CN111553759A - Product information pushing method, device, equipment and storage medium - Google Patents
Product information pushing method, device, equipment and storage medium Download PDFInfo
- Publication number
- CN111553759A CN111553759A CN202010220066.7A CN202010220066A CN111553759A CN 111553759 A CN111553759 A CN 111553759A CN 202010220066 A CN202010220066 A CN 202010220066A CN 111553759 A CN111553759 A CN 111553759A
- Authority
- CN
- China
- Prior art keywords
- product
- vector
- fusion
- feature vector
- numerical
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 62
- 239000013598 vector Substances 0.000 claims abstract description 281
- 230000004927 fusion Effects 0.000 claims abstract description 109
- 238000013528 artificial neural network Methods 0.000 claims abstract description 58
- 238000012545 processing Methods 0.000 claims abstract description 48
- 238000007781 pre-processing Methods 0.000 claims abstract description 13
- 238000012549 training Methods 0.000 claims description 33
- 238000004590 computer program Methods 0.000 claims description 17
- 230000007246 mechanism Effects 0.000 claims description 9
- 238000013507 mapping Methods 0.000 claims description 8
- 238000007405 data analysis Methods 0.000 abstract description 3
- 230000008569 process Effects 0.000 description 21
- 238000010606 normalization Methods 0.000 description 11
- 238000010586 diagram Methods 0.000 description 10
- 239000003795 chemical substances by application Substances 0.000 description 7
- 230000008520 organization Effects 0.000 description 7
- 230000006870 function Effects 0.000 description 6
- 230000004913 activation Effects 0.000 description 5
- 238000007477 logistic regression Methods 0.000 description 5
- 230000009977 dual effect Effects 0.000 description 4
- 230000000694 effects Effects 0.000 description 4
- 238000004364 calculation method Methods 0.000 description 3
- 238000012360 testing method Methods 0.000 description 3
- 238000012163 sequencing technique Methods 0.000 description 2
- 238000013459 approach Methods 0.000 description 1
- 238000003491 array Methods 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 230000008034 disappearance Effects 0.000 description 1
- 239000006185 dispersion Substances 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/06—Buying, selling or leasing transactions
- G06Q30/0601—Electronic shopping [e-shopping]
- G06Q30/0631—Item recommendations
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Finance (AREA)
- Accounting & Taxation (AREA)
- General Physics & Mathematics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Development Economics (AREA)
- Economics (AREA)
- Marketing (AREA)
- Strategic Management (AREA)
- General Business, Economics & Management (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The application relates to the field of data analysis, in particular to a method, a device, equipment and a storage medium for pushing product information, wherein the method comprises the following steps: acquiring user characteristic information from a target user terminal, wherein the user characteristic information comprises numerical type characteristics and classification characteristics; performing data preprocessing on the numerical features and the classification features to construct fusion vectors; processing the fusion vector based on a deep neural network submodel in a product recommendation model to obtain a deep feature vector, and processing the fusion vector based on a cross neural network submodel in the product recommendation model to obtain a cross feature vector; constructing a fusion feature vector according to the depth feature vector and the cross feature vector, and predicting respective recommendation scores of a plurality of products to be recommended according to the fusion feature vector; and sending the product to be recommended to the target terminal according to the recommendation score of the product to be recommended. More types of user attributes can be fused to participate in decision making so as to improve the accuracy of product recommendation.
Description
Technical Field
The present application relates to the field of data analysis, and in particular, to a method, an apparatus, a device, and a storage medium for pushing product information.
Background
The conventional product recommendation system determines the products which are purchased next by the user by analyzing the historical purchase condition and consultation condition of the products of the user by the service staff. For the mode of manually recommending products, a large amount of labor needs to be pumped, the labor cost is increased, and meanwhile poor experience is brought to users. Because of the large number of users, the manual recommendation mode cannot meet the requirements of the current users, and an intelligent product automatic recommendation scheme is generated.
In the existing intelligent automatic product recommendation scheme, the association degree between the user attribute and the product attribute is generally determined through big data analysis, and then the product which the user may be interested in is determined according to the user attribute, the product attribute and the association degree. However, the types of user attributes which can participate in decision making are limited, and more reference information can be omitted in the recommendation decision making process, so that the determined products are probably not really interested by the user, and the recommendation accuracy is low.
Therefore, how to improve the accuracy of product recommendation becomes an urgent problem to be solved.
Disclosure of Invention
The application provides a product information pushing method, a product information pushing device and a product information pushing storage medium, which can be used for fusing more types of user attributes to participate in decision making so as to improve the accuracy of product recommendation.
In a first aspect, the present application provides a product information pushing method, including:
acquiring user characteristic information from a target user terminal, wherein the user characteristic information comprises numerical type characteristics and classification characteristics;
performing data preprocessing on the numerical type features and the classification type features to construct fusion vectors;
processing the fusion vector based on a deep neural network submodel in a product recommendation model to obtain a deep feature vector, and processing the fusion vector based on a cross neural network submodel in the product recommendation model to obtain a cross feature vector;
constructing a fusion feature vector according to the depth feature vector and the cross feature vector, and predicting respective recommendation scores of a plurality of products to be recommended according to the fusion feature vector;
and sending the product information of at least one product to be recommended to the target user terminal according to the respective recommendation scores of the products to be recommended.
In a second aspect, the present application further provides a product information pushing device, where the device includes:
the data acquisition module is used for acquiring user characteristic information from a target user terminal, wherein the user characteristic information comprises numerical characteristics and classification characteristics;
the data processing module is used for carrying out data preprocessing on the numerical characteristic and the classification characteristic to construct a fusion vector;
the vector processing module is used for processing the fusion vector based on a deep neural network submodel in the product recommendation model to obtain a deep feature vector and processing the fusion vector based on a cross neural network submodel in the product recommendation model to obtain a cross feature vector;
the score prediction module is used for constructing a fusion feature vector according to the depth feature vector and the cross feature vector and predicting the respective recommendation scores of a plurality of products to be recommended according to the fusion feature vector;
and the product recommending module is used for sending the product information of at least one product to be recommended to the target user terminal according to the respective recommendation scores of the products to be recommended.
In a third aspect, the present application further provides a computer device comprising a memory and a processor; the memory is used for storing a computer program; the processor is configured to execute the computer program and implement the product information pushing method when executing the computer program.
In a fourth aspect, the present application further provides a computer-readable storage medium, which stores a computer program, and when the computer program is executed by a processor, the processor is caused to implement the product information pushing method as described above.
The application discloses a product information pushing method, a device, equipment and a storage medium, which are characterized in that user characteristic information is obtained from a target terminal, the user characteristic information comprises numerical type characteristics and classification characteristics, the numerical type characteristics and the classification characteristics are respectively processed to obtain a fusion vector, the fusion vector is respectively processed by a deep neural network sub-model and a cross neural network sub-model in a pre-trained product recommendation model to obtain a deep characteristic vector and a cross characteristic vector, the fusion characteristic vector is constructed according to the deep characteristic vector and the cross characteristic vector, so that the respective recommendation scores of a plurality of products to be recommended are predicted according to the fusion characteristic vector, finally, the product information of at least one product to be recommended is sent to the target user terminal according to the respective recommendation scores of the recommended products, and more types of user attributes can be fused to participate in decision making, to thereby improve the accuracy of product recommendations.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed to be used in the description of the embodiments are briefly introduced below, and it is obvious that the drawings in the following description are some embodiments of the present application, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.
Fig. 1 is a schematic flowchart of a product information pushing method provided in an embodiment of the present application;
FIG. 2 is a schematic view of a scenario of a product information pushing method applied to a server;
FIG. 3 is a sub-flow diagram of the data pre-processing of the numerical and categorical features to construct a fusion vector of FIG. 1;
FIG. 4 is a block diagram schematically illustrating a structure of a product recommendation model according to an embodiment of the present disclosure;
FIG. 5 is a schematic flow chart diagram of a product recommendation model training process provided by an embodiment of the present application;
fig. 6 is a schematic block diagram of a product information pushing apparatus provided in an embodiment of the present application;
fig. 7 is a schematic block diagram of another product information pushing device provided in the embodiment of the present application;
fig. 8 is a block diagram schematically illustrating a structure of a computer device according to an embodiment of the present application.
Detailed Description
The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are some, but not all, embodiments of the present application. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.
The flow diagrams depicted in the figures are merely illustrative and do not necessarily include all of the elements and operations/steps, nor do they necessarily have to be performed in the order depicted. For example, some operations/steps may be decomposed, combined or partially combined, so that the actual execution sequence may be changed according to the actual situation.
It is to be understood that the terminology used in the description of the present application herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. As used in the specification of the present application and the appended claims, the singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise.
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.
The embodiment of the application provides a product information pushing method and device, computer equipment and a storage medium. The product information pushing method can be applied to a server and used for pushing proper product information for the user according to the user characteristic information of the user and the like so as to improve the accuracy of product recommendation.
For example, a product information push method is used for the server; the servers may be independent servers or server clusters. The following embodiments will be described in detail with reference to a product information push method applied to a server.
Some embodiments of the present application will be described in detail below with reference to the accompanying drawings. The embodiments described below and the features of the embodiments can be combined with each other without conflict.
Referring to fig. 1, fig. 1 is a schematic flowchart of a product information pushing method according to an embodiment of the present disclosure.
Fig. 2 is a schematic view of a scenario of a product information pushing method applied to a server.
As shown in fig. 1 and fig. 2, the method for pushing product information specifically includes: step S101 to step S105.
Step S101, user characteristic information is obtained from a target user terminal, and the user characteristic information comprises numerical type characteristics and classification characteristics.
Illustratively, the user characteristic information may include vehicle information of the user. In some embodiments, the user may enter vehicle information on the terminal he uses, and the terminal then sends the vehicle information to the server.
Illustratively, when a user clicks a button for recommending a product on the terminal, a vehicle information input interface can be displayed on the terminal; the user can input the vehicle information at the input interface, and then the terminal transmits the vehicle information to the server.
In some embodiments, the vehicle information may include purchase price, seat number, steam discharge amount, organization code, broker code, agent code, carrier code, vehicle family, brand, belonging property, vehicle type, new energy vehicle logo, driving license owner gender, and the like.
Since the purchase price, the seat number, and the steam discharge amount are quantized values, the vehicle information such as the purchase price, the seat number, and the steam discharge amount can be determined as numerical characteristics. The discomfort-quantified characteristics, such as organization code, broker code, agent code, carrier code, vehicle family, brand, belonged property, vehicle type, new energy vehicle logo, and sex of driver of driving license, may be represented by numerical and/or alphabetic codes, and thus may be determined as the typing characteristics.
And S102, carrying out data preprocessing on the numerical characteristic and the classification characteristic to construct a fusion vector.
Since the numerical features and the classification features are different data types, different preprocessing methods are required to process the numerical features and the classification features respectively.
In some embodiments, referring to fig. 3, the pre-processing the numerical features and the classification features to construct a fusion vector includes: step S1021 to step S1023.
And S1021, carrying out standardization processing on the numerical characteristic to obtain numerical data corresponding to the numerical characteristic.
In some embodiments, the numerical features are normalized by a max-min normalization method.
max-min normalization, also known as dispersion normalization, is a linear transformation of the raw data, with the resulting values mapped between [0, 1 ]. The transfer function is as follows: x' ═ x-min ÷ (max-min). Wherein max represents a maximum value of the numerical characteristic, min represents a minimum value of the numerical characteristic, x represents a numerical value of the numerical characteristic in the vehicle information acquired from the target user terminal, and x' represents numerical data corresponding to the numerical characteristic.
For example, if the maximum value of the purchase price of the vehicle is 110 ten thousand renminbi and the minimum value is 10 ten thousand renminbi, and if the purchase price of the vehicle information acquired from the target user terminal is 30 ten thousand renminbi, the numerical data obtained by normalizing the purchase price is 0.2.
In other embodiments, the numerical characteristic is normalized such that the numerical data corresponding to the numerical characteristic satisfies a normal distribution.
Illustratively, the conversion function of the normalization process may be expressed as: where μ denotes a mean value of the numerical type feature sample data, σ denotes a standard deviation of the numerical type feature sample data, x denotes a numerical value of the numerical type feature in the vehicle information acquired from the target user terminal, and x' denotes numerical value data corresponding to the numerical type feature.
By carrying out standardization processing on the numerical characteristics, the problem that the accuracy of product recommendation is poor when the numerical characteristics are small due to overlarge numerical difference of different numerical characteristics can be solved; the numerical values of the numerical characteristics with large differences are compressed to a smaller range through standardization processing, so that the differences among the numerical values of different numerical characteristics can be kept, and the differences among the numerical values of different numerical characteristics are reduced, so that the product information pushing method can keep enough precision in more scenes.
Illustratively, the purchase price, the seat number, and the steam discharge amount in the vehicle information are standardized, and the corresponding obtained numerical data are D1, D2, and D3, respectively.
And step S1022, performing vectorization processing on the classification features to obtain classification vectors corresponding to the classification features.
In some embodiments, the server stores a mapping table of the classification type features and the classification type vectors in advance, and the classification type vectors corresponding to the classification type features can be obtained through table lookup.
In other embodiments, the type classification features are vectorized based on a vector embedding layer in a product recommendation model, so as to obtain a classification vector corresponding to the classification features. Through a product recommendation model trained by the model, the vector embedding layer learns the corresponding relation between the classification features and the classification vectors in the training sample, so that the corresponding classification vectors can be determined according to the classification features.
Illustratively, the organization code, the broker code, the agent code, the carrier code and the train department in the vehicle information are respectively vectorized to obtain classified vectors E1, E2, E3, E4 and E5.
And S1023, constructing a fusion vector according to the numerical data and the typing vector.
In some embodiments, the numerical data and the categorical vector are used as input vectors of a product recommendation model by end-to-end splicing. Illustratively, the fusion vectors are represented as D1, D2, D3, E1, E2, E3, E4, E5.
According to the fusion vector constructed by the numerical data and the classification vector, as the relevant information of the numerical characteristic and the classification characteristic of the user, such as purchase price, seat number, steam discharge amount, organization code, broker code, agent code, vehicle series, brand, belonging property, vehicle type, new energy vehicle mark, driving license vehicle owner gender and the like, is kept, more types of user attributes can be better fused to participate in decision making, so that more accurate user interest portrait can be obtained. Based on the user interest figures, products are recommended for the users, and the product recommendation accuracy can be improved. In some embodiments, before constructing the fusion vector according to the numerical data and the classification vector, the method further comprises the following steps:
and acquiring product category information from the target user terminal, and performing vectorization processing on the product category information to obtain a product category vector corresponding to the product category information.
Illustratively, the user also inputs the category of the product to be recommended at the target user terminal, so that the server can obtain product category information according to the user input, such as vehicle insurance risk types, which can be life insurance, vehicle insurance and the like.
In some embodiments, the vectorizing the product category information to obtain a product category vector corresponding to the product category information specifically includes the following steps: acquiring mapping relation data between preset product category information and product category vectors; and vectorizing the product category information according to the mapping relation data to obtain a product category vector corresponding to the product category information. The server stores a mapping table of the product category information and the product category vector in advance, and the product category vector corresponding to the product category information can be obtained through table lookup.
In other embodiments, the vectorizing the product category information to obtain a product category vector corresponding to the product category information specifically includes the following steps: and vectorizing the product category information based on a vector embedding layer in the product recommendation model to obtain a corresponding product category vector. Through the product recommendation model trained by the model, the vector embedding layer learns the corresponding relation between the product category information and the product category vector in the training sample, so that the corresponding product category vector can be determined according to the product category information.
In some embodiments, the constructing a fusion vector according to the numerical data and the classification vector specifically includes the following steps:
and constructing a fusion vector according to the numerical data, the classification vector and the product category vector.
Illustratively, the product category information obtained from the target user terminal is car insurance, the product category vector corresponding to the car insurance is F1, the numerical data, the classification vector and the product category vector are spliced end to construct a fusion vector, i.e. the fusion vector is used as an input vector of the product recommendation model, and is represented as D1, D2, D3, E1, E2, E3, E4, E5 and F1.
According to the numerical data, the classification vector and the product category vector, the constructed fusion vector retains the relevant information of the numerical characteristic and the classification characteristic of the user and the product category information, so that the product which the user may be interested in can be further analyzed according to the category characteristic corresponding to the product on the basis of the user interest portrait, and the accuracy of product recommendation is improved. In some embodiments, before constructing the fusion vector according to the numerical data and the classification vector, the method further comprises the following steps:
and acquiring a historical product purchase record corresponding to the target user terminal, and performing self-attention mechanism processing on the historical product purchase record to obtain a purchase record vector corresponding to the historical product purchase record.
Illustratively, according to the user information of the target user terminal, the product purchased by the user is obtained from the customer database, and the historical product purchase record corresponding to the user is generated.
If the target user is a renewal user and the target user has historically purchased more car insurance risk categories, there is a greater likelihood that the target user will continue to purchase the risk categories the next time the car insurance risk categories are purchased. In the renewal user car insurance risk recommendation, the importance of each risk can be evaluated according to the condition of historical product purchase records. Therefore, for the renewal user, the risk importance ranking is performed by adopting an attention mechanism according to the historical car risk purchase records. If the user has a large number of car risk types purchased historically, the risk types will receive a high score. Meanwhile, the processed result of the attention mechanism can be used as the input of the product recommendation model, so that the model can be predicted to achieve the optimal effect by using the attention mechanism.
In some embodiments, the constructing a fusion vector according to the numerical data and the classification vector specifically includes the following steps:
constructing a fusion vector according to the numerical data, the classification vector and the product category vector and/or the purchase record vector.
Illustratively, the product category information obtained from the target user terminal is car insurance, the product category vector corresponding to the car insurance is F1, the purchase record vector corresponding to the historical product purchase record is G1, the numerical data, the classification type vector, and the product category vector and/or the purchase record vector are spliced end to serve as an input vector of the product recommendation model, for example, a fusion vector may be represented as D1, D2, D3, E1, E2, E3, E4, E5, F1, G1, or as D1, D2, D3, E1, E2, E3, E4, E5, G1.
Step S103, processing the fusion vector based on a deep neural network submodel in the product recommendation model to obtain a deep feature vector, and processing the fusion vector based on a cross neural network submodel in the product recommendation model to obtain a cross feature vector.
In some embodiments, as shown in FIG. 4, the product recommendation model includes a deep neural network sub-model deepnetwork and a cross neural network sub-model cross network. Wherein the core idea of the crossbar network is to apply explicit feature interleaving in an efficient way. The crossbar network consists of crossbar layers, each crossbar layer having the following formula:
wherein, Xl、Xl+1Is a column vector representing the output from the cross-layer cross layer of the l-th and (l +1) -th layers, respectively, X0Is a predetermined vector, Wl、blFor the connection parameters between the two cross-layer cross layers of the l-th layer and the (l +1) -th layer, f () represents the feature cross referencing to fit the residual error between the output of this layer and the output of the previous layer, i.e., Xl+1-XlThe residual error of (a).
The capability of the model is limited due to the small number of parameters of the cross network. In order to capture the high-order nonlinear intersection, a deep network is introduced in parallel in the product recommendation model. The depth network is a fully connected feedforward neural network, and each depth layer has the following formula:
hl+1＝f(WlHl+bl)
wherein h isl、hl+1Output of hidden layer, W, of l-th and (l +1) -th layers, respectivelyl、blFor a connection parameter between two depth layers deep layer of l-th layer and (l +1) -th layer, f () represents an activation function-ReLUfunction.
The product recommendation model is obtained by training a cross network and a deep neural network deep network jointly (joint), so that the capability of the deep neural network deep network for capturing complex feature combinations is reserved, and meanwhile, each layer in the cross network has feature cross learning, so that cross features can be learned without manual feature engineering. Therefore, the product recommendation model has small memory, is very efficient in learning the combination features of a specific order, and introduces very little additional complexity.
And step S104, constructing a fusion feature vector according to the depth feature vector and the cross feature vector, and predicting the respective recommendation scores of a plurality of products to be recommended according to the fusion feature vector.
Examples of the inventionAnd the cross feature vector output by the cross neural network submodel in the product recommendation model is Xl1+The depth feature vector output by the deep neural network submodel is hl+1Splicing the outputs of deep neural network submodel deep and cross neural network submodels cross in the product recommendation model to obtain a fusion feature vector, which can be expressed as X for examplel+1，hl+1。
In some embodiments, the product recommendation model further comprises a logistic regression sub-model comprising a linear network layer and a sigmoid activation layer.
And inputting the fusion feature vector into the logistic regression sub-model, and processing by the linear network layer according to the fusion feature vector to obtain an output vector containing a plurality of elements. Illustratively, the elements correspond to the products to be recommended one by one, wherein the elements are in a direct proportion relationship with the recommendation scores of the products to be recommended, that is, the larger the element corresponding to a product to be recommended is, the higher the recommendation score corresponding to the product to be recommended is.
Illustratively, the output vector of the linear network layer is processed by a sigmoid activation function to calculate the recommendation score corresponding to each product to be recommended, so that the recommendation score can be processed into a numerical value between 0 and 1.
And S105, sending the product information of at least one product to be recommended to the target user terminal according to the respective recommendation scores of the plurality of products to be recommended.
For example, the product to be recommended with the largest recommended score may be determined as the target product, then the product information of the target product, such as the agreement, price, etc. of the car insurance, may be obtained, and then the product information may be sent to the target user terminal.
In some embodiments, the sending, according to the recommendation score of each of the plurality of products to be recommended, product information of at least one of the products to be recommended to the target user terminal includes the following steps:
and if the recommendation score of the product to be recommended is larger than a preset threshold value, sending the product to be recommended to the target user terminal.
Presetting a lowest recommendation score, such as 0.75, and if the recommendation score of the product to be recommended is greater than 0.75, sending the product information of the product to be recommended to the target user terminal; and if the recommendation score of no product to be recommended is larger than 0.75, sending the product information corresponding to the product to be recommended with the maximum recommendation score to the target user terminal.
In some embodiments, please refer to fig. 5, the product information pushing method further includes a training process of a product recommendation model, for example, may include step 201 to step S206.
It is understood that the training process of the product recommendation model can be implemented by a terminal or a server; the product recommendation model obtained after the training is finished may be deployed in a terminal or a server for implementing the product information push method of the foregoing embodiment.
Step 201, obtaining training sample data, where the training sample data includes product information of a plurality of users and interested products, and the product information includes numerical features and classification features.
The product information comprises vehicle information of a user, and the vehicle information comprises purchase price, seat number, steam discharge amount, organization code, broker code, agent code, vehicle manufacturer code, vehicle series, brand, belonging property, vehicle type, new energy vehicle mark, sex of a driving license owner and the like. Since the purchase price, the seat number, and the steam discharge amount are quantized values, the vehicle information such as the purchase price, the seat number, and the steam discharge amount can be determined as numerical characteristics. The discomfort-quantified characteristics, such as organization code, broker code, agent code, carrier code, vehicle family, brand, belonged property, vehicle type, new energy vehicle logo, and sex of driver of driving license, may be represented by numerical and/or alphabetic codes, and thus may be determined as the typing characteristics.
Step 202, performing standardization processing on the numerical characteristics to obtain numerical data corresponding to the numerical characteristics, and performing vectorization processing on the classification characteristics to obtain a classification vector corresponding to the classification characteristics.
In some embodiments, the numerical features are normalized by a max-min normalization method. The reason is that if the data difference between a plurality of features is large, the convergence speed is slow, and in a model related to distance calculation, if the numerical difference between a plurality of features is large, the influence of the small-numerical features on the distance is small, and the accuracy of the model is influenced.
And performing random initialization processing on the classification features to obtain corresponding classification vectors, and updating the classification vectors in the model training process. Illustratively, the classification features are processed by applying L2 regularization, so that the complexity and instability degree of the model are reduced in the model training learning process, and the risk of overfitting is avoided.
Illustratively, a corresponding vehicle risk category feature vector is initialized for each vehicle risk category, the classification variables are adopted to express the vehicle risk category information, and the initialization dimension of the features of the classification variables is 10 dimensions. Meanwhile, in the model training process, the classification characteristic vector value is updated, and the initialization characteristic value is between 0 and 0.001. When the eigenvalue is between 0 and 0.001, the model training converges fastest.
Illustratively, the purchase price, the seat number and the steam discharge amount in the vehicle information of a certain user in the training sample data are standardized, the obtained numerical data are respectively D1, D2 and D3, and the organization code, the broker code, the agent code, the vehicle quotient code and the vehicle system in the vehicle information of the user are respectively vectorized to obtain corresponding classification vectors E1, E2, E3, E4 and E5.
And step 203, constructing a fusion vector according to the numerical data and the classified vector of the user.
In some embodiments, the numerical data and the categorical vector are used as input vectors of a product recommendation model by end-to-end splicing. Illustratively, the fusion vectors are represented as D1, D2, D3, E1, E2, E3, E4, E5.
And 204, respectively inputting the fusion vector of the user into a deep neural network submodel and a cross neural network submodel in the product recommendation model to respectively obtain a deep feature vector and a cross feature vector.
Exemplarily, after the fusion vector of the user is input into a cross neural network submodel in a product recommendation model, the output cross feature vector is Xl+1(ii) a Inputting the fusion vector of the user into a deep neural network submodel in a product recommendation model, and outputting a depth feature vector hl+1。
Step 205, constructing a fusion feature vector according to the depth feature vector and the cross feature vector of the user, and predicting respective recommendation scores of a plurality of products to be recommended according to the fusion feature vector.
Illustratively, the cross feature vector output by the cross neural network submodel in the product recommendation model is Xl+lThe depth feature vector output by the deep neural network submodel is hl+1Splicing the outputs of deep neural network submodel deep and cross neural network submodels cross in the product recommendation model to obtain a fusion feature vector, which can be expressed as X for examplel+1，hl+1。
In some embodiments, the product recommendation model further comprises a logistic regression sub-model, wherein the logistic regression sub-model comprises a linear network layer and a sigmoid activation layer. And inputting the fusion feature vector into the logistic regression sub-model, and processing by the linear network layer according to the fusion feature vector to obtain an output vector containing a plurality of elements. And calculating the recommendation score corresponding to each product to be recommended by the output vector of the linear network layer through a sigmoid activation function, so that the recommendation score can be processed into a numerical value between 0 and 1.
And step 206, calculating a loss value according to the respective recommendation scores of the products to be recommended corresponding to the user and the interested products of the user, and adjusting the model parameters of the product recommendation model according to the loss value.
In some embodiments, the acquired dataset is divided into a training dataset, a verification dataset, and a test dataset. The training data set is used for model training, the checking data set is used for model selection, and the testing data set is used for model testing.
In some embodiments, the trained question may be treated as a non-dual question or a dual question, the non-dual question being in constructed positive and negative sample pairs, if the user purchases some sort of risk, which is a positive sample, we label that sort of risk as 1, and if the user does not purchase some sort of risk, which is a negative sample, we label that sort of risk as 0. In the dual problem, each sample pair corresponds to a positive sample and a negative sample and serves as a sequencing problem, and in the training process of the model, the score of the positive sample is higher than that of the negative sample so as to achieve the sequencing effect of the model.
In some embodiments, in the model training process, first, through calculation by the deep neural network, the number of layers of the deep neural network may be adjusted as needed, for example, a linear neural network with two layers is selected, each layer of the neural network corresponds to a weight and a Bias, and in the model training, the values of the weight and the Bias are updated. The hidden layer dimension size can be adjusted as needed, for example, the hidden layer dimension is set to 50. The final input dimension of the model is 100 dimensions, and in order to achieve the ideal effect of the model, the middle layer output layer dimension is set to be 50, and the input layer dimension is set to be 10, wherein the output layer dimension can be adjusted according to the requirement. Through the deep neural network, the model can learn feature information in a deep sense.
In some embodiments, in the deep neural network, a Batch Normalization mechanism may be selectively used, and in the training process of the neural network, as the depth is increased, the distribution of input values may shift and approach to the upper and lower ends of the value range. In order to solve the problem, Batch Normalization forcibly pulls back the distribution of input values of each layer of neural network to a standard normal distribution with a mean value of 0 and a variance of 1 through a certain Normalization means, so that the distribution returns to a region where a nonlinear function is sensitive to input, and a loss function can be greatly changed, such as a gradient is increased, and the problem of gradient disappearance is avoided. Meanwhile, the gradient is increased, so that the convergence speed of the model can be increased, and the training speed is increased.
In some embodiments, in the model training process, in order to learn cross information between features, a cross neural network mechanism is used, the number of layers of the cross neural network can be adjusted as required, and the implementation mechanism of the cross neural network is to multiply a matrix with the same dimension according to the result of the last layer of the cross neural network each time, and add the output result of the last layer of the cross neural network and a corresponding offset.
Illustratively, an optimizer is selected for model training, wherein the optimizer can be selected from, for example, sgd optimizer, and the learning rate of the optimizer can be adjusted according to actual needs so as to improve the model training effect. sgd optimizer has the characteristics of fast training speed and difficult falling into local optimal solution.
Illustratively, a dynamic learning rate scheme may be employed in the product recommendation model, such as a 10-fold decay in learning rate per 100 training iterations of the model. The purpose of setting the dynamic learning rate is to make the model more closely fit into the optimal solution.
The product information pushing method provided in the above embodiment obtains user feature information from a target terminal, where the user feature information includes numerical features and classification features, respectively processes the numerical features and the classification features to obtain a fusion vector, respectively processes the fusion vector by using a deep neural network sub-model and a cross neural network sub-model in a pre-trained product recommendation model to obtain a deep feature vector and a cross feature vector, constructs the fusion feature vector according to the deep feature vector and the cross feature vector, so as to predict respective recommendation scores of a plurality of products to be recommended according to the fusion feature vector, and finally sends product information of at least one product to be recommended to the target user terminal according to respective recommendation scores of the recommended products, so that more types of user attributes can be fused to participate in decision making, thereby improving the accuracy of product recommendation.
Referring to fig. 6, fig. 6 is a schematic block diagram of a product information pushing device according to an embodiment of the present application, where the product information pushing device is configured to execute the product information pushing method. The product information pushing device can be configured in a server or a terminal.
The server may be an independent server or a server cluster. The terminal can be an electronic device such as a mobile phone, a tablet computer, a notebook computer, a desktop computer, a personal digital assistant and a wearable device.
As shown in fig. 6, the product information push apparatus 300 includes: a data acquisition module 301, a data processing module 302, a vector processing module 303, a score prediction module 304, and a product recommendation module 305.
A data obtaining module 301, configured to obtain user feature information from a target user terminal, where the user feature information includes a numeric feature and a typing feature.
A data processing module 302, configured to perform data preprocessing on the numerical features and the classification features to construct a fusion vector.
As shown in fig. 7, the data processing module 302 includes a normalization processing sub-module 3021, a vectorization processing sub-module 3022, and a data fusion sub-module 3023.
Specifically, the normalization processing submodule 3021 is configured to perform normalization processing on the numerical characteristic to obtain numerical data corresponding to the numerical characteristic; a vectorization processing submodule 3022, configured to perform vectorization processing on the classification features to obtain classification vectors corresponding to the classification features; a data fusion sub-module 3023, configured to construct a fusion vector according to the numerical data and the typing vector.
And the vector processing module 303 is configured to process the fusion vector based on a deep neural network submodel in the product recommendation model to obtain a deep feature vector, and process the fusion vector based on a cross neural network submodel in the product recommendation model to obtain a cross feature vector.
And the score prediction module 304 is configured to construct a fusion feature vector according to the depth feature vector and the cross feature vector, and predict respective recommendation scores of a plurality of products to be recommended according to the fusion feature vector.
The product recommending module 305 is configured to send product information of at least one product to be recommended to the target user terminal according to the respective recommendation scores of the plurality of products to be recommended.
As shown in fig. 7, the product recommending module 305 includes a product determining sub-module 3051 and a product pushing sub-module 3052.
Specifically, the product determination submodule 3051 is configured to determine whether a recommendation score of a product to be recommended is greater than a preset threshold; and the product pushing submodule 3052 is configured to send the product to be recommended, of which the recommendation score is greater than a preset threshold, to the target user terminal.
The product information pushing device 300 comprises a model optimization module, which is configured to: acquiring training sample data, wherein the training sample data comprises product information of a plurality of users and interested products, and the product information comprises numerical characteristics and classification characteristics; standardizing the numerical characteristics to obtain numerical data corresponding to the numerical characteristics, and vectorizing the classification characteristics to obtain classification vectors corresponding to the classification characteristics; constructing a fusion vector according to the numerical data and the classified vector of the user; inputting the fusion vector of the user into a deep neural network submodel and a cross neural network submodel in the product recommendation model respectively to obtain a deep feature vector and a cross feature vector respectively; constructing a fusion feature vector according to the depth feature vector and the cross feature vector of the user, and predicting respective recommendation scores of a plurality of products to be recommended according to the fusion feature vector; calculating a loss value according to the respective recommendation scores of the products to be recommended corresponding to the user and the interested products of the user, and adjusting the model parameters of the product recommendation model according to the loss value.
It should be clearly understood by those skilled in the art that, for convenience and brevity of description, the specific working processes of the product information pushing apparatus and each module described above may refer to corresponding processes in the foregoing embodiment of the product information pushing method, and are not described herein again.
The product information pushing apparatus may be implemented in the form of a computer program, which can run on a computer device as shown in fig. 8.
Referring to fig. 8, fig. 8 is a schematic block diagram of a computer device according to an embodiment of the present disclosure. The computer device may be a server or a terminal.
Referring to fig. 8, the computer device includes a processor, a memory, and a network interface connected through a system bus, wherein the memory may include a nonvolatile storage medium and an internal memory.
The non-volatile storage medium may store an operating system and a computer program. The computer program comprises program instructions which, when executed, can cause a processor to execute any one of the product information pushing methods.
The processor is used for providing calculation and control capability and supporting the operation of the whole computer equipment.
The internal memory provides an environment for running a computer program in the non-volatile storage medium, and the computer program, when executed by the processor, can cause the processor to execute any one of the product information pushing methods.
The network interface is used for network communication, such as sending assigned tasks and the like. Those skilled in the art will appreciate that the architecture shown in fig. 8 is merely a block diagram of some of the structures associated with the disclosed aspects and is not intended to limit the computing devices to which the disclosed aspects apply, as particular computing devices may include more or less components than those shown, or may combine certain components, or have a different arrangement of components.
It should be understood that the Processor may be a Central Processing Unit (CPU), and the Processor may be other general purpose processors, Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), Field Programmable Gate Arrays (FPGAs) or other Programmable logic devices, discrete Gate or transistor logic devices, discrete hardware components, etc. Wherein a general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.
Wherein, in one embodiment, the processor is configured to execute a computer program stored in the memory to implement the steps of:
acquiring user characteristic information from a target user terminal, wherein the user characteristic information comprises numerical type characteristics and classification characteristics;
performing data preprocessing on the numerical type features and the classification type features to construct fusion vectors;
processing the fusion vector based on a deep neural network submodel in a product recommendation model to obtain a deep feature vector, and processing the fusion vector based on a cross neural network submodel in the product recommendation model to obtain a cross feature vector;
constructing a fusion feature vector according to the depth feature vector and the cross feature vector, and predicting respective recommendation scores of a plurality of products to be recommended according to the fusion feature vector;
and sending the product information of at least one product to be recommended to the target user terminal according to the respective recommendation scores of the products to be recommended.
In one embodiment, the processor, in performing the data pre-processing on the numerical features and the categorical features to construct a fusion vector, is configured to perform:
carrying out standardization processing on the numerical characteristic to obtain numerical data corresponding to the numerical characteristic;
vectorizing the classification features to obtain classification vectors corresponding to the classification features;
and constructing a fusion vector according to the numerical data and the typing vector.
In one embodiment, the processor, prior to implementing the constructing a fused vector from the numerical data and the subtype vector, is further configured to implement:
acquiring product category information from the target user terminal, and performing vectorization processing on the product category information to obtain a product category vector corresponding to the product category information;
the constructing a fusion vector according to the numerical data and the typing vector comprises:
and constructing a fusion vector according to the numerical data, the classification vector and the product category vector.
In an embodiment, when the processor implements vectorization processing on the product category information to obtain a product category vector corresponding to the product category information, the processor is configured to implement:
acquiring mapping relation data between preset product category information and product category vectors;
and vectorizing the product category information according to the mapping relation data to obtain a product category vector corresponding to the product category information.
In one embodiment, the processor, prior to implementing the constructing a fused vector from the numerical data and the typing vector, is configured to implement:
acquiring a historical product purchase record corresponding to the target user terminal, and performing self-attention mechanism processing on the historical product purchase record to obtain a purchase record vector corresponding to the historical product purchase record;
the constructing a fusion vector according to the numerical data and the typing vector comprises:
constructing a fusion vector according to the numerical data, the classification vector and the product category vector and/or the purchase record vector.
In one embodiment, when the processor sends the product information of at least one product to be recommended to the target user terminal according to the recommendation scores of the products to be recommended, the processor is configured to:
and if the recommendation score of the product to be recommended is larger than a preset threshold value, sending the product to be recommended to the target user terminal.
In one embodiment, when the processor implements the product information pushing method, the processor is further configured to implement:
acquiring training sample data, wherein the training sample data comprises product information of a plurality of users and interested products, and the product information comprises numerical characteristics and classification characteristics;
standardizing the numerical characteristics to obtain numerical data corresponding to the numerical characteristics, and vectorizing the classification characteristics to obtain classification vectors corresponding to the classification characteristics;
constructing a fusion vector according to the numerical data and the classified vector of the user;
inputting the fusion vector of the user into a deep neural network submodel and a cross neural network submodel in the product recommendation model respectively to obtain a deep feature vector and a cross feature vector respectively;
constructing a fusion feature vector according to the depth feature vector and the cross feature vector of the user, and predicting respective recommendation scores of a plurality of products to be recommended according to the fusion feature vector;
calculating a loss value according to the respective recommendation scores of the products to be recommended corresponding to the user and the interested products of the user, and adjusting the model parameters of the product recommendation model according to the loss value.
The embodiment of the application further provides a computer-readable storage medium, wherein a computer program is stored in the computer-readable storage medium, the computer program comprises program instructions, and the processor executes the program instructions to realize any product information pushing method provided by the embodiment of the application.
The computer-readable storage medium may be an internal storage unit of the computer device described in the foregoing embodiment, for example, a hard disk or a memory of the computer device. The computer readable storage medium 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.
The product information pushing device, the storage medium, and the computer device provided in the foregoing embodiment obtain user feature information from a target terminal, where the user feature information includes numerical features and classification features, respectively process the numerical features and the classification features to obtain a fusion vector, respectively process the fusion vector by using a deep neural network sub-model and a cross neural network sub-model in a pre-trained product recommendation model to obtain a deep feature vector and a cross feature vector, construct the fusion feature vector according to the deep feature vector and the cross feature vector, predict respective recommendation scores of a plurality of products to be recommended according to the fusion feature vector, and finally send product information of at least one product to be recommended to the target user terminal according to respective recommendation scores of the recommended products, so as to fuse more types of user attributes to participate in a decision, thereby improving the accuracy of product recommendation.
While the invention has been described with reference to specific embodiments, the scope of the invention is not limited thereto, and those skilled in the art can easily conceive various equivalent modifications or substitutions within the technical scope of the invention. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.
Claims (10)
1. A product information pushing method is characterized by comprising the following steps:
acquiring user characteristic information from a target user terminal, wherein the user characteristic information comprises numerical type characteristics and classification characteristics;
performing data preprocessing on the numerical type features and the classification type features to construct fusion vectors;
processing the fusion vector based on a deep neural network submodel in a product recommendation model to obtain a deep feature vector, and processing the fusion vector based on a cross neural network submodel in the product recommendation model to obtain a cross feature vector;
constructing a fusion feature vector according to the depth feature vector and the cross feature vector, and predicting respective recommendation scores of a plurality of products to be recommended according to the fusion feature vector;
and sending the product information of at least one product to be recommended to the target user terminal according to the respective recommendation scores of the products to be recommended.
2. The product information pushing method according to claim 1, wherein the data preprocessing the numerical features and the classification-type features to construct a fusion vector comprises:
carrying out standardization processing on the numerical characteristic to obtain numerical data corresponding to the numerical characteristic;
vectorizing the classification features to obtain classification vectors corresponding to the classification features;
and constructing a fusion vector according to the numerical data and the typing vector.
3. The product information pushing method according to claim 2, wherein before constructing a fusion vector from the numerical data and the typing vector, further comprising:
acquiring product category information from the target user terminal, and performing vectorization processing on the product category information to obtain a product category vector corresponding to the product category information;
the constructing a fusion vector according to the numerical data and the typing vector comprises:
and constructing a fusion vector according to the numerical data, the classification vector and the product category vector.
4. The product information pushing method according to claim 3, wherein the vectorizing the product category information to obtain a product category vector corresponding to the product category information includes:
acquiring mapping relation data between preset product category information and product category vectors;
and vectorizing the product category information according to the mapping relation data to obtain a product category vector corresponding to the product category information.
5. The product information pushing method according to any one of claims 1 to 4, wherein before constructing a fusion vector from the numerical data and the typing vector, further comprising:
acquiring a historical product purchase record corresponding to the target user terminal, and performing self-attention mechanism processing on the historical product purchase record to obtain a purchase record vector corresponding to the historical product purchase record;
the constructing a fusion vector according to the numerical data and the typing vector comprises:
constructing a fusion vector according to the numerical data, the classification vector and the product category vector and/or the purchase record vector.
6. The method according to claim 1, wherein the sending the product information of at least one of the products to be recommended to the target user terminal according to the recommendation score of each of the plurality of products to be recommended comprises:
and if the recommendation score of the product to be recommended is larger than a preset threshold value, sending the product to be recommended to the target user terminal.
7. The product information pushing method according to claim 1, further comprising:
acquiring training sample data, wherein the training sample data comprises product information of a plurality of users and interested products, and the product information comprises numerical characteristics and classification characteristics;
standardizing the numerical characteristics to obtain numerical data corresponding to the numerical characteristics, and vectorizing the classification characteristics to obtain classification vectors corresponding to the classification characteristics;
constructing a fusion vector according to the numerical data and the classified vector of the user;
inputting the fusion vector of the user into a deep neural network submodel and a cross neural network submodel in the product recommendation model respectively to obtain a deep feature vector and a cross feature vector respectively;
constructing a fusion feature vector according to the depth feature vector and the cross feature vector of the user, and predicting respective recommendation scores of a plurality of products to be recommended according to the fusion feature vector;
calculating a loss value according to the respective recommendation scores of the products to be recommended corresponding to the user and the interested products of the user, and adjusting the model parameters of the product recommendation model according to the loss value.
8. A product information pushing apparatus, comprising:
the data acquisition module is used for acquiring user characteristic information from a target user terminal, wherein the user characteristic information comprises numerical characteristics and classification characteristics;
the data processing module is used for carrying out data preprocessing on the numerical characteristic and the classification characteristic to construct a fusion vector;
the vector processing module is used for processing the fusion vector based on a deep neural network submodel in the product recommendation model to obtain a deep feature vector and processing the fusion vector based on a cross neural network submodel in the product recommendation model to obtain a cross feature vector;
the score prediction module is used for constructing a fusion feature vector according to the depth feature vector and the cross feature vector and predicting the respective recommendation scores of a plurality of products to be recommended according to the fusion feature vector;
and the product recommending module is used for sending the product information of at least one product to be recommended to the target user terminal according to the respective recommendation scores of the products to be recommended.
9. A computer device, wherein the computer device comprises a memory and a processor;
the memory is used for storing a computer program;
the processor, configured to execute the computer program and implement the product information pushing method according to any one of claims 1 to 7 when executing the computer program.
10. A computer-readable storage medium, characterized in that the computer-readable storage medium stores a computer program which, when executed by a processor, causes the processor to implement the product information push method according to any one of claims 1 to 7.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010220066.7A CN111553759A (en) | 2020-03-25 | 2020-03-25 | Product information pushing method, device, equipment and storage medium |
| PCT/CN2020/136326 WO2021189976A1 (en) | 2020-03-25 | 2020-12-15 | Product information pushing method and apparatus, device, and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010220066.7A CN111553759A (en) | 2020-03-25 | 2020-03-25 | Product information pushing method, device, equipment and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111553759A true CN111553759A (en) | 2020-08-18 |
Family
ID=72003754
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010220066.7A Pending CN111553759A (en) | 2020-03-25 | 2020-03-25 | Product information pushing method, device, equipment and storage medium |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN111553759A (en) |
| WO (1) | WO2021189976A1 (en) |
Cited By (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112163165A (en) * | 2020-10-21 | 2021-01-01 | 腾讯科技（深圳）有限公司 | Information recommendation method, device, equipment and computer readable storage medium |
| CN112163963A (en) * | 2020-09-27 | 2021-01-01 | 中国平安财产保险股份有限公司 | Service recommendation method and device, computer equipment and storage medium |
| CN112231550A (en) * | 2020-09-11 | 2021-01-15 | 重庆誉存大数据科技有限公司 | Credit financial product recommendation processing method and device |
| CN112307341A (en) * | 2020-11-02 | 2021-02-02 | 沈阳民航东北凯亚有限公司 | Flight pushing method and device |
| CN112417760A (en) * | 2020-11-20 | 2021-02-26 | 哈尔滨工程大学 | Warship control method based on competitive hybrid network |
| CN112949824A (en) * | 2021-02-07 | 2021-06-11 | 北京淇瑀信息科技有限公司 | Neural network-based multi-output multi-task feature evaluation method and device and electronic equipment |
| CN112990584A (en) * | 2021-03-19 | 2021-06-18 | 山东大学 | Automatic production decision system and method based on deep reinforcement learning |
| CN113159201A (en) * | 2021-04-28 | 2021-07-23 | 上海德衡数据科技有限公司 | Method and device for computing intelligent fusion information based on cloud edge and cloud server |
| CN113298661A (en) * | 2021-07-28 | 2021-08-24 | 北京芯盾时代科技有限公司 | Artificial intelligence based product recommendation method and device and electronic equipment |
| CN113298121A (en) * | 2021-04-30 | 2021-08-24 | 上海淇玥信息技术有限公司 | Message sending method and device based on multi-data source modeling and electronic equipment |
| WO2021189976A1 (en) * | 2020-03-25 | 2021-09-30 | 平安科技（深圳）有限公司 | Product information pushing method and apparatus, device, and storage medium |
| CN113610580A (en) * | 2021-08-10 | 2021-11-05 | 平安科技（深圳）有限公司 | Product recommendation method and device, electronic equipment and readable storage medium |
| CN113672803A (en) * | 2021-08-02 | 2021-11-19 | 杭州网易云音乐科技有限公司 | Recommendation method and device, computing equipment and storage medium |
| CN113781139A (en) * | 2020-10-19 | 2021-12-10 | 北京沃东天骏信息技术有限公司 | Item recommendation method, item recommendation device, equipment and medium |
| CN113868542A (en) * | 2021-11-25 | 2021-12-31 | 平安科技（深圳）有限公司 | Attention model-based push data acquisition method, device, equipment and medium |
| CN115034847A (en) * | 2022-05-25 | 2022-09-09 | 山东大学 | Product recommendation method, system, storage medium and equipment based on deep learning |
| WO2023051678A1 (en) * | 2021-09-29 | 2023-04-06 | 华为技术有限公司 | Recommendation method and related device |
Families Citing this family (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114520826B (en) * | 2021-12-31 | 2024-04-05 | 珠海华发金融科技研究院有限公司 | Cross-platform information matching method and device and cloud intelligent robot |
| CN114581196A (en) * | 2022-03-10 | 2022-06-03 | 广州华多网络科技有限公司 | Commodity sorting processing method and device, equipment, medium and product thereof |
| CN114756762B (en) * | 2022-06-13 | 2022-09-02 | 腾讯科技（深圳）有限公司 | Data processing method, device, equipment and storage medium |
| CN116662556B (en) * | 2023-08-02 | 2023-10-20 | 天河超级计算淮海分中心 | Text data processing method integrating user attributes |
| CN117609620A (en) * | 2023-11-29 | 2024-02-27 | 广州方舟信息科技有限公司 | Medical commodity information pushing method and device, electronic equipment and storage medium |
| CN117952724B (en) * | 2024-03-21 | 2024-06-14 | 烟台大学 | Article recommendation method, system and equipment based on knowledge graph and neural network |
| CN118035859B (en) * | 2024-04-15 | 2024-06-11 | 普益智慧云科技(成都)有限公司 | User portrait construction method, system, computer equipment and storage medium |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110263821A (en) * | 2019-05-29 | 2019-09-20 | 阿里巴巴集团控股有限公司 | Transaction feature generates the generation method and device of the training of model, transaction feature |
| CN110458649A (en) * | 2019-07-11 | 2019-11-15 | 北京三快在线科技有限公司 | Information recommendation method, device, electronic equipment and readable storage medium storing program for executing |
| CN110888911A (en) * | 2019-10-11 | 2020-03-17 | 平安科技（深圳）有限公司 | Sample data processing method and device, computer equipment and storage medium |
Family Cites Families (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101140651A (en) * | 2006-09-04 | 2008-03-12 | 腾讯科技（深圳）有限公司 | Display objects ordering method and system |
| CN101551825A (en) * | 2009-05-15 | 2009-10-07 | 中国科学技术大学 | Personalized film recommendation system and method based on attribute description |
| TWI612488B (en) * | 2016-12-05 | 2018-01-21 | 財團法人資訊工業策進會 | Computer device and method for predicting market demand of commodities |
| CN108985830B (en) * | 2018-07-05 | 2021-03-02 | 北京邮电大学 | Recommendation scoring method and device based on heterogeneous information network |
| CN109710768B (en) * | 2019-01-10 | 2020-07-28 | 西安交通大学 | Tax payer industry two-level classification method based on MIMO recurrent neural network |
| CN109977151B (en) * | 2019-03-28 | 2020-02-07 | 北京九章云极科技有限公司 | Data analysis method and system |
| CN110119477B (en) * | 2019-05-14 | 2024-02-27 | 腾讯科技（深圳）有限公司 | Information pushing method, device and storage medium |
| CN111553759A (en) * | 2020-03-25 | 2020-08-18 | 平安科技（深圳）有限公司 | Product information pushing method, device, equipment and storage medium |
-
2020
- 2020-03-25 CN CN202010220066.7A patent/CN111553759A/en active Pending
- 2020-12-15 WO PCT/CN2020/136326 patent/WO2021189976A1/en active Application Filing
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110263821A (en) * | 2019-05-29 | 2019-09-20 | 阿里巴巴集团控股有限公司 | Transaction feature generates the generation method and device of the training of model, transaction feature |
| CN110458649A (en) * | 2019-07-11 | 2019-11-15 | 北京三快在线科技有限公司 | Information recommendation method, device, electronic equipment and readable storage medium storing program for executing |
| CN110888911A (en) * | 2019-10-11 | 2020-03-17 | 平安科技（深圳）有限公司 | Sample data processing method and device, computer equipment and storage medium |
Cited By (24)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2021189976A1 (en) * | 2020-03-25 | 2021-09-30 | 平安科技（深圳）有限公司 | Product information pushing method and apparatus, device, and storage medium |
| CN112231550A (en) * | 2020-09-11 | 2021-01-15 | 重庆誉存大数据科技有限公司 | Credit financial product recommendation processing method and device |
| CN112231550B (en) * | 2020-09-11 | 2024-05-17 | 重庆誉存大数据科技有限公司 | Credit financial product recommendation processing method and device |
| CN112163963A (en) * | 2020-09-27 | 2021-01-01 | 中国平安财产保险股份有限公司 | Service recommendation method and device, computer equipment and storage medium |
| CN112163963B (en) * | 2020-09-27 | 2024-04-02 | 中国平安财产保险股份有限公司 | Service recommendation method, device, computer equipment and storage medium |
| CN113781139A (en) * | 2020-10-19 | 2021-12-10 | 北京沃东天骏信息技术有限公司 | Item recommendation method, item recommendation device, equipment and medium |
| CN112163165A (en) * | 2020-10-21 | 2021-01-01 | 腾讯科技（深圳）有限公司 | Information recommendation method, device, equipment and computer readable storage medium |
| CN112163165B (en) * | 2020-10-21 | 2024-05-17 | 腾讯科技（深圳）有限公司 | Information recommendation method, device, equipment and computer readable storage medium |
| CN112307341A (en) * | 2020-11-02 | 2021-02-02 | 沈阳民航东北凯亚有限公司 | Flight pushing method and device |
| CN112417760A (en) * | 2020-11-20 | 2021-02-26 | 哈尔滨工程大学 | Warship control method based on competitive hybrid network |
| CN112417760B (en) * | 2020-11-20 | 2023-01-17 | 哈尔滨工程大学 | Warship control method based on competitive hybrid network |
| CN112949824A (en) * | 2021-02-07 | 2021-06-11 | 北京淇瑀信息科技有限公司 | Neural network-based multi-output multi-task feature evaluation method and device and electronic equipment |
| CN112990584A (en) * | 2021-03-19 | 2021-06-18 | 山东大学 | Automatic production decision system and method based on deep reinforcement learning |
| CN113159201A (en) * | 2021-04-28 | 2021-07-23 | 上海德衡数据科技有限公司 | Method and device for computing intelligent fusion information based on cloud edge and cloud server |
| CN113298121A (en) * | 2021-04-30 | 2021-08-24 | 上海淇玥信息技术有限公司 | Message sending method and device based on multi-data source modeling and electronic equipment |
| CN113298121B (en) * | 2021-04-30 | 2023-08-18 | 上海淇玥信息技术有限公司 | Message sending method and device based on multi-data source modeling and electronic equipment |
| CN113298661A (en) * | 2021-07-28 | 2021-08-24 | 北京芯盾时代科技有限公司 | Artificial intelligence based product recommendation method and device and electronic equipment |
| CN113672803A (en) * | 2021-08-02 | 2021-11-19 | 杭州网易云音乐科技有限公司 | Recommendation method and device, computing equipment and storage medium |
| CN113610580B (en) * | 2021-08-10 | 2023-09-19 | 平安科技（深圳）有限公司 | Product recommendation method and device, electronic equipment and readable storage medium |
| CN113610580A (en) * | 2021-08-10 | 2021-11-05 | 平安科技（深圳）有限公司 | Product recommendation method and device, electronic equipment and readable storage medium |
| WO2023051678A1 (en) * | 2021-09-29 | 2023-04-06 | 华为技术有限公司 | Recommendation method and related device |
| CN113868542B (en) * | 2021-11-25 | 2022-03-11 | 平安科技（深圳）有限公司 | Attention model-based push data acquisition method, device, equipment and medium |
| CN113868542A (en) * | 2021-11-25 | 2021-12-31 | 平安科技（深圳）有限公司 | Attention model-based push data acquisition method, device, equipment and medium |
| CN115034847A (en) * | 2022-05-25 | 2022-09-09 | 山东大学 | Product recommendation method, system, storage medium and equipment based on deep learning |
Also Published As
| Publication number | Publication date |
|---|---|
| WO2021189976A1 (en) | 2021-09-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111553759A (en) | Product information pushing method, device, equipment and storage medium | |
| CN112559784B (en) | Image classification method and system based on incremental learning | |
| CN111639516B (en) | Analysis platform based on machine learning | |
| CN109800885A (en) | It is determined for the rule of black box machine learning model | |
| EP4322031A1 (en) | Recommendation method, recommendation model training method, and related product | |
| CN114022176A (en) | Method for predicting commodity sales on e-commerce platform and electronic equipment | |
| CN112085565A (en) | Deep learning-based information recommendation method, device, equipment and storage medium | |
| CN112861936A (en) | Graph node classification method and device based on graph neural network knowledge distillation | |
| CN112529071B (en) | Text classification method, system, computer equipment and storage medium | |
| CN111582932A (en) | Inter-scene information pushing method and device, computer equipment and storage medium | |
| CN114997916A (en) | Prediction method, system, electronic device and storage medium of potential user | |
| CN112749737A (en) | Image classification method and device, electronic equipment and storage medium | |
| CN112487284A (en) | Bank customer portrait generation method, equipment, storage medium and device | |
| Lataniotis | Data-driven uncertainty quantification for high-dimensional engineering problems | |
| US11144938B2 (en) | Method and system for predictive modeling of consumer profiles | |
| US20200160200A1 (en) | Method and System for Predictive Modeling of Geographic Income Distribution | |
| CN112991026A (en) | Commodity recommendation method, system, equipment and computer readable storage medium | |
| CN113159213A (en) | Service distribution method, device and equipment | |
| CN117078007A (en) | Multi-scale wind control system integrating scale labels and method thereof | |
| US20240161117A1 (en) | Trigger-Based Electronic Fund Transfers | |
| US20230027530A1 (en) | Artificial intelligence (ai) engine assisted creation of production descriptions | |
| US11921821B2 (en) | System and method for labelling data for trigger identification | |
| US20230076149A1 (en) | Methods and apparatus for data imputation of a sparse time series data set | |
| CN117252665B (en) | Service recommendation method and device, electronic equipment and storage medium | |
| KR20200029647A (en) | Generalization method for curated e-Commerce system by user personalization |
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 | ||
| REG | Reference to a national code |
Ref country code: HK Ref legal event code: DE Ref document number: 40032024 Country of ref document: HK |