CN112507216B - Data object recommendation method, device, equipment and storage medium - Google Patents
Data object recommendation method, device, equipment and storage medium Download PDFInfo
- Publication number
- CN112507216B CN112507216B CN202011388041.4A CN202011388041A CN112507216B CN 112507216 B CN112507216 B CN 112507216B CN 202011388041 A CN202011388041 A CN 202011388041A CN 112507216 B CN112507216 B CN 112507216B
- Authority
- CN
- China
- Prior art keywords
- data
- user
- model
- data object
- prediction data
- 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.)
- Active
Links
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
- G06F16/40—Information retrieval; Database structures therefor; File system structures therefor of multimedia data, e.g. slideshows comprising image and additional audio data
- G06F16/43—Querying
- G06F16/435—Filtering based on additional data, e.g. user or group profiles
-
- 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
- G06N3/045—Combinations of networks
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Databases & Information Systems (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Biophysics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Computational Linguistics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Multimedia (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The embodiment of the invention discloses a data object recommendation method, a device, equipment and a storage medium. The method comprises the following steps: receiving a user feature vector from data object recommendation equipment and object feature vectors respectively corresponding to a plurality of data objects; generating model prediction data corresponding to each data object according to the user feature vector and the object feature vector corresponding to each data object respectively; and according to the model prediction data corresponding to each data object, determining the user interest degree corresponding to each data object and sending the user interest degree to the data object recommendation equipment so that the data object recommendation equipment executes data object recommendation operation according to the user interest degree corresponding to each data object. The method and the device only need to receive one user feature vector, do not need to receive a plurality of vector combinations, reduce the receiving times of the user feature vector, do not need to repeatedly calculate the same user feature vector, reduce the calculated amount of data object recommendation, and improve the recommendation efficiency.
Description
Technical Field
The present invention relates to the field of deep learning technologies, and in particular, to a data object recommendation method, apparatus, device, and storage medium.
Background
In the deep learning field, the recommendation class model can predict the interest degree of a user in each data object, and data object recommendation is performed according to the interest degree of the user in each data object. The data objects may be video, images and text. Such as: predicting the interest degree of the user on the videos, and after obtaining the interest degree of the user on a plurality of videos, sorting the videos with high interest degree to the user according to the interest degree from large to small.
When a user browses or acquires a data object, the data object with high user interest degree can be recommended to the user. When predicting the interest degree of a user on a data object, collecting the user characteristics of the user and the object characteristics of the data object, generating a fusion characteristic vector comprising the user characteristics and the object characteristics, and performing predictive analysis on the fusion characteristic vector by using a preset prediction algorithm to determine the interest degree of the user on the data object. Further, when recommending data objects, for the same user, the user's interest degree in hundreds of data objects is predicted at the same time, so that a fused feature vector including the feature vectors of the user features and the object features needs to be generated for each data object, that is, the fused feature vectors corresponding to the hundreds of data objects respectively include the user features, which results in that hundreds of times of repeated analysis and calculation on the user features are required, increasing the calculation amount of a prediction algorithm, and reducing the recommendation efficiency of the data objects.
Disclosure of Invention
The embodiment of the invention aims to provide a data object recommending method, a device, equipment and a storage medium, so as to solve the problem that the existing data object recommending mode requires repeated calculation of user characteristics.
Aiming at the technical problems, the embodiment of the invention is solved by the following technical scheme:
in a first aspect of the present invention, there is provided a data object recommendation method, including: receiving a user feature vector from data object recommendation equipment and object feature vectors respectively corresponding to a plurality of data objects; copying the user feature vectors to enable the number of the user feature vectors to be the same as the number of the object feature vectors; fusing each user feature vector with an object feature vector corresponding to one data object to obtain model prediction data corresponding to each data object; or, carrying out feature processing on the user feature vector to obtain user prediction data; performing feature processing on each object feature vector to obtain object prediction data corresponding to each data object; copying the user prediction data so that the number of the user prediction data is the same as the number of the object prediction data; fusing each piece of user prediction data with object prediction data corresponding to one piece of data object to obtain model prediction data corresponding to each piece of data object; and according to the model prediction data corresponding to each data object, determining the user interest degree corresponding to each data object and sending the user interest degree to the data object recommending equipment so that the data object recommending equipment executes data object recommending operation according to the user interest degree corresponding to each data object.
The determining the user interested degree corresponding to each data object according to the model predicted data corresponding to each data object comprises the following steps: and sequentially inputting model prediction data corresponding to each data object into a preset recommendation class model, or inputting model prediction data corresponding to a plurality of data objects into the recommendation class model together, and determining the user interested degree corresponding to each data object through the recommendation class model.
Wherein the copying of the user feature vectors makes the number of the user feature vectors the same as the number of the object feature vectors; fusing each user feature vector with an object feature vector corresponding to one data object to obtain model prediction data corresponding to each data object, wherein the model prediction data comprises: inputting the user feature vectors and the object feature vectors corresponding to each data object together into a preset recommendation model, and copying the user feature vectors through a first layer sub-model in the recommendation model to enable the number of the user feature vectors to be the same as the number of the object feature vectors; fusing each user feature vector with an object feature vector corresponding to one data object to obtain model prediction data corresponding to each data object; the determining the user interest degree corresponding to each data object according to the model prediction data corresponding to each data object comprises the following steps: and sequentially inputting the model prediction data corresponding to each data object into a subsequent sub-model of the first-layer sub-model, or inputting the model prediction data corresponding to a plurality of data objects into the subsequent sub-model of the first-layer sub-model together, and determining the user interest degree corresponding to each data object through the subsequent sub-model.
The fusing each user feature vector with an object feature vector corresponding to one data object to obtain model prediction data corresponding to each data object includes: and according to a preset vector arrangement sequence, arranging and splicing the user feature vector and the object feature vector corresponding to the data object to obtain model prediction data corresponding to the data object.
The user characteristic vector is subjected to characteristic processing to obtain user prediction data; performing feature processing on each object feature vector to obtain object prediction data corresponding to each data object; copying the user prediction data so that the number of the user prediction data is the same as the number of the object prediction data; fusing each piece of user prediction data with object prediction data corresponding to one piece of data object to obtain model prediction data corresponding to each piece of data object, wherein the method comprises the following steps: inputting the user feature vector and the object feature vector corresponding to each data object into a preset recommendation model, and executing the following steps through a first-layer sub-model in the recommendation model: performing feature processing on the user feature vector through a first model unit in the first layer sub-model to obtain user prediction data; performing feature processing on each object feature vector through a second model unit in the first layer sub-model to obtain object prediction data corresponding to each data object; copying the user prediction data to enable the quantity of the user prediction data to be the same as the quantity of the object prediction data; for each data object, according to a preset parameter matrix, fusing each piece of user prediction data with object prediction data corresponding to one data object to obtain model prediction data corresponding to each data object; the determining the user interest degree corresponding to each data object according to the model prediction data corresponding to each data object comprises the following steps: and sequentially inputting the model prediction data corresponding to each data object into a subsequent sub-model of the first-layer sub-model, or inputting the model prediction data corresponding to a plurality of data objects into the subsequent sub-model of the first-layer sub-model together, and determining the user interest degree corresponding to each data object through the subsequent sub-model.
Wherein the parameter matrix comprises: a user parameter matrix and an object parameter matrix; fusing each piece of user prediction data with object prediction data corresponding to one data object according to a preset parameter matrix to obtain model prediction data corresponding to each data object, wherein the method comprises the following steps: determining the product of the user prediction data and the user parameter matrix to obtain a first matrix product; determining the product of the object prediction data and the object parameter matrix to obtain a second matrix product; and taking the sum of the first matrix product and the second matrix product as model prediction data corresponding to the data object.
Wherein the categories of the data objects include: video data objects, picture data objects, and text data objects.
In a second aspect of the present invention, there is provided a data object recommendation apparatus, comprising: the receiving module is used for receiving a user characteristic vector from the data object recommending equipment and object characteristic vectors respectively corresponding to a plurality of data objects; the generation module is used for copying the user characteristic vectors, so that the number of the user characteristic vectors is the same as the number of the object characteristic vectors; fusing each user feature vector with an object feature vector corresponding to one data object to obtain model prediction data corresponding to each data object; or the generating module is used for carrying out feature processing on the user feature vector to obtain user prediction data; performing feature processing on each object feature vector to obtain object prediction data corresponding to each data object; copying the user prediction data so that the number of the user prediction data is the same as the number of the object prediction data; fusing each piece of user prediction data with object prediction data corresponding to one piece of data object to obtain model prediction data corresponding to each piece of data object; and the sending module is used for determining the user interest degree corresponding to each data object according to the model prediction data corresponding to each data object and sending the data object to the data object recommending equipment so that the data object recommending equipment executes data object recommending operation according to the user interest degree corresponding to each data object.
In a third aspect of the present invention, there is provided an electronic device including a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other via the communication bus; a memory for storing a computer program; and the processor is used for realizing any one of the data object recommending method steps when executing the program stored in the memory.
In a fourth aspect of the invention, there is provided a computer readable storage medium having instructions stored therein which, when executed on a computer, cause the computer to perform any of the data object recommendation method steps described above.
In a fifth aspect of the invention there is provided a computer program product comprising instructions which, when run on a computer, cause the computer to perform the data object recommendation method steps of any of the above.
According to the data object recommendation method, device, equipment and storage medium provided by the embodiment of the invention, the user interested degree corresponding to each data object is determined according to one user characteristic vector and the object characteristic vectors corresponding to the data objects respectively, in the process, only one user characteristic vector is received, a plurality of vector combinations (combination of the user characteristic vector and the object characteristic vector) are not required to be received, the receiving times of the user characteristic vector are reduced, the bandwidth occupied by the user characteristic vector is reduced, the calculation of the same user characteristic vector is not required to be repeated, the calculated amount of data object recommendation is reduced, and the recommendation efficiency of the data object is improved.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below.
FIG. 1 is a flow chart of a data object recommendation method according to an embodiment of the invention;
FIG. 2 is a flowchart illustrating a data object recommendation method according to an embodiment of the present invention;
FIG. 3 is a network architecture diagram of a data object recommendation method according to an embodiment of the present invention;
FIG. 4 is a specific flow chart of a data object recommendation method according to another embodiment of the present invention;
FIG. 5 is a schematic diagram of a data object recommendation method according to an embodiment of the invention;
FIG. 6 is a detailed flow chart of a data object recommendation method according to yet another embodiment of the present invention;
FIG. 7 is a schematic diagram of a data object recommendation method according to another embodiment of the present invention;
FIG. 8 is a block diagram of a data object recommendation device according to an embodiment of the present invention;
fig. 9 is a block diagram of an electronic device according to an embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be described below with reference to the accompanying drawings of the embodiments of the present invention.
According to the embodiment of the invention, a data object recommendation method is provided. The execution subject of the invention is a recommendation server. Wherein the recommendation server may run an online recommendation algorithm. The online recommendation algorithm may be a pre-trained recommendation class model. The recommended type model is, for example, a Tensorflow (TF) model. The recommendation class model may be provided in a TF service container of a recommendation server (physical machine).
FIG. 1 is a flow chart of a data object recommendation method according to an embodiment of the invention.
In step S110, a user feature vector and object feature vectors corresponding to a plurality of data objects from the data object recommendation device are received.
And the data object recommendation device is used for acquiring the user characteristics of the user and generating a user characteristic vector, and acquiring the object characteristics of the data object and generating an object characteristic vector.
A data object recommendation device comprising: a client and a data server. When the data object recommending device is a client, the client is responsible for acquiring the user characteristics of the login user, and acquiring the object characteristics corresponding to the candidate recommended data objects from the data server. When the data object recommending device is a data server, the data server is responsible for acquiring the user characteristics of the login user and the object characteristics corresponding to the candidate recommended data objects.
User features include, but are not limited to: gender, age, region, and preference of the user.
The data object refers to a candidate recommended data object. Object features include, but are not limited to: attributes of the data object. Wherein the types of data objects include, but are not limited to: video data objects, picture data objects, and text data objects.
When the data object is a video data object, the object feature vectors include, but are not limited to: the name, category (e.g., administrative, entertainment, sports, etc.), and duration of the video data object.
When the data object is a picture data object, the object feature vectors include, but are not limited to: the names and categories of the video data objects (e.g., nature scene, character writing, etc.).
When the data object is a text data object, the object feature vectors include, but are not limited to: the name and category of the video data object (e.g., fantasy, history, mood, etc.).
Step S120, copying the user feature vectors to enable the number of the user feature vectors to be the same as the number of the object feature vectors; fusing each user feature vector with an object feature vector corresponding to one data object to obtain model prediction data corresponding to each data object; or, carrying out feature processing on the user feature vector to obtain user prediction data; performing feature processing on each object feature vector to obtain object prediction data corresponding to each data object; copying the user prediction data so that the number of the user prediction data is the same as the number of the object prediction data; and fusing each piece of user prediction data with object prediction data corresponding to one data object to obtain model prediction data corresponding to each data object.
According to the user feature vector and the object feature vector corresponding to each data object, model prediction data corresponding to each data object are generated.
The model prediction data is used to predict the user's level of interest in the data object.
The model prediction data is a fusion feature vector or an intermediate processing result.
The fusion feature vector is obtained by fusing the user feature vector and the object fusion vector.
The intermediate processing result is obtained by respectively processing the user feature vector and the object feature vector and then fusing the two obtained results.
The manner in which the model prediction data is generated will be described later, and therefore will not be described in detail herein.
Step S130, determining a user interest degree corresponding to each data object according to the model prediction data corresponding to each data object, and sending the data object prediction data to the data object recommendation device, so that the data object recommendation device executes a data object recommendation operation according to the user interest degree corresponding to each data object.
The user interest level refers to the user's interest level in the data object.
In the embodiment of the invention, the user interested degree corresponding to each data object is determined according to one user characteristic vector and the object characteristic vectors corresponding to the plurality of data objects respectively, in the process, only one user characteristic vector is received without receiving a plurality of vector combinations (combination of the user characteristic vector and the object characteristic vector), so that the receiving times of the user characteristic vector are reduced, the occupied bandwidth of the user characteristic vector is reduced, the calculation of the same user characteristic vector is not required to be repeated, the calculation amount of data object recommendation is reduced, and the recommendation efficiency of the data object is improved.
A more specific data object recommendation method is provided below. According to the embodiment of the invention, model prediction data are input into a recommendation model, and the user interest degree corresponding to the data object is determined through the recommendation model.
FIG. 2 is a flowchart of a data object recommendation method according to an embodiment of the present invention.
In step S210, a user feature vector and object feature vectors corresponding to a plurality of data objects from the data object recommendation device are received.
A client in the data object recommendation device sends a recommendation prediction request to a recommendation server. The recommendation prediction request comprises a user characteristic vector and object characteristic vectors respectively corresponding to a plurality of data objects.
When the client sends the recommendation prediction request, the number of the user feature vectors is not required to be aligned with the number of the object feature vectors, and only one user feature vector is required to be sent.
And step S220, copying the user characteristic vectors so that the number of the user characteristic vectors is the same as the number of the object characteristic vectors.
That is, before the model prediction data is input to a preset recommendation type model, the user feature vectors need to be copied so that the number of the user feature vectors is the same as the number of the object feature vectors.
Step S230, fusing each piece of user prediction data with the object prediction data corresponding to one data object to obtain model prediction data corresponding to each data object.
Since the number of the user feature vectors is the same as the number of the object feature vectors, the number of vector combinations may be divided, each vector combination including one user feature vector and one object feature vector; and splicing the user characteristic vector and the object characteristic vector in each vector combination into a fusion characteristic vector, wherein the spliced characteristic vector is model prediction data.
Further, according to a preset vector arrangement sequence, the user feature vectors and the object feature vectors corresponding to the data objects are arranged and spliced to obtain model prediction data corresponding to the data objects.
For example: the user feature vector is in front, the object feature vector is in back, and the fusion feature vector is formed.
Step S240, sequentially inputting the model prediction data corresponding to each data object into a preset recommendation class model, or inputting the model prediction data corresponding to each data object into the recommendation class model, and determining the user interest level corresponding to each data object through the recommendation class model.
Before model prediction data corresponding to a plurality of data objects are input into the recommendation model together, the model prediction data corresponding to the plurality of data objects can be fused into a model prediction matrix; wherein each row or column in the model prediction matrix is one model prediction data.
In this embodiment, a user feature vector and a plurality of object feature vectors are received, so that the transmission amount of the user feature vector is reduced, and the transmission bandwidth of the vector is reduced. Before the recommendation type model is used for determining the user interest degree, the user feature vector is copied, so that the recommendation type model can normally use the fusion feature vector for determining the user interest degree, and the request delay is reduced.
In the embodiment of the present invention, the steps of the data object recommendation method of the present embodiment may be executed in parallel by multiple processes. As shown in fig. 3, a network architecture diagram of a data object recommendation method according to an embodiment of the present invention is shown. In fig. 3, the recommendation class model is set in a TF service container in a physical machine, and a plurality of clients (clients are set in a data object recommendation device) are connected to the TF service container, respectively, so as to transmit and receive data between the clients and the TF service container. Of course, the client in this embodiment may be a data server.
Another more specific data object recommendation method is provided below. The recommendation class model is used for determining the user interest degree corresponding to the data object according to the user feature vector and the object feature vector.
FIG. 4 is a flowchart illustrating a data object recommendation method according to another embodiment of the present invention.
In step S410, a user feature vector and object feature vectors corresponding to a plurality of data objects from the data object recommendation device are received.
Step S420, inputting the user feature vector and the object feature vector corresponding to each data object into a preset recommendation model.
In this embodiment, after the user feature vector and the object feature vector corresponding to each data object are input into a preset recommendation class model, the following operations from step S430 to step S450 are performed through a first layer sub-model in the recommendation class model.
The first sub-model and the subsequent sub-model are divided in the recommended class model in advance. The first-layer sub-model refers to a sub-model for first executing operation in the recommended class model. The subsequent sub-model refers to the remaining models in the recommended class model other than the first-layer sub-model.
In step S430, the first layer sub-model replicates the user feature vectors so that the number of user feature vectors is the same as the number of object feature vectors.
In step S440, the first-layer sub-model fuses each user feature vector with an object feature vector corresponding to one data object, to obtain model prediction data corresponding to each data object.
And when each user characteristic vector is fused with an object characteristic vector corresponding to one data object, arranging and splicing the user characteristic vector and the object characteristic vector corresponding to the data object according to a preset vector arrangement sequence to obtain model prediction data corresponding to the data object.
The vector arrangement order is, for example, that the user feature vector is preceding and the object feature vector is following; alternatively, the object feature vector is forward and the user feature vector is backward.
In step S450, the first-layer sub-model sequentially inputs the model prediction data corresponding to each data object into a subsequent sub-model of the first-layer sub-model, or inputs the model prediction data corresponding to each data object into the subsequent sub-model of the first-layer sub-model, and determines the user interest level corresponding to each data object through the subsequent sub-model.
In this embodiment, a user feature vector and a plurality of object feature vectors are received, so that the transmission amount of the user feature vector is reduced, and the transmission bandwidth of the vector is reduced. And copying the user feature vector through a first-layer sub-model in the recommendation model, so that a subsequent sub-model in the recommendation model can normally use the fusion feature vector to determine the user interest degree.
For example: FIG. 5 is a schematic diagram of a data object recommendation method according to an embodiment of the invention. Assuming that the user has 3 user features, the names of the 3 user features are user feature1, user feature2 and user feature3, respectively, and the values corresponding to the 3 user features are u1, u2 and u3, respectively. Suppose there are 3 videos, v1, v2 and v3, respectively. Each video has 3 object features, the names of the 3 object features are video feature1, video feature3 and video feature3, and the values of the video v1 corresponding to the above 3 object features are v11, v21 and v31 respectively; the values of the video v2 corresponding to the above 3 object features are v12, v22 and v32 respectively; the values of the video v3 corresponding to the above 3 object features are v13, v23 and v33, respectively. The first layer sub-model is a first layer of the neural network, and before the first layer of the neural network, u1, u2 and u3 are duplicated to obtain three groups of u1, u2 and u3, and then u1, u2, u3, v11, v21 and v31 can be fused into fusion feature vectors corresponding to the video v 1; fusing u1, u2, u3, v12, v22 and v32 into a fusion feature vector corresponding to the video v 2; the u1, u2, u3, v13, v23 and v33 are fused into a fused feature vector corresponding to the video v3. The 3 fusion feature vectors are sequentially input into a first layer of the neural network, and the user interested degrees respectively corresponding to the video v1, v2 and v3 can be output from a second layer of the neural network to an nth layer of the neural network.
Another more specific data object recommendation method is provided below. In this embodiment, after receiving a user feature vector from a data object recommendation device and object feature vectors corresponding to a plurality of data objects, the user feature vector and the object feature vector corresponding to each data object are input into a preset recommendation model, and the steps of fig. 6 are executed through the recommendation model.
FIG. 6 is a flowchart illustrating a data object recommendation method according to a further embodiment of the present invention.
In step S610, a user feature vector input to the recommendation class model and object feature vectors corresponding to the plurality of data objects are received.
And step S620, performing feature processing on the user feature vector through a first model unit in the first layer sub-model to obtain user prediction data.
Step S630, performing feature processing on each object feature vector through a second model unit in the first layer sub-model, to obtain object prediction data corresponding to each data object; wherein the first model unit and the second model unit are in a first layer sub-model of the recommendation class model.
And step S640, copying the user prediction data so that the number of the user prediction data is the same as the number of the object prediction data.
Step S650, for each data object, according to a preset parameter matrix, fuses each piece of user prediction data with the object prediction data corresponding to one data object, so as to obtain model prediction data corresponding to each data object.
In the first layer sub-model, it includes: a user parameter matrix and an object parameter matrix; determining the product of the user prediction data and the user parameter matrix to obtain a first matrix product; determining the product of the object prediction data and the object parameter matrix to obtain a second matrix product; and taking the sum of the first matrix product and the second matrix product as model prediction data corresponding to the data object.
For example: model prediction data = user prediction data x user parameter matrix + object prediction data x object parameter matrix.
Step S660, sequentially inputting the model prediction data corresponding to each data object into a subsequent sub-model of the first-layer sub-model, or inputting the model prediction data corresponding to each of the plurality of data objects into the subsequent sub-model of the first-layer sub-model, and determining the user interest level corresponding to each data object through the subsequent sub-model.
In this embodiment, a user feature vector and a plurality of object feature vectors are received, so that the transmission amount of the user feature vector is reduced, and the transmission bandwidth of the vector is reduced. Processing the user feature vector through a first layer sub-model in the recommendation type model to obtain user prediction data, and processing the object feature vector to obtain object prediction data; before the user prediction data and the object feature data are input into the subsequent sub-model, the user prediction data are copied, so that the user prediction data and the object prediction data can be combined one by one, and each combination is fused into model prediction data, the user feature vector only needs to be processed once, the subsequent sub-model in the recommendation type model can be used for determining the user interest degree according to the model prediction data as usual, and the request delay can be reduced.
For example: fig. 7 is a schematic diagram of a data object recommendation method according to another embodiment of the present invention. Assuming that the user has 3 user features, the names of the 3 user features are user feature1, user feature2 and user feature3, respectively, and the values corresponding to the 3 user features are u1, u2 and u3, respectively. Suppose there are 3 videos, v1, v2 and v3, respectively. Each video has 3 object features, the names of the 3 object features are video feature1, video feature3 and video feature3, and the values of the video v1 corresponding to the above 3 object features are v11, v21 and v31 respectively; the values of the video v2 corresponding to the above 3 object features are v12, v22 and v32 respectively; the values of the video v3 corresponding to the above 3 object features are v13, v23 and v33, respectively. The first layer sub-model is a first layer of the neural network, and the first layer of the neural network comprises a first layer A of the neural network and a first layer B of the neural network. The first layer A of the neural network processes u1, u2 and u3 to obtain user prediction data u1', u2' and u3'. The first layer B of the neural network respectively processes v11, v21 and v31 corresponding to the video v1 to obtain object prediction data v11', v21' and v31' corresponding to the video v 1; processing v12, v22 and v32 corresponding to the video v2 to obtain object prediction data v12', v22' and v32' corresponding to the video v 2; and processing v13, v23 and v33 corresponding to the video v3 to obtain object prediction data v13', v23' and v33' corresponding to the video v3. Copying user prediction data u1', u2' and u3 'to obtain three groups of u1', u2 'and u3', and fusing u1', u2' and u3', v11', v21 'and v31' into model prediction data corresponding to the video v 1; fusing u1', u2' and u3', v12', v22 'and v32' into model prediction data corresponding to the video v 2; and u1', u2' and u3', v13', v23 'and v33' are fused into model prediction data corresponding to the video v3. 3 model prediction data are sequentially input into a second layer of the neural network, and the user interested degree corresponding to the video v1, v2 and v3 respectively can be output from the second layer of the neural network to the nth layer of the neural network.
The embodiment of the invention also provides a data object recommending device. Fig. 8 is a block diagram of a data object recommending apparatus according to an embodiment of the present invention.
The data object recommendation device includes: a receiving module 810, a generating module 820 and a transmitting module 830.
The receiving module 810 is configured to receive a user feature vector from the data object recommendation device and object feature vectors corresponding to a plurality of data objects respectively.
A generating module 820, configured to copy the user feature vectors so that the number of the user feature vectors is the same as the number of the object feature vectors; fusing each user feature vector with an object feature vector corresponding to one data object to obtain model prediction data corresponding to each data object; or the generating module is used for carrying out feature processing on the user feature vector to obtain user prediction data; performing feature processing on each object feature vector to obtain object prediction data corresponding to each data object; copying the user prediction data so that the number of the user prediction data is the same as the number of the object prediction data; and fusing each piece of user prediction data with object prediction data corresponding to one data object to obtain model prediction data corresponding to each data object.
And the sending module 830 is configured to determine, according to the model prediction data corresponding to each data object, a user interest degree corresponding to each data object and send the data object prediction data to the data object recommendation device, so that the data object recommendation device executes a data object recommendation operation according to the user interest degree corresponding to each data object.
The functions of the apparatus according to the embodiments of the present invention have been described in the foregoing method embodiments, so that details of the descriptions of the embodiments of the present invention are not detailed, and reference may be made to the related descriptions in the foregoing embodiments, which are not repeated herein.
The embodiment of the invention also provides an electronic device, as shown in fig. 9, which includes a processor 910, a communication interface 920, a memory 930, and a communication bus 940, where the processor 910, the communication interface 920, and the memory 930 implement communication between each other through the communication bus 940.
Memory 930 for storing the computer programs.
Processor 910, when executing a program stored on memory 930, performs the following steps:
receiving a user feature vector from data object recommendation equipment and object feature vectors respectively corresponding to a plurality of data objects; copying the user feature vectors to enable the number of the user feature vectors to be the same as the number of the object feature vectors; fusing each user feature vector with an object feature vector corresponding to one data object to obtain model prediction data corresponding to each data object; or, carrying out feature processing on the user feature vector to obtain user prediction data; performing feature processing on each object feature vector to obtain object prediction data corresponding to each data object; copying the user prediction data so that the number of the user prediction data is the same as the number of the object prediction data; fusing each piece of user prediction data with object prediction data corresponding to one piece of data object to obtain model prediction data corresponding to each piece of data object; and according to the model prediction data corresponding to each data object, determining the user interest degree corresponding to each data object and sending the user interest degree to the data object recommending equipment so that the data object recommending equipment executes data object recommending operation according to the user interest degree corresponding to each data object.
The determining the user interested degree corresponding to each data object according to the model predicted data corresponding to each data object comprises the following steps: and sequentially inputting model prediction data corresponding to each data object into a preset recommendation class model, or inputting model prediction data corresponding to a plurality of data objects into the recommendation class model together, and determining the user interested degree corresponding to each data object through the recommendation class model.
Wherein the copying of the user feature vectors makes the number of the user feature vectors the same as the number of the object feature vectors; fusing each user feature vector with an object feature vector corresponding to one data object to obtain model prediction data corresponding to each data object, wherein the model prediction data comprises: inputting the user feature vectors and the object feature vectors corresponding to each data object together into a preset recommendation model, and copying the user feature vectors through a first layer sub-model in the recommendation model to enable the number of the user feature vectors to be the same as the number of the object feature vectors; fusing each user feature vector with an object feature vector corresponding to one data object to obtain model prediction data corresponding to each data object; the determining the user interest degree corresponding to each data object according to the model prediction data corresponding to each data object comprises the following steps: and sequentially inputting the model prediction data corresponding to each data object into a subsequent sub-model of the first-layer sub-model, or inputting the model prediction data corresponding to a plurality of data objects into the subsequent sub-model of the first-layer sub-model together, and determining the user interest degree corresponding to each data object through the subsequent sub-model.
The fusing each user feature vector with an object feature vector corresponding to one data object to obtain model prediction data corresponding to each data object includes: and according to a preset vector arrangement sequence, arranging and splicing the user feature vector and the object feature vector corresponding to the data object to obtain model prediction data corresponding to the data object.
The user characteristic vector is subjected to characteristic processing to obtain user prediction data; performing feature processing on each object feature vector to obtain object prediction data corresponding to each data object; copying the user prediction data so that the number of the user prediction data is the same as the number of the object prediction data; fusing each piece of user prediction data with object prediction data corresponding to one piece of data object to obtain model prediction data corresponding to each piece of data object, wherein the method comprises the following steps: inputting the user feature vector and the object feature vector corresponding to each data object into a preset recommendation model, and executing the following steps through a first-layer sub-model in the recommendation model: performing feature processing on the user feature vector through a first model unit in the first layer sub-model to obtain user prediction data; performing feature processing on each object feature vector through a second model unit in the first layer sub-model to obtain object prediction data corresponding to each data object; copying the user prediction data to enable the quantity of the user prediction data to be the same as the quantity of the object prediction data; for each data object, according to a preset parameter matrix, fusing each piece of user prediction data with object prediction data corresponding to one data object to obtain model prediction data corresponding to each data object; the determining the user interest degree corresponding to each data object according to the model prediction data corresponding to each data object comprises the following steps: and sequentially inputting the model prediction data corresponding to each data object into a subsequent sub-model of the first-layer sub-model, or inputting the model prediction data corresponding to a plurality of data objects into the subsequent sub-model of the first-layer sub-model together, and determining the user interest degree corresponding to each data object through the subsequent sub-model.
Wherein the parameter matrix comprises: a user parameter matrix and an object parameter matrix; fusing each piece of user prediction data with object prediction data corresponding to one data object according to a preset parameter matrix to obtain model prediction data corresponding to each data object, wherein the method comprises the following steps: determining the product of the user prediction data and the user parameter matrix to obtain a first matrix product; determining the product of the object prediction data and the object parameter matrix to obtain a second matrix product; and taking the sum of the first matrix product and the second matrix product as model prediction data corresponding to the data object.
Wherein the categories of the data objects include: video data objects, picture data objects, and text data objects.
The communication bus mentioned by the above terminal may be a peripheral component interconnect standard (Peripheral Component Interconnect, abbreviated as PCI) bus or an extended industry standard architecture (Extended Industry Standard Architecture, abbreviated as EISA) bus, etc. The communication bus may be classified as an address bus, a data bus, a control bus, or the like. For ease of illustration, the figures are shown with only one bold line, but not with only one bus or one type of bus.
The communication interface is used for communication between the terminal and other devices.
The memory may include random access memory (Random Access Memory, RAM) or non-volatile memory (non-volatile memory), such as at least one disk memory. Optionally, the memory may also be at least one memory device located remotely from the aforementioned processor.
The processor may be a general-purpose processor, including a central processing unit (Central Processing Unit, CPU for short), a network processor (Network Processor, NP for short), etc.; but also digital signal processors (Digital Signal Processing, DSP for short), application specific integrated circuits (Application Specific Integrated Circuit, ASIC for short), field-programmable gate arrays (Field-Programmable Gate Array, FPGA for short) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.
In yet another embodiment of the present invention, a computer readable storage medium is provided, in which instructions are stored, which when run on a computer, cause the computer to perform the data object recommendation method according to any of the above embodiments.
In yet another embodiment of the present invention, a computer program product comprising instructions which, when run on a computer, cause the computer to perform the data object recommendation method of any of the above embodiments is also provided.
In the above embodiments, it may be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented in software, may be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When loaded and executed on a computer, produces a flow or function in accordance with embodiments of the present invention, in whole or in part. The computer may be a general purpose computer, a special purpose computer, a computer network, or other programmable apparatus. The computer instructions may be stored in or transmitted from one computer-readable storage medium to another, for example, by wired (e.g., coaxial cable, optical fiber, digital Subscriber Line (DSL)), or wireless (e.g., infrared, wireless, microwave, etc.). The computer readable storage medium may be any available medium that can be accessed by a computer or a data storage device such as a server, data center, etc. that contains an integration of one or more available media. The usable medium may be a magnetic medium (e.g., floppy Disk, hard Disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid State Disk (SSD)), etc.
It is noted that relational terms such as first and second, and the like are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.
In this specification, each embodiment is described in a related manner, and identical and similar parts of each embodiment are all referred to each other, and each embodiment mainly describes differences from other embodiments. In particular, for system embodiments, since they are substantially similar to method embodiments, the description is relatively simple, as relevant to see a section of the description of method embodiments.
The foregoing description is only of the preferred embodiments of the present invention and is not intended to limit the scope of the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention are included in the protection scope of the present invention.
Claims (9)
1. A data object recommendation method, comprising:
receiving a user feature vector from data object recommendation equipment and object feature vectors respectively corresponding to a plurality of data objects;
copying the user feature vectors to enable the number of the user feature vectors to be the same as the number of the object feature vectors; fusing each user feature vector with an object feature vector corresponding to one data object to obtain model prediction data corresponding to each data object, wherein the model prediction data comprises: inputting the user feature vectors and the object feature vectors corresponding to each data object together into a preset recommendation model, and copying the user feature vectors through a first layer sub-model in the recommendation model to enable the number of the user feature vectors to be the same as the number of the object feature vectors; fusing each user feature vector with an object feature vector corresponding to one data object to obtain model prediction data corresponding to each data object;
Or, carrying out feature processing on the user feature vector to obtain user prediction data; performing feature processing on each object feature vector to obtain object prediction data corresponding to each data object; copying the user prediction data so that the number of the user prediction data is the same as the number of the object prediction data; fusing each piece of user prediction data with object prediction data corresponding to one piece of data object to obtain model prediction data corresponding to each piece of data object;
according to the model prediction data corresponding to each data object, determining the user interest degree corresponding to each data object and sending the user interest degree to the data object recommending equipment so that the data object recommending equipment executes data object recommending operation according to the user interest degree corresponding to each data object;
the determining the user interest degree corresponding to each data object according to the model prediction data corresponding to each data object comprises the following steps: and sequentially inputting the model prediction data corresponding to each data object into a subsequent sub-model of the first-layer sub-model, or inputting the model prediction data corresponding to a plurality of data objects into the subsequent sub-model of the first-layer sub-model together, and determining the user interest degree corresponding to each data object through the subsequent sub-model.
2. The method of claim 1, wherein determining the user interest level for each of the data objects based on model prediction data for each of the data objects comprises:
and sequentially inputting model prediction data corresponding to each data object into a preset recommendation class model, or inputting model prediction data corresponding to a plurality of data objects into the recommendation class model together, and determining the user interested degree corresponding to each data object through the recommendation class model.
3. The method according to any one of claims 1-2, wherein the fusing each user feature vector with an object feature vector corresponding to one data object to obtain model prediction data corresponding to each data object includes:
and according to a preset vector arrangement sequence, arranging and splicing the user feature vector and the object feature vector corresponding to the data object to obtain model prediction data corresponding to the data object.
4. The method of claim 1, wherein the step of determining the position of the substrate comprises,
performing feature processing on the user feature vector to obtain user prediction data; performing feature processing on each object feature vector to obtain object prediction data corresponding to each data object; copying the user prediction data so that the number of the user prediction data is the same as the number of the object prediction data; fusing each piece of user prediction data with object prediction data corresponding to one piece of data object to obtain model prediction data corresponding to each piece of data object, wherein the method comprises the following steps:
Inputting the user feature vector and the object feature vector corresponding to each data object into a preset recommendation model, and executing the following steps through a first-layer sub-model in the recommendation model:
performing feature processing on the user feature vector through a first model unit in the first layer sub-model to obtain user prediction data;
performing feature processing on each object feature vector through a second model unit in the first layer sub-model to obtain object prediction data corresponding to each data object;
copying the user prediction data to enable the quantity of the user prediction data to be the same as the quantity of the object prediction data;
and fusing each piece of user prediction data with the object prediction data corresponding to one data object according to a preset parameter matrix aiming at each data object to obtain model prediction data corresponding to each data object.
5. The method of claim 4, wherein the step of determining the position of the first electrode is performed,
the parameter matrix comprises: a user parameter matrix and an object parameter matrix;
fusing each piece of user prediction data with object prediction data corresponding to one data object according to a preset parameter matrix to obtain model prediction data corresponding to each data object, wherein the method comprises the following steps:
Determining the product of the user prediction data and the user parameter matrix to obtain a first matrix product;
determining the product of the object prediction data and the object parameter matrix to obtain a second matrix product;
and taking the sum of the first matrix product and the second matrix product as model prediction data corresponding to the data object.
6. The method according to any one of claims 1-2, 4-5, wherein the categories of data objects include: video data objects, picture data objects, and text data objects.
7. A data object recommendation device, comprising:
the receiving module is used for receiving a user characteristic vector from the data object recommending equipment and object characteristic vectors respectively corresponding to a plurality of data objects;
the generation module is used for copying the user characteristic vectors, so that the number of the user characteristic vectors is the same as the number of the object characteristic vectors; fusing each user feature vector with an object feature vector corresponding to one data object to obtain model prediction data corresponding to each data object, wherein the model prediction data comprises: inputting the user feature vectors and the object feature vectors corresponding to each data object together into a preset recommendation model, and copying the user feature vectors through a first layer sub-model in the recommendation model to enable the number of the user feature vectors to be the same as the number of the object feature vectors; fusing each user feature vector with an object feature vector corresponding to one data object to obtain model prediction data corresponding to each data object; or the generating module is used for carrying out feature processing on the user feature vector to obtain user prediction data; performing feature processing on each object feature vector to obtain object prediction data corresponding to each data object; copying the user prediction data so that the number of the user prediction data is the same as the number of the object prediction data; fusing each piece of user prediction data with object prediction data corresponding to one piece of data object to obtain model prediction data corresponding to each piece of data object;
The sending module is used for determining the user interest degree corresponding to each data object according to the model prediction data corresponding to each data object and sending the data object to the data object recommending equipment so that the data object recommending equipment executes data object recommending operation according to the user interest degree corresponding to each data object;
the determining the user interest degree corresponding to each data object according to the model prediction data corresponding to each data object comprises the following steps: and sequentially inputting the model prediction data corresponding to each data object into a subsequent sub-model of the first-layer sub-model, or inputting the model prediction data corresponding to a plurality of data objects into the subsequent sub-model of the first-layer sub-model together, and determining the user interest degree corresponding to each data object through the subsequent sub-model.
8. The electronic equipment is characterized by comprising a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory are communicated with each other through the communication bus;
a memory for storing a computer program;
a processor for implementing the data object recommendation method steps of any of claims 1-6 when executing a program stored on a memory.
9. A computer readable storage medium having stored thereon a computer program, which when executed by a processor implements a data object recommendation method according to any of claims 1-6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011388041.4A CN112507216B (en) | 2020-12-01 | 2020-12-01 | Data object recommendation method, device, equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011388041.4A CN112507216B (en) | 2020-12-01 | 2020-12-01 | Data object recommendation method, device, equipment and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112507216A CN112507216A (en) | 2021-03-16 |
| CN112507216B true CN112507216B (en) | 2023-07-18 |
Family
ID=74969208
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011388041.4A Active CN112507216B (en) | 2020-12-01 | 2020-12-01 | Data object recommendation method, device, equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112507216B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113722537B (en) * | 2021-08-11 | 2024-04-26 | 北京奇艺世纪科技有限公司 | Short video ordering and model training method and device, electronic equipment and storage medium |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10244286B1 (en) * | 2018-01-30 | 2019-03-26 | Fmr Llc | Recommending digital content objects in a network environment |
Family Cites Families (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| HUP1000408A2 (en) * | 2010-07-30 | 2012-03-28 | Gravity Res & Dev Kft | Recommender systems and methods |
| CN110209922B (en) * | 2018-06-12 | 2023-11-10 | 中国科学院自动化研究所 | Object recommendation method and device, storage medium and computer equipment |
| US10475106B1 (en) * | 2019-01-16 | 2019-11-12 | Capital One Services, Llc | Generating a product recommendation based on a user reaction |
| CN109996122B (en) * | 2019-04-12 | 2021-07-20 | 北京奇艺世纪科技有限公司 | Video recommendation method and device, server and storage medium |
| CN110149541B (en) * | 2019-04-23 | 2021-08-03 | 腾讯科技（深圳）有限公司 | Video recommendation method and device, computer equipment and storage medium |
| CN111046286B (en) * | 2019-12-12 | 2023-04-18 | 腾讯科技（深圳）有限公司 | Object recommendation method and device and computer storage medium |
| CN111274474A (en) * | 2020-01-16 | 2020-06-12 | 平安科技（深圳）有限公司 | Object recommendation method, electronic device and computer-readable storage medium |
| CN111310056B (en) * | 2020-03-11 | 2023-04-14 | 深圳市雅阅科技有限公司 | Information recommendation method, device, equipment and storage medium based on artificial intelligence |
| CN111709810A (en) * | 2020-06-17 | 2020-09-25 | 腾讯云计算（北京）有限责任公司 | Object recommendation method and device based on recommendation model |
-
2020
- 2020-12-01 CN CN202011388041.4A patent/CN112507216B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10244286B1 (en) * | 2018-01-30 | 2019-03-26 | Fmr Llc | Recommending digital content objects in a network environment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112507216A (en) | 2021-03-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| JP7052037B2 (en) | Fusion data processing method and information recommendation system | |
| CN110399728B (en) | Edge computing node trust evaluation method, device, equipment and storage medium | |
| CN107305611B (en) | Method and device for establishing model corresponding to malicious account and method and device for identifying malicious account | |
| CN110110139B (en) | Method and device for explaining recommendation result and electronic equipment | |
| CN112868004A (en) | Resource recommendation method and device, electronic equipment and storage medium | |
| WO2023116219A1 (en) | Cdn node distribution method and apparatus, electronic device, medium and program product | |
| CN112507216B (en) | Data object recommendation method, device, equipment and storage medium | |
| WO2024051707A1 (en) | Recommendation model training method and apparatus, and resource recommendation method and apparatus | |
| CN112437148B (en) | Service request processing method and device, service server and system | |
| CN110827078A (en) | Information recommendation method, device, equipment and storage medium | |
| CN111461825B (en) | Virtual resource generation method and device, electronic equipment and storage medium | |
| CN111814051B (en) | Resource type determining method and device | |
| CN111008667B (en) | Feature extraction method and device and electronic equipment | |
| CN115858911A (en) | Information recommendation method and device, electronic equipment and computer-readable storage medium | |
| CN106776947A (en) | Resource acquiring method, device and terminal | |
| CN112860999A (en) | Information recommendation method, device, equipment and storage medium | |
| CN113342859B (en) | Information processing method and device and electronic equipment | |
| CN116911954B (en) | Method and device for recommending items based on interests and popularity | |
| CN116911928B (en) | Training method and device of advertisement recommendation model based on creative features | |
| CN116800834B (en) | Virtual gift merging method, device, electronic equipment and computer readable medium | |
| CN117217852B (en) | Behavior recognition-based purchase willingness prediction method and device | |
| CN113609332B (en) | Video live broadcast resource recommendation method, system and device | |
| CN117150257A (en) | Sequence generation, model training and video recommendation methods, devices, equipment and media | |
| CN116089706A (en) | Training method and device for preset rearrangement model and content recommendation method and device | |
| CN117313821A (en) | Online learning method, video recommendation method and related devices |
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 |