augmentation probability of the node.
5. The item recommendation method according to claim 1, wherein performing any one augmentation process on the interaction graph data according to the augmentation probabilities of all the nodes comprises:
aiming at any node, taking the augmentation probability of the node as the node abandon probability of the node;
according to the node abandon probability of the node, sampling whether the node is abandoned or not, and if the sampling result is abandoned, abandoning the node and an edge connected with the node from the interactive graph data;
and determining the discarded interactive map data as the interactive map data after the primary augmentation processing.
6. The item recommendation method according to claim 1, wherein performing any one augmentation process on the interaction graph data according to the augmentation probabilities of all the nodes comprises:
regarding any edge, taking the average value of the augmentation probabilities of the nodes connected with the edge as the edge abandon probability of the edge;
according to the edge rejection probability of the edge, sampling whether the edge is rejected, and if the sampling result is rejection, rejecting the edge from the interactive map data;
and determining the discarded interactive map data as the interactive map data after the augmentation processing.
7. The item recommendation method according to any one of claims 1 to 6, wherein the recommendation model comprises a first graph encoder and a second graph encoder;
the step of forming a training data set by using the preset augmented interactive map data, and updating the parameters of the recommendation model by using a back propagation algorithm based on a contrast loss function comprises the following steps:
combining any two augmented interactive graph data into a training sample to obtain M training samples, wherein M is an integer greater than zero;
inputting two pieces of augmentation data in any training sample into the first image encoder and the second image encoder respectively to obtain a first feature vector and a second feature vector;
and calculating the contrast loss function through the first feature vector and the second feature vector, and updating the parameters of the first image encoder and the second image encoder by adopting a back propagation algorithm on the basis of the contrast loss function.
8. An artificial intelligence based item recommendation device, characterized in that the item recommendation device comprises:
the graph data construction module is used for constructing interactive graph data, the interactive graph data comprises nodes and edges, the nodes comprise project nodes and user nodes, the node information of the project nodes comprises project information of a project, the node information of the user nodes comprises user information of a user, the edges are connected with the project nodes and the user nodes, and the edges are used for representing the interactive degree between the project nodes and the user nodes;
the probability calculation module is used for calculating the augmentation probability of any node according to the edge connected with the node;
the augmentation processing module is used for carrying out preset augmentation processing on the interactive map data according to the augmentation probabilities of all the nodes to obtain preset augmented interactive map data;
the model training module is used for forming a training data set by using the preset augmented interactive map data, updating parameters of a recommendation model by adopting a back propagation algorithm based on a comparison loss function until the comparison loss function is converged to obtain a trained recommendation model;
and the item recommendation module is used for inputting the interaction diagram data into the trained recommendation model to obtain an item recommendation result of each user in the interaction diagram data, wherein the item recommendation result is at least one item information in the interaction diagram data.
9. A computer device, characterized in that the computer device comprises a processor, a memory and a computer program stored in the memory and executable on the processor, the processor implementing the item recommendation method according to any one of claims 1 to 7 when executing the computer program.
10. A computer-readable storage medium, in which a computer program is stored which, when being executed by a processor, carries out a method of item recommendation according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210702599.8A CN114780863B (en) | 2022-06-21 | 2022-06-21 | Project recommendation method and device based on artificial intelligence, computer equipment and medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210702599.8A CN114780863B (en) | 2022-06-21 | 2022-06-21 | Project recommendation method and device based on artificial intelligence, computer equipment and medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114780863A true CN114780863A (en) | 2022-07-22 |
| CN114780863B CN114780863B (en) | 2022-09-13 |
Family
ID=82420510
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210702599.8A Active CN114780863B (en) | 2022-06-21 | 2022-06-21 | Project recommendation method and device based on artificial intelligence, computer equipment and medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114780863B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115809364A (en) * | 2022-09-30 | 2023-03-17 | 北京百度网讯科技有限公司 | Object recommendation method and model training method |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20130041876A1 (en) * | 2011-08-08 | 2013-02-14 | Paul Alexander Dow | Link recommendation and densification |
| US20140244667A1 (en) * | 2011-10-05 | 2014-08-28 | Telefonaktiebolaget L M Ericsson (Publ) | Method and Apparatuses for Enabling Recommendations |
| CN110879864A (en) * | 2019-10-29 | 2020-03-13 | 南京大学 | Context recommendation method based on graph neural network and attention mechanism |
| CN111428091A (en) * | 2020-03-19 | 2020-07-17 | 腾讯科技（深圳）有限公司 | Encoder training method, information recommendation method and related device |
| WO2021062219A1 (en) * | 2019-09-25 | 2021-04-01 | Google Llc | Clustering data using neural networks based on normalized cuts |
| CN113822444A (en) * | 2021-02-09 | 2021-12-21 | 日本电气株式会社 | Method, apparatus and computer-readable storage medium for model training and data processing |
| CN114372566A (en) * | 2022-03-21 | 2022-04-19 | 支付宝(杭州)信息技术有限公司 | Augmentation of graph data, graph neural network training method, device and equipment |
-
2022
- 2022-06-21 CN CN202210702599.8A patent/CN114780863B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20130041876A1 (en) * | 2011-08-08 | 2013-02-14 | Paul Alexander Dow | Link recommendation and densification |
| US20140244667A1 (en) * | 2011-10-05 | 2014-08-28 | Telefonaktiebolaget L M Ericsson (Publ) | Method and Apparatuses for Enabling Recommendations |
| WO2021062219A1 (en) * | 2019-09-25 | 2021-04-01 | Google Llc | Clustering data using neural networks based on normalized cuts |
| CN110879864A (en) * | 2019-10-29 | 2020-03-13 | 南京大学 | Context recommendation method based on graph neural network and attention mechanism |
| CN111428091A (en) * | 2020-03-19 | 2020-07-17 | 腾讯科技（深圳）有限公司 | Encoder training method, information recommendation method and related device |
| CN113822444A (en) * | 2021-02-09 | 2021-12-21 | 日本电气株式会社 | Method, apparatus and computer-readable storage medium for model training and data processing |
| CN114372566A (en) * | 2022-03-21 | 2022-04-19 | 支付宝(杭州)信息技术有限公司 | Augmentation of graph data, graph neural network training method, device and equipment |
Non-Patent Citations (1)
| Title |
|---|
| 于宏全等: "基于深度学习的铸件缺陷检测方法", 《电子机械工程》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115809364A (en) * | 2022-09-30 | 2023-03-17 | 北京百度网讯科技有限公司 | Object recommendation method and model training method |
| CN115809364B (en) * | 2022-09-30 | 2023-12-08 | 北京百度网讯科技有限公司 | Object recommendation method and model training method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114780863B (en) | 2022-09-13 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111523621B (en) | Image recognition method and device, computer equipment and storage medium | |
| CN108509463B (en) | Question response method and device | |
| CN108960409B (en) | Method and device for generating annotation data and computer-readable storage medium | |
| CN112434721A (en) | Image classification method, system, storage medium and terminal based on small sample learning | |
| US20220222918A1 (en) | Image retrieval method and apparatus, storage medium, and device | |
| CN112164391A (en) | Statement processing method and device, electronic equipment and storage medium | |
| CN111563192B (en) | Entity alignment method, device, electronic equipment and storage medium | |
| WO2018196718A1 (en) | Image disambiguation method and device, storage medium, and electronic device | |
| CN116664719B (en) | Image redrawing model training method, image redrawing method and device | |
| CN112819023A (en) | Sample set acquisition method and device, computer equipment and storage medium | |
| CN110728319B (en) | Image generation method and device and computer storage medium | |
| CN112528637A (en) | Text processing model training method and device, computer equipment and storage medium | |
| CN113408570A (en) | Image category identification method and device based on model distillation, storage medium and terminal | |
| CN110532562B (en) | Neural network training method, idiom misuse detection method and device and electronic equipment | |
| CN114780863B (en) | Project recommendation method and device based on artificial intelligence, computer equipment and medium | |
| CN112508048A (en) | Image description generation method and device | |
| CN115081615A (en) | Neural network training method, data processing method and equipment | |
| CN113408282B (en) | Method, device, equipment and storage medium for topic model training and topic prediction | |
| CN113726545B (en) | Network traffic generation method and device for generating countermeasure network based on knowledge enhancement | |
| CN112328881B (en) | Article recommendation method, device, terminal equipment and storage medium | |
| CN115618043B (en) | Text operation graph mutual inspection method and model training method, device, equipment and medium | |
| JP2022541832A (en) | Method and apparatus for retrieving images | |
| CN116958626A (en) | Image classification model training, image classification method and device and electronic equipment | |
| CN116958590A (en) | Media resource processing method and device, storage medium and electronic equipment | |
| CN112861474B (en) | Information labeling method, device, equipment and computer readable storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |