CN108765076B - Mother and infant content recommendation method and device and readable storage medium - Google Patents
Mother and infant content recommendation method and device and readable storage medium Download PDFInfo
- Publication number
- CN108765076B CN108765076B CN201810509212.0A CN201810509212A CN108765076B CN 108765076 B CN108765076 B CN 108765076B CN 201810509212 A CN201810509212 A CN 201810509212A CN 108765076 B CN108765076 B CN 108765076B
- Authority
- CN
- China
- Prior art keywords
- infant
- user
- commodity
- layer
- attribute
- 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
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
Landscapes
- Business, Economics & Management (AREA)
- Accounting & Taxation (AREA)
- Finance (AREA)
- Development Economics (AREA)
- Economics (AREA)
- Marketing (AREA)
- Strategic Management (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The embodiment of the invention provides a mother and infant content recommendation method and device and a readable storage medium. The method comprises the following steps: acquiring an embedded matrix representation of a user to be predicted, wherein the embedded matrix representation comprises commodity data related to the user to be predicted; inputting the embedded matrix representation into an infant attribute prediction model, and outputting a prediction result, wherein the prediction result comprises at least one infant attribute of the user to be predicted; and recommending corresponding maternal and infant contents to the user to be predicted based on the prediction result. Therefore, matched community content can be provided for the user, the flow guide of the mother and infant e-commerce user to the mother and infant community and the interest degree of the community content are improved, the click rate of the community content is further improved, and the cognition of the user to the content community is cultured.
Description
Technical Field
The invention relates to the field of mother and infant content recommendation, in particular to a mother and infant content recommendation method and device and a readable storage medium.
Background
With the development of vertical mother and infant e-commerce, the purchasing function of the e-commerce is provided simply and is more and more limited by the competitiveness of the expansion of varieties and supply chains, and the natural community attribute with high viscosity of mother and infant user groups can improve the user activity and the APP stay time through the development of the mother and infant community, so that the e-commerce platform is better backed up. The mode of content plus e-commerce gradually becomes the standard distribution of the vertical e-commerce industry, and the efficiency of improving mutual flow guide between e-commerce and content communities is very valuable.
In the shopping process of mother-infant e-commerce users, the cognition and the demand of the content community are weak, the browsing behaviors of the users mainly take e-commerce sales commodities and public praise comments as main parts, the search also takes e-commerce SKUs as main parts, and the cognition is lacked on the community content attached to an e-commerce platform. How to provide matched community content for a user and improve the diversion of a mother-infant e-commerce user to a mother-infant community is a technical problem to be solved urgently by technical personnel in the field.
Disclosure of Invention
In order to overcome the defects in the prior art, the invention aims to provide a mother and infant content recommendation method, a mother and infant content recommendation device and a readable storage medium, which can provide matched community content for users, improve the diversion of mother and infant e-commerce users to mother and infant communities and the interest degree of the mother and infant e-commerce users in the community content, further improve the click rate of the community content, and train the cognition of the users to the content community.
In order to achieve the above purpose, the embodiment of the present invention adopts the following technical solutions:
in a first aspect, an embodiment of the present invention provides a mother and infant content recommendation method, which is applied to a server, where an infant attribute prediction model is configured in the server, and the method includes:
acquiring an embedded matrix representation of a user to be predicted, wherein the embedded matrix representation comprises commodity data related to the user to be predicted;
inputting the embedded matrix representation into the infant attribute prediction model, and outputting a prediction result, wherein the prediction result comprises at least one infant attribute of the user to be predicted;
and recommending corresponding maternal and infant contents to the user to be predicted based on the prediction result.
Optionally, before the step of obtaining the embedding matrix of the user to be predicted, the method further includes:
configuring the infant attribute prediction model, wherein the infant attribute prediction model comprises an input layer, a convolution layer, a pooling layer, a full-link layer and an output layer;
the manner of configuring the infant attribute prediction model includes:
obtaining a training sample comprising a plurality of users having infant attributes;
processing the training sample to obtain an embedded matrix representation to be input;
inputting the embedded matrix representation to the convolutional layer through the input layer;
performing convolution operation on the embedded matrix representation through the convolution layer, and outputting a corresponding characteristic diagram to the pooling layer;
converting the characteristic graph into a corresponding multi-dimensional vector through the pooling layer and outputting the multi-dimensional vector to the full connection layer, wherein nodes of the full connection layer are connected with nodes of the pooling layer;
the node with the preset proportion discarded randomly at the full connection layer calculates the characteristic graph, outputs the calculation result to a piecewise linear function for calculation processing, and outputs the calculation processing result to the output layer;
the output layer inputs the calculation processing result into a softmax classification function to calculate a probability value of each infant attribute category;
in the training process, calculating a cross entropy loss value according to the probability value of each infant attribute category and the infant attributes of the corresponding user, updating parameters of each layer of the infant attribute prediction model by using a random gradient descent algorithm according to the cross entropy loss value, stopping training until the cross entropy loss value meets a training termination condition, and outputting target model parameters and a calculation graph;
and obtaining the infant attribute prediction model according to the target model parameters and the calculation chart.
Optionally, the step of processing the training sample to obtain an embedded matrix representation to be input includes:
for each user with infant attributes, acquiring commodity data associated with the user, wherein the commodity data comprises one or more combinations of commodity browsing data, commodity collection data and commodity purchasing data, and the commodity browsing data, the commodity collection data and the commodity purchasing data respectively comprise commodity names and corresponding commodity attributes;
filtering the commodity data, and respectively carrying out numerical processing on commodity names and corresponding commodity attributes in the filtered commodity data to generate a dictionary file, wherein the dictionary file comprises commodity name codes and commodity attribute codes of each commodity;
and processing the dictionary file to obtain an embedded matrix representation to be input.
Optionally, the manner of filtering the commodity data includes:
and filtering the commodities of which the occurrence times of the commodity names and the commodity attributes in the commodity data are less than the preset times.
Optionally, the step of processing the dictionary file to obtain an embedded matrix representation to be input includes:
counting the commodity name number and the commodity attribute number in the dictionary file to generate a counting result;
generating the length to be input of the infant attribute prediction model according to the statistical result;
searching each commodity name and each embedded vector of the commodity attribute in the dictionary file from a pre-configured word embedded matrix;
and generating an embedded matrix representation to be input based on the length to be input and the found embedded vector of each commodity name and each commodity attribute.
Optionally, the infant attribute prediction model includes an input layer, a convolutional layer, a pooling layer, a fully-connected layer, and an output layer, and the step of inputting the embedded matrix representation into the infant attribute prediction model and outputting a prediction result includes:
inputting the embedded matrix representation to the convolutional layer through the input layer;
performing convolution operation on the embedded matrix representation through the convolution layer, and outputting a corresponding characteristic diagram to the pooling layer;
converting the characteristic graph into a corresponding multi-dimensional vector through the pooling layer and outputting the multi-dimensional vector to the full connection layer, wherein nodes of the full connection layer are connected with nodes of the pooling layer;
the full connection layer calculates the characteristic graph by using all nodes, outputs a calculation result to a piecewise linear function for calculation processing, and outputs a corresponding calculation processing result to the output layer;
the output layer inputs the calculation processing result into a softmax classification function to calculate a probability value of each infant attribute category;
and selecting the infant attribute with the highest probability value as a prediction result.
Optionally, the step of recommending corresponding maternal and infant content to the user to be predicted based on the prediction result includes:
recommending the UGC content and the PGC content which are matched to the user based on the prediction result, wherein the PGC content is associated with corresponding infant attribute options;
and acquiring the corresponding user set and the commodity data of each user in the user set based on the prediction result, and recommending matched commodity contents to the user according to the corresponding user set and the commodity data of each user in the user set.
Optionally, the step of recommending the matched commodity content to the user according to the corresponding user set and the commodity data of each user in the user set includes:
classifying the user set and the commodity data of each user in the user set according to a confidence threshold value to obtain a classification result;
and recommending the matched commodity content to the user according to the classification result.
In a second aspect, an embodiment of the present invention further provides a maternal-infant content recommendation device, which is applied to a server, where an infant attribute prediction model is configured in the server, and the device includes:
the system comprises an acquisition module, a prediction module and a prediction module, wherein the acquisition module is used for acquiring an embedded matrix representation of a user to be predicted, and the embedded matrix representation comprises commodity data related to the user to be predicted;
the input module is used for inputting the embedded matrix representation into the infant attribute prediction model and outputting a prediction result, wherein the prediction result comprises at least one infant attribute of the user to be predicted;
and the recommending module is used for recommending corresponding maternal and infant contents to the user to be predicted based on the prediction result.
In a third aspect, an embodiment of the present invention further provides a readable storage medium, where a computer program is stored in the readable storage medium, and when the computer program is executed, the method for recommending maternal and infant content as described above is implemented.
Compared with the prior art, the invention has the following beneficial effects:
according to the mother and infant content recommendation method, device and readable storage medium provided by the embodiment of the invention, the embedded matrix representation of the user to be predicted is obtained and input into the infant attribute prediction model, the prediction result is output, the prediction result comprises at least one infant attribute of the user to be predicted, and then the corresponding mother and infant content is recommended to the user to be predicted based on the prediction result. Therefore, matched community content can be provided for the user according to the predicted attributes of the user and the infant, the flow guide of the mother and infant e-commerce user to the mother and infant community and the interest degree of the mother and infant e-commerce user to the community content are improved, the click rate of the community content is further improved, and the cognition of the user to the content community is cultured.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings that are required to be used in the embodiments will be briefly described below, it should be understood that the following drawings only illustrate some embodiments of the present invention and therefore should not be considered as limiting the scope, and for those skilled in the art, other related drawings can be obtained according to the drawings without inventive efforts.
Fig. 1 is a schematic view of an application scenario of a mother-infant content recommendation method according to an embodiment of the present invention;
fig. 2 is a schematic flow chart of a method for recommending mother and infant content according to an embodiment of the present invention;
FIG. 3 is a block diagram of an infant attribute prediction model according to an embodiment of the present invention;
fig. 4 is a block diagram of a structure of a server for implementing the maternal-infant content recommendation method according to an embodiment of the present invention;
fig. 5 is a functional block diagram of the mother-infant content recommendation apparatus shown in fig. 4.
Icon: 100-a server; 110-a bus; 120-a processor; 130-a storage medium; 140-bus interface; 150-a network adapter; 160-a user interface; 200-mother and infant content recommendation device; 210-an obtaining module; 220-an input module; 230-a recommendation module; 300-user terminal.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, not all, embodiments of the present invention. The components of embodiments of the present invention generally described and illustrated in the figures herein may be arranged and designed in a wide variety of different configurations.
Thus, the following detailed description of the embodiments of the present invention, presented in the figures, is not intended to limit the scope of the invention, as claimed, but is merely representative of selected embodiments of the invention. 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 invention.
It should be noted that: like reference numbers and letters refer to like items in the following figures, and thus, once an item is defined in one figure, it need not be further defined and explained in subsequent figures. Meanwhile, in the description of the present invention, the terms "first", "second", and the like are used only for distinguishing the description, and are not to be construed as indicating or implying relative importance.
At present, with the development of vertical mother and infant e-commerce, the purchasing function of the e-commerce is simply provided, and the e-commerce is more and more limited by the competitiveness of the expansion of categories and a supply chain, and the natural community attribute with high viscosity of a mother and infant user group can improve the user activity and the APP stay time through the development of the mother and infant community, so that the e-commerce platform is better backed up. The mode of content plus e-commerce gradually becomes the standard distribution of the vertical e-commerce industry, and the efficiency of improving mutual flow guide between e-commerce and content communities is very valuable.
The inventor finds that the mother-infant community has a large characteristic that users are grouped in the community, and users with different infant attributes have large purchasing and community demand differences. For example, the user during pregnancy pays more attention to health care goods such as maternity dress, radiation protection clothes, folic acid and the like, and pays more attention to items needing attention during pregnancy in a mother-infant community, such as incapability of eating certain food, whether certain changes of the body are normal or not, items needing attention during pregnancy inspection and key index interpretation. Mothers who are 0-1 year old infants pay more attention to breast feeding, infant growth indexes, knowledge that the beginners and mothers must read and the like, and shopping is mainly favored by milk powder, paper diapers and other commodities. Mothers of 1-3 years old infants pay more attention to infant cognitive culture, and can purchase related products such as picture books, toys and children, and more related knowledge such as children, tourism and infant education is mainly used in the community. Mothers of infants of 3-6 years old pay more attention to children's clothes of the infants and infant nutrition-related commodities, and pay more attention to education and child activity-related contents of the infants in the ages in the community. Therefore, the inventor draws a conclusion that grouping the users can better help the e-commerce and the community to carry out accurate flow distribution, and the user viscosity is improved.
However, the inventors have found that the following problems exist in the prior art: the method is characterized in that cognition and demand of a community of a mother-infant e-commerce user are weak in the shopping process, the browsing behavior of the user mainly takes e-commerce sale of commodities and public praise comments as main parts, searching is also mainly based on e-commerce SKUs, the community content attached to an e-commerce platform is lack of cognition, and if general community content is provided and fine matching is not performed on the attributes of the user infants, the interest degree of most users in the community content is not high. In view of this, the inventor believes that if the user can depict accurate user portrait and infant attributes of the user based on the browsing behavior, the collection behavior and the shopping behavior of the user when browsing e-commerce commodities and public praise, related community contents such as accurate knowledge and topics can be provided, so that the interest of the e-commerce user in the community contents can be improved, the click rate of the user on the community contents can be further improved, and the cognition of the user on the content community can be developed.
Specifically, the standard matching of the mother-infant community is that the initial user needs to set infant attribute information when using, so that the users can be grouped, and more accurate content is provided. However, it is not a common practice to set infant attributes on a platform mainly based on e-commerce, because even if a setting entrance is provided, the filling intention of a user is not strong relative to a mother-infant community platform, and because the community of the mother-infant e-commerce platform lacks a flow entrance, less infant attribute information can be acquired, but the infant attribute information cannot be acquired, more accurate contents cannot be provided to attract the user to browse, and a closed loop with negative feedback is formed. In order to solve the problems, the inventor finds that the mother-infant e-commerce platform has abundant e-commerce user behaviors in research, so that the e-commerce user behaviors with existing infant attributes can be labeled, and the user infant attributes without written information can be estimated by adding new user behaviors to a deep learning model. Meanwhile, the community content browsed by the existing infant attributes can form sample labels of resource types, and the infant attributes mapped by the part of resources are carved, so that the part of content can be accurately distributed to users with proper infant attributes, and the flow of mother and infant e-commerce users to the mother and infant community is improved. Therefore, the present invention proposes the following embodiments to solve many disadvantages in the prior art.
The above prior art solutions have shortcomings which are the results of practical and careful study of the inventor, and therefore, the discovery process of the above problems and the solutions proposed by the following embodiments of the present invention to the above problems should be the contribution of the inventor to the present invention in the course of the present invention.
Fig. 1 is a schematic view of an application scenario of the mother-infant content recommendation method according to an embodiment of the present invention. In this embodiment, the application scenario includes at least one user terminal 300 and a server 100 in communication connection with the at least one user terminal 300.
In this embodiment, the user terminal 300 may be a personal computer, a smart phone, a tablet computer, a wearable device, and the like, which is not limited in detail in this embodiment.
According to some embodiments of the present invention, the user terminal 300 may include: a processing device including an application processing section and a radio frequency/digital signal processor; a display screen; a keypad that may include physical keys, touch keys overlaid on a display, or a combination thereof; a subscriber identity module card; memory devices that may include ROM, RAM, flash memory, or any combination thereof; a Wi-Fi and/or Bluetooth interface; the NFC chip, the wireless power receiving coil used for wireless charging and the wireless telephone interface; a power management circuit with an associated battery; a USB interface and a connector; an audio management system with associated microphone, speaker and headphone jack; and various optional accessory components such as cameras, global positioning systems, accelerators, etc. In addition, various client applications may be installed on the user terminal 300, which may be used to allow the user terminal 300 to be used to communicate commands suitable for operation with other devices. Such applications may be downloaded from the server 100 and installed in the memory of the user terminal 300, or may have been previously installed on the user terminal 300. In the embodiment of the present invention, the user terminal 300 is installed with a mother-and-infant content recommendation application, and the mother-and-infant content recommendation application can point a user to implement functions of user registration, user login, commodity collection, commodity purchase, mother-and-infant community content browsing, user invitation, and the like.
In this embodiment, the server 100 should be understood as a service point providing processes, databases, and communication facilities. By way of example, server 100 may refer to a single physical processor with associated communication and data storage and library facilities, or it may refer to a networked or clustered collection of processors, associated networks, and storage devices, and operating on software and one or more library systems and application software that support the services provided by server 100. The servers 100 may vary widely in configuration or performance, but the servers 100 may generally include one or more central processing units and memory units. The Server 100 may also include one or more mass storage devices, one or more power supplies, one or more wired or wireless network components, one or more input/output components, or one or more operating systems, such as Windows Server, Mac OS X, Unix, Linux, FreeBSD, and so forth.
Further, please refer to fig. 2, which is a flowchart illustrating a method for recommending maternal and infant content according to an embodiment of the present invention, the method is executed by the server 100 shown in fig. 1. It should be noted that, the method for recommending maternal and infant content provided in the embodiment of the present invention is not limited by fig. 2 and the following specific sequence, and the method for recommending maternal and infant content may be implemented by the following steps:
step S210, acquiring an embedded matrix representation of the user to be predicted.
In this embodiment, the embedded matrix representation includes commodity data associated with the user to be predicted, where the commodity data may represent a user behavior of the user to be predicted, and may specifically include one or more combinations of commodity browsing data, commodity collection data, and commodity purchase data. For example, the sku-id, sku attribute id, collected sku-id, sku attribute id, purchased sku-id, sku attribute id, etc. that the user is e-commerce browsing. The sku-id may be understood as a specific name of the article and the sku attribute id may be understood as an attribute of the article, for example, for infant clothing, the sku attribute id may include color, size, style, and the like.
In detail, before further describing step S210, a description will be given of an arrangement manner of the infant attribute prediction model. Referring to fig. 3, in the present embodiment, the infant attribute prediction model may include an input layer, a convolutional layer, a pooling layer, a full link layer, and an output layer.
First, a training sample is obtained, the training sample includes a plurality of users with infant attributes, which may be defined and extended according to an actual scene, for example, the infant attributes may include: the age, sex, constellation and age classification of the infant may be a pregnancy preparation period, a pregnancy period, 0-1 year, 1-3 years and 3-6 years, the sex of the infant is mainly male and female, the constellation of the infant is mainly twelve constellation, and the like, and is not limited in detail herein.
And then, processing the training sample to obtain an embedded matrix representation to be input. The following describes the generation process of the embedded matrix representation with reference to fig. 3, and it should be noted that the generation of the embedded matrix representation of the user to be predicted may also be realized in the following manner.
First, for each user having an infant attribute, product data having an association with the user is acquired. The commodity data comprises one or more of commodity browsing data, commodity collection data and commodity purchasing data. The commodity browsing data, the commodity collection data and the commodity purchase data respectively comprise commodity names sku-id and corresponding commodity attribute sku attribute ids.
And then, filtering the commodity data, and respectively carrying out numerical processing on the commodity names and the corresponding commodity attributes in the filtered commodity data, namely, uniformly coding the sku-id and the sku attribute id, replacing the sku-id and the sku attribute id with the numbers of the sku-id and the sku attribute id in a dictionary file, generating sku-id and sku attribute id codes, digitizing the sku-id and sku attribute id codes and generating a dictionary file D, wherein the dictionary file comprises the commodity name codes and the commodity attribute codes of each commodity. Optionally, the commodity data may be filtered in a manner that the number of occurrences of the commodity name and the commodity attribute in the commodity data is less than a preset number of times (e.g., twice), so that the accuracy of the training sample may be improved.
And then, processing the dictionary file to obtain an embedded matrix representation to be input. As an implementation manner, a statistical result may be generated by counting the number of names of goods and the number of attributes of goods in the dictionary file, and the length to be input of the infant attribute prediction model may be generated according to the statistical result. For example, the number of sku and the number of sku attributes of each week visited by all users may be counted, the average length of the number of sku and the number of sku attributes after the number of sku attributes is selected as the length L to be input of the infant attribute prediction model, a section that is longer than L may be performed, and a title that is shorter than L is filled with 0. And then, randomly initializing a word embedding matrix E with dimension of D multiplied by S, and searching an embedding vector of each commodity name and commodity attribute in the dictionary file D from the word embedding matrix E, namely, searching an embedding vector with length of S in the corresponding word embedding matrix E through corresponding codes for each sku-id and sku attribute id. And finally, combining a matrix X with the dimension of L multiplied by S based on the length to be input and the found embedded vector of each commodity name and each commodity attribute, wherein the matrix X is also represented by the embedded matrix to be input, and the infant attribute of the user is used as a training label.
The embedded matrix representation X is then input as input to the convolutional layer through the input layer, while the user's infant attributes are used as training labels.
Next, convolution operation is performed on the embedded matrix representation by the convolutional layer, and a corresponding feature map is output to the pooling layer. In one embodiment, the convolution layer may perform convolution operation with step size 1 on the input layer by using 128 one-dimensional convolution kernels with width 5 and 128 one-dimensional convolution kernels with width 2, where the convolution kernels with different widths are respectively responsible for extracting local features with different ranges, and when performing convolution operation, zero padding is required to be performed on the embedding matrix representation so that the length of the output feature map is equal to the length of the input embedding matrix representation. After convolution operation, two L multiplied by 128 feature maps can be obtained, and the two feature maps are simply spliced to form the L multiplied by 256 feature map.
And then, converting the characteristic diagram into a corresponding multidimensional vector through the pooling layer and outputting the multidimensional vector to the full-connection layer. Therefore, the L multiplied by 256 feature maps are converted into 256-dimensional vectors through the maximized pooling operation, the pooling operation reduces the dimensionality of the feature maps, and the generalization capability of the network is improved. Wherein the nodes of the full connection layer are connected with the nodes of the pooling layer.
Then, the fully connected layer randomly discards a predetermined proportion (e.g., 50%) of the nodes to calculate the feature map, outputs the calculation result to a piecewise linear function (ReLU) for calculation processing, and outputs the calculation processing result to the output layer. Wherein the formula of the piecewise linear function (ReLU) is f (x) max (0, x), where x is the output of the fully connected layer, and f (x) is the output of the piecewise linear function.
Then, the output layer inputs the calculation processing result into a softmax classification function to calculate a probability value of each infant attribute category, wherein the calculation formula of the softmax is
In the training process, a cross-entropy loss value (cross-entropy cost) is calculated according to the probability value of each infant attribute category and the infant attributes of the corresponding users, each layer of parameters of the infant attribute prediction model are updated by using a random gradient descent algorithm according to the cross-entropy loss value, the training is stopped until the cross-entropy loss value meets the training termination condition, and target model parameters and a calculation graph are output. Therefore, the infant attribute prediction model can be obtained according to the target model parameters and the calculation graph.
After the configuration of the infant attribute prediction model is completed, firstly, a calculation graph and target model parameters of the infant attribute prediction model are loaded when the embedded matrix representation of a user to be predicted is obtained.
And S220, inputting the embedded matrix representation into the infant attribute prediction model, and outputting a prediction result.
In implementation, the embedded matrix representation is input to the convolutional layer through the input layer, convolution operation is performed on the embedded matrix representation through the convolutional layer, a corresponding feature map is output to the pooling layer, and the feature map is converted into a corresponding multi-dimensional vector through the pooling layer and then output to the full connection layer. And then the full connection layer calculates the characteristic graph by using all nodes, outputs the calculation result to a piecewise linear function for calculation processing, and outputs a corresponding calculation processing result to the output layer. And the output layer inputs the calculation processing result into a softmax classification function to calculate a probability value of each infant attribute category, and selects the infant attribute with the highest probability value as a prediction result, wherein the prediction result comprises at least one infant attribute of the user to be predicted.
The specific operation method of the above process may refer to the corresponding detailed description in the configuration manner of the infant attribute prediction model, and will not be repeated herein.
Therefore, the embodiment predicts the classification score of the sku-id and sku attribute id accessed, collected and purchased by the user without infant attributes by using an infant attribute prediction model to obtain a result set P { P1，p2，…，pjAnd (6) taking the result set as the attributes of the infants of the userThe final classification of (1). E.g. p1Indicating gender attribute as female, p2Representing an attribute of 0-1 year old, the user's infant attribute characterizes a baby girl of 0-1 year old.
And step S230, recommending corresponding maternal and infant contents to the user to be predicted based on the prediction result.
In this embodiment, the mother and infant content can be divided into two forms: the content management system comprises production content and consumption content, wherein the production content is UGC content and PGC content, and the consumption content is content actively browsed, praised and collected by a user. As an embodiment, after obtaining the prediction result, the matching UGC content and PGC content may be recommended to the user based on the prediction result.
The UGC (user Generated content) content refers to original content of the user, that is, when other users with infant attributes issue corresponding UGC content in the maternal-infant community, if the infant attributes of the user are similar to the infant attributes of the user to be predicted, the server 100 may push the UGC content issued by the user to the user terminal 300 of the user to be predicted.
PGC (professional Generated content) content refers to professional produced content. The PGC content is associated with corresponding infant attribute options, that is, relevant experts may select a matched infant attribute when creating the PGC content, and if the selected infant attribute is similar to the infant attribute of the user to be predicted, the server 100 may push the PGC content to the user terminal 300 of the user to be predicted. Therefore, accurate UGC content and PGC content can be pushed for the user to be predicted.
For the consumption content, the commodity data of each user in the corresponding user set and the user set can be obtained based on the prediction result, and the matched commodity content can be recommended to the user according to the commodity data of each user in the corresponding user set and the user set. For example, a set of all users U { U } for the consumed content may be obtained1，u2…unInfant attribute P { P }u1，pu2…punGet p according to a confidence threshold μuiSet of trusted classification resultsC{c│c>Mu, and then recommending the matched commodity content to the user according to the classification result.
Therefore, through the steps, the embodiment can provide matched community content for the user according to the predicted attributes of the user and the infant, the flow guide of the mother and infant e-commerce user to the mother and infant community and the interest degree of the mother and infant e-commerce user to the community are improved, the click rate of the community content is further improved, and the cognition of the user to the content community is cultured.
Referring to fig. 4, a block diagram of a server 100 according to a preferred embodiment of the invention is shown. As shown in FIG. 4, the server 100 may be implemented by a bus 110 as a general bus architecture. The bus 110 may include any number of interconnecting buses and bridges depending on the specific application of the server 100 and the overall design constraints. Bus 110 connects various circuits together, including processor 120, storage medium 130, and bus interface 140. Alternatively, the server 100 may connect a network adapter 150 or the like via the bus 110 using the bus interface 140. The network adapter 150 may be used to implement signal processing functions of a physical layer in a wireless communication network and implement transmission and reception of radio frequency signals through an antenna. The user interface 160 may connect external devices such as: a keyboard, a display, a mouse or a joystick, etc. The bus 110 may also connect various other circuits such as timing sources, peripherals, voltage regulators, or power management circuits, which are well known in the art, and therefore, will not be described in detail.
Alternatively, the server 100 may be configured as a general purpose processing system, such as what is commonly referred to as a chip, including: one or more microprocessors providing processing functions, and an external memory providing at least a portion of storage medium 130, all connected together with other support circuits through an external bus architecture.
Alternatively, the server 100 may be implemented using: an ASIC (application specific integrated circuit) having a processor 120, a bus interface 140, a user interface 160; and at least a portion of the storage medium 130 integrated in a single chip, or the server 100 may be implemented using: one or more FPGAs (field programmable gate arrays), PLDs (programmable logic devices), controllers, state machines, gate logic, discrete hardware components, any other suitable circuitry, or any combination of circuitry capable of performing the various functions described throughout this disclosure.
Among other things, processor 120 is responsible for managing bus 110 and general processing (including the execution of software stored on storage medium 130). Processor 120 may be implemented using one or more general-purpose processors and/or special-purpose processors. Examples of processor 120 include microprocessors, microcontrollers, DSP processors, and other circuits capable of executing software. Software should be construed broadly to mean instructions, data, or any combination thereof, whether referred to as software, firmware, middleware, microcode, hardware description language, or otherwise.
The processor 120 may execute the above embodiments, specifically, the storage medium 130 may store the maternal-infant content recommendation device 200 therein, and the processor 120 may be configured to execute the maternal-infant content recommendation device 200.
Further, referring to fig. 5, the mother-infant content recommendation device 200 may include:
an obtaining module 210, configured to obtain an embedded matrix representation of a user to be predicted, where the embedded matrix representation includes commodity data associated with the user to be predicted;
an input module 220, configured to input the embedded matrix representation into the infant attribute prediction model, and output a prediction result, where the prediction result includes at least one infant attribute of the user to be predicted;
and the recommending module 230 is configured to recommend the corresponding maternal and infant content to the user to be predicted based on the prediction result.
It can be understood that, for the specific operation method of each functional module in this embodiment, reference may be made to the detailed description of the corresponding step in the foregoing method embodiment, and no repeated description is provided herein.
The embodiment of the invention also provides a readable storage medium, wherein a computer program is stored in the readable storage medium, and when the computer program is executed, the method for recommending the maternal and infant content is realized.
In summary, according to the maternal-infant content recommendation method, the device and the readable storage medium provided by the embodiments of the present invention, the embedded matrix representation of the user to be predicted is obtained and input into the infant attribute prediction model, the prediction result is output, the prediction result includes at least one infant attribute of the user to be predicted, and then the corresponding maternal-infant content is recommended to the user to be predicted based on the prediction result. Therefore, matched community content can be provided for the user according to the predicted attributes of the user and the infant, the flow guide of the mother and infant e-commerce user to the mother and infant community and the interest degree of the mother and infant e-commerce user to the community content are improved, the click rate of the community content is further improved, and the cognition of the user to the content community is cultured.
In the embodiments provided in the present invention, it should be understood that the disclosed apparatus and method can be implemented in other ways. The apparatus and method embodiments described above are illustrative only, as the flowcharts and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present invention. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and/or flowchart illustration, and combinations of blocks in the block diagrams and/or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.
In addition, the functional modules in the embodiments of the present invention may be integrated together to form an independent part, or each module may exist separately, or two or more modules may be integrated to form an independent part.
Alternatively, all or part of the implementation may be in 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, cause the processes or functions described in accordance with the embodiments of the invention to occur, in whole or in part. The computer may be a general purpose computer, a special purpose computer, a network of computers, or other programmable device. The computer instructions may be stored in a computer readable storage medium or transmitted from one computer readable storage medium to another, for example, from one website site, computer, server, or data center to another website site, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, Digital Subscriber Line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device, such as a server, a data center, etc., that incorporates one or more of the 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)), among others
It should be noted that, in this document, 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 an … …" does not exclude the presence of other identical elements in a process, method, article, or apparatus that comprises the element.
It will be evident to those skilled in the art that the invention is not limited to the details of the foregoing illustrative embodiments, and that the present invention may be embodied in other specific forms without departing from the spirit or essential attributes thereof. The present embodiments are therefore to be considered in all respects as illustrative and not restrictive, the scope of the invention being indicated by the appended claims rather than by the foregoing description, and all changes which come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein. Any reference sign in a claim should not be construed as limiting the claim concerned.
Claims (9)
1. A mother-infant content recommendation method is applied to a server, wherein an infant attribute prediction model is configured in the server, and the method comprises the following steps:
acquiring an embedded matrix representation of a user to be predicted, wherein the embedded matrix representation comprises commodity data related to the user to be predicted;
inputting the embedded matrix representation into the infant attribute prediction model, and outputting a prediction result, wherein the prediction result comprises at least one infant attribute of the user to be predicted;
the infant attribute prediction model comprises an input layer, a convolution layer, a pooling layer, a full-link layer and an output layer, wherein the step of inputting the embedded matrix representation into the infant attribute prediction model and outputting a prediction result comprises the following steps:
inputting the embedded matrix representation to the convolutional layer through the input layer;
performing convolution operation on the embedded matrix representation through the convolution layer, and outputting a corresponding characteristic diagram to the pooling layer;
converting the characteristic graph into a corresponding multi-dimensional vector through the pooling layer and outputting the multi-dimensional vector to the full connection layer, wherein nodes of the full connection layer are connected with nodes of the pooling layer;
the full connection layer calculates the characteristic graph by using all nodes, outputs a calculation result to a piecewise linear function for calculation processing, and outputs a corresponding calculation processing result to the output layer;
the output layer inputs the calculation processing result into a softmax classification function to calculate a probability value of each infant attribute category;
selecting the infant attribute with the highest probability value as a prediction result;
and recommending corresponding maternal and infant contents to the user to be predicted based on the prediction result.
2. The method for recommending maternal and infant content according to claim 1, wherein before the step of obtaining an embedding matrix of the user to be predicted, the method further comprises:
configuring the infant attribute prediction model, wherein the infant attribute prediction model comprises an input layer, a convolution layer, a pooling layer, a full-link layer and an output layer;
the manner of configuring the infant attribute prediction model includes:
obtaining a training sample comprising a plurality of users having infant attributes;
training the infant attribute prediction model by using the training samples, wherein the training process comprises the following steps:
processing the training sample to obtain an embedded matrix representation to be input;
inputting the embedded matrix representation to the convolutional layer through the input layer;
performing convolution operation on the embedded matrix representation through the convolution layer, and outputting a corresponding characteristic diagram to the pooling layer;
converting the characteristic graph into a corresponding multi-dimensional vector through the pooling layer and outputting the multi-dimensional vector to the full connection layer, wherein nodes of the full connection layer are connected with nodes of the pooling layer;
the node with the preset proportion discarded randomly at the full connection layer calculates the characteristic graph, outputs the calculation result to a piecewise linear function for calculation processing, and outputs the calculation processing result to the output layer;
the output layer inputs the calculation processing result into a softmax classification function to calculate a probability value of each infant attribute category;
in the training process, calculating a cross entropy loss value according to the probability value of each infant attribute category and the infant attributes of the corresponding user, updating parameters of each layer of the infant attribute prediction model by using a random gradient descent algorithm according to the cross entropy loss value, stopping training until the cross entropy loss value meets a training termination condition, and outputting target model parameters and a calculation graph;
and obtaining the trained infant attribute prediction model according to the target model parameters and the calculation chart.
3. The method for recommending maternal and infant content according to claim 2, wherein the step of processing the training samples to obtain an embedded matrix representation to be input comprises:
for each user with infant attributes, acquiring commodity data associated with the user, wherein the commodity data comprises one or more combinations of commodity browsing data, commodity collection data and commodity purchasing data, and the commodity browsing data, the commodity collection data and the commodity purchasing data respectively comprise commodity names and corresponding commodity attributes;
filtering the commodity data, and respectively carrying out numerical processing on commodity names and corresponding commodity attributes in the filtered commodity data to generate a dictionary file, wherein the dictionary file comprises commodity name codes and commodity attribute codes of each commodity;
and processing the dictionary file to obtain an embedded matrix representation to be input.
4. The method for recommending maternal and infant content according to claim 3, wherein said filtering said commodity data comprises:
and filtering the commodities of which the occurrence times of the commodity names and the commodity attributes in the commodity data are less than the preset times.
5. The method for recommending mother and infant contents according to claim 3, wherein the step of processing the dictionary file to obtain the embedded matrix representation to be inputted comprises:
counting the commodity name number and the commodity attribute number in the dictionary file to generate a counting result;
generating the length to be input of the infant attribute prediction model according to the statistical result;
searching each commodity name and each embedded vector of the commodity attribute in the dictionary file from a pre-configured word embedded matrix;
and generating an embedded matrix representation to be input based on the length to be input and the found embedded vector of each commodity name and each commodity attribute.
6. The method for recommending maternal and infant content according to claim 1, wherein the step of recommending the corresponding maternal and infant content to the user to be predicted based on the prediction result comprises:
recommending the UGC content and the PGC content which are matched to the user based on the prediction result, wherein the PGC content is associated with corresponding infant attribute options;
and acquiring the corresponding user set and the commodity data of each user in the user set based on the prediction result, and recommending matched commodity contents to the user according to the corresponding user set and the commodity data of each user in the user set.
7. The maternal and infant content recommendation method according to claim 6, wherein the step of recommending to the user the matched commodity content according to the corresponding user set and the commodity data of each user in the user set comprises:
classifying the user set and the commodity data of each user in the user set according to a confidence threshold value to obtain a classification result;
and recommending the matched commodity content to the user according to the classification result.
8. A maternal and infant content recommendation device, applied to a server in which an infant attribute prediction model is configured, the device comprising:
the system comprises an acquisition module, a prediction module and a prediction module, wherein the acquisition module is used for acquiring an embedded matrix representation of a user to be predicted, and the embedded matrix representation comprises commodity data related to the user to be predicted;
the input module is used for inputting the embedded matrix representation into the infant attribute prediction model and outputting a prediction result, wherein the prediction result comprises at least one infant attribute of the user to be predicted;
the infant attribute prediction model comprises an input layer, a convolutional layer, a pooling layer, a full-link layer and an output layer, and the mode of inputting the embedded matrix representation into the infant attribute prediction model and outputting a prediction result comprises the following steps:
inputting the embedded matrix representation to the convolutional layer through the input layer;
performing convolution operation on the embedded matrix representation through the convolution layer, and outputting a corresponding characteristic diagram to the pooling layer;
converting the characteristic graph into a corresponding multi-dimensional vector through the pooling layer and outputting the multi-dimensional vector to the full connection layer, wherein nodes of the full connection layer are connected with nodes of the pooling layer;
the full connection layer calculates the characteristic graph by using all nodes, outputs a calculation result to a piecewise linear function for calculation processing, and outputs a corresponding calculation processing result to the output layer;
the output layer inputs the calculation processing result into a softmax classification function to calculate a probability value of each infant attribute category;
selecting the infant attribute with the highest probability value as a prediction result;
and the recommending module is used for recommending corresponding maternal and infant contents to the user to be predicted based on the prediction result.
9. A readable storage medium, wherein a computer program is stored in the readable storage medium, and when the computer program is executed, the method for recommending maternal and infant content according to any one of claims 1 to 7 is implemented.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810509212.0A CN108765076B (en) | 2018-05-24 | 2018-05-24 | Mother and infant content recommendation method and device and readable storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810509212.0A CN108765076B (en) | 2018-05-24 | 2018-05-24 | Mother and infant content recommendation method and device and readable storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108765076A CN108765076A (en) | 2018-11-06 |
| CN108765076B true CN108765076B (en) | 2021-08-03 |
Family
ID=64005714
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810509212.0A Active CN108765076B (en) | 2018-05-24 | 2018-05-24 | Mother and infant content recommendation method and device and readable storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108765076B (en) |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110119935A (en) * | 2019-03-20 | 2019-08-13 | 中国空气动力研究与发展中心高速空气动力研究所 | A kind of science and technology item declares method for procedure tracking and device |
| CN111274497B (en) * | 2020-01-22 | 2023-08-18 | 北京百度网讯科技有限公司 | Community recommendation and model training method and device, electronic equipment and storage medium |
| CN111292168B (en) * | 2020-02-06 | 2022-08-02 | 腾讯科技（深圳）有限公司 | Data processing method, device and equipment |
| CN111861178B (en) * | 2020-07-13 | 2024-06-07 | 北京嘀嘀无限科技发展有限公司 | Training method of service matching model, service matching method, equipment and medium |
| CN112506975A (en) * | 2020-11-30 | 2021-03-16 | 橙脑教育科技(上海)有限公司 | Information recommendation method, device and system |
| CN115619046B (en) * | 2022-12-05 | 2023-03-28 | 湖南工商大学 | Commercial configuration method for community life circle |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105868317B (en) * | 2016-03-25 | 2017-04-12 | 华中师范大学 | Digital education resource recommendation method and system |
| CN106874522A (en) * | 2017-03-29 | 2017-06-20 | 珠海习悦信息技术有限公司 | Information recommendation method, device, storage medium and processor |
| CN107330115B (en) * | 2017-07-12 | 2020-04-28 | 广东工业大学 | Information recommendation method and device |
| CN107766439A (en) * | 2017-09-21 | 2018-03-06 | 汉鼎宇佑互联网股份有限公司 | A kind of personalized recommendation method of fusion structure feature and implicit feedback |
| CN107992531B (en) * | 2017-11-21 | 2020-11-27 | 吉浦斯信息咨询（深圳）有限公司 | News personalized intelligent recommendation method and system based on deep learning |
-
2018
- 2018-05-24 CN CN201810509212.0A patent/CN108765076B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN108765076A (en) | 2018-11-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108765076B (en) | Mother and infant content recommendation method and device and readable storage medium | |
| Lakshmanan et al. | Machine learning design patterns | |
| CN109299994B (en) | Recommendation method, device, equipment and readable storage medium | |
| CN106651057B (en) | Mobile terminal user age prediction method based on installation package sequence list | |
| CN111523010A (en) | Recommendation method and device, terminal equipment and computer storage medium | |
| CN110851713A (en) | Information processing method, recommendation method and related equipment | |
| CN111225009B (en) | Method and device for generating information | |
| CN108984688B (en) | Mother and infant knowledge topic recommendation method and device | |
| CN111368072A (en) | Microblog hot topic discovery algorithm based on linear fusion of BTM and GloVe similarity | |
| CN113611405A (en) | Physical examination item recommendation method, device, equipment and medium | |
| CN107145485A (en) | Method and apparatus for compressing topic model | |
| CN112288554B (en) | Commodity recommendation method and device, storage medium and electronic device | |
| US20240143684A1 (en) | Information presentation method and apparatus, and device and medium | |
| CN118132752B (en) | Commodity description word classification method and device | |
| CN112132633B (en) | Consumption intention identification and prediction method based on consumption affair map | |
| CN110852785A (en) | User grading method, device and computer readable storage medium | |
| CN110473073A (en) | The method and device that linear weighted function is recommended | |
| US9058328B2 (en) | Search device, search method, search program, and computer-readable memory medium for recording search program | |
| CN110363206A (en) | Cluster, data processing and the data identification method of data object | |
| CN116340643B (en) | Object recommendation adjustment method and device, storage medium and electronic equipment | |
| CN117196716A (en) | Digital signage advertisement theme recommendation method based on Transformer network model | |
| CN112633988A (en) | User product recommendation method and device, electronic equipment and readable storage medium | |
| CN113744002A (en) | Method, device, equipment and computer readable medium for pushing information | |
| CN108319721B (en) | Application program recommendation method and device, user terminal and server | |
| JP6541737B2 (en) | Selection apparatus, selection method, selection program, model and learning data |
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 | ||
| TR01 | Transfer of patent right | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20220920 Address after: 311100 rooms 809 and 810, building 8, Xixi bafangcheng, Wuchang Street, Yuhang District, Hangzhou City, Zhejiang Province Patentee after: Hangzhou Mitu network technology (Group) Co.,Ltd. Address before: 100000 room 1606, 16th floor, building 1, yard 16, Taiyanggong Middle Road, Chaoyang District, Beijing Patentee before: MIYA BAOBEI (BEIJING) NETWORK TECHNOLOGY Co.,Ltd. |