CN112507234B - Material pushing method and device, storage medium and electronic equipment - Google Patents
Material pushing method and device, storage medium and electronic equipment Download PDFInfo
- Publication number
- CN112507234B CN112507234B CN202011519601.5A CN202011519601A CN112507234B CN 112507234 B CN112507234 B CN 112507234B CN 202011519601 A CN202011519601 A CN 202011519601A CN 112507234 B CN112507234 B CN 112507234B
- Authority
- CN
- China
- Prior art keywords
- layer
- prediction model
- training
- information
- target user
- 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
- 239000000463 material Substances 0.000 title claims abstract description 56
- 238000000034 method Methods 0.000 title claims abstract description 42
- 239000000523 sample Substances 0.000 claims description 60
- 238000012549 training Methods 0.000 claims description 54
- 230000006399 behavior Effects 0.000 claims description 40
- 238000011176 pooling Methods 0.000 claims description 32
- 230000015654 memory Effects 0.000 claims description 26
- 238000006243 chemical reaction Methods 0.000 claims description 25
- 238000004891 communication Methods 0.000 claims description 16
- 230000006870 function Effects 0.000 claims description 16
- 238000012545 processing Methods 0.000 claims description 14
- 238000010606 normalization Methods 0.000 claims description 13
- 238000004590 computer program Methods 0.000 claims description 11
- 238000007477 logistic regression Methods 0.000 claims description 10
- 230000004931 aggregating effect Effects 0.000 claims description 9
- 230000008569 process Effects 0.000 claims description 6
- 230000001737 promoting effect Effects 0.000 claims description 2
- 238000005516 engineering process Methods 0.000 abstract description 4
- 238000013473 artificial intelligence Methods 0.000 abstract description 2
- 238000010586 diagram Methods 0.000 description 10
- 230000005540 biological transmission Effects 0.000 description 6
- 230000003287 optical effect Effects 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 2
- 238000013145 classification model Methods 0.000 description 2
- 230000008878 coupling Effects 0.000 description 2
- 238000010168 coupling process Methods 0.000 description 2
- 238000005859 coupling reaction Methods 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 230000000306 recurrent effect Effects 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- 239000013598 vector Substances 0.000 description 2
- 108700019146 Transgenes Proteins 0.000 description 1
- 230000001133 acceleration Effects 0.000 description 1
- 230000006978 adaptation Effects 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 239000012769 display material Substances 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000010295 mobile communication Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000008520 organization Effects 0.000 description 1
- 230000006403 short-term memory Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- 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
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0241—Advertisements
- G06Q30/0251—Targeted advertisements
- G06Q30/0269—Targeted advertisements based on user profile or attribute
- G06Q30/0271—Personalized advertisement
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Databases & Information Systems (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Accounting & Taxation (AREA)
- Development Economics (AREA)
- Finance (AREA)
- Strategic Management (AREA)
- General Physics & Mathematics (AREA)
- Entrepreneurship & Innovation (AREA)
- General Business, Economics & Management (AREA)
- Marketing (AREA)
- Economics (AREA)
- Game Theory and Decision Science (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a material pushing method and device, a storage medium and electronic equipment, and belongs to the field of artificial intelligence. Wherein the method comprises the following steps: acquiring historical behavior information of a target user aiming at a plurality of entities; according to the historical behavior information, a trend prediction model is adopted to output trend values of the target user for a plurality of entity types, wherein the trend prediction model comprises a multiplexer; and selecting a target type with the highest tendency value from the entity types, and pushing the promotion materials of the target type to the target user. According to the invention, the technical problem of low efficiency of predicting the user behavior tendency in the related technology is solved, the quick and accurate prediction of the user tendency to each type of entity is realized, the accurate pushing of the materials is realized, and the pushing amount of the garbage materials is reduced.
Description
Technical Field
The invention relates to the field of artificial intelligence, in particular to a pushing method and device of materials, a storage medium and electronic equipment.
Background
In the related art, consumer brand tendencies are an important issue in the field of CRM (customer relationship management), and consumer tendencies of brands are generally measured by purchasing behavior, as well as understanding and predicting the tendencies or favorites of CRM members for each brand of product. In many cases of brands, a large number of consumers do not have purchasing behavior on the brands, and the brand tendencies cannot be directly counted according to order data, so that modeling is required based on consumer behavior data, and the purchasing probability of each brand is predicted to represent the brand tendencies. In the CRM advertising platform, a machine learning model can be built to predict consumer brand tendencies by utilizing behavior log data of users' browsing and clicking on various brands of advertisements.
In the related art, logistic Regression (logistic regression, LR) is a commonly used machine learning classification model, and because behavior log data of clicking advertisements by a user of a CRM platform mainly contains various ids related to the advertisements, the ids tend to be sparse, and LR has strong robustness to sparse input processing, the model is commonly used in modeling tasks in the advertisement field. And LR is a linear model, and the model has simple structure, fast training and interpretability. Logistic Regression logistic regression remains essentially a simple linear model, with less fitting capability than deep learning models, and it can only capture user behavior information for advertisements, but cannot capture user behavior information for advertisements in time series.
In the related art, the long-short-term memory (Long Short Term Memory, LSTM) network is a special recurrent neural network (Recurrent Neural Network, RNN) model, and its special structural design makes it possible to avoid long-term dependency, and to remember that the information at a very early time is the default behavior of LSTM, without paying a great cost for this. LSTM is effective in capturing sequence information. In the predicted scene of brand tendencies, the user has a time sequence relation with the behavior information of the advertisement, namely, the user belongs to the sequence information. In LSTM training, they are difficult to train because of the large memory bandwidth required, which is not friendly to hardware design. Recursion is inherently non-parallelizable and therefore limits the acceleration of parallel computation by GPUs and the like, which is a very large limitation for model training, especially system tuning.
In view of the above problems in the related art, no effective solution has been found yet.
Disclosure of Invention
The embodiment of the invention provides a pushing method and device of materials, a storage medium and electronic equipment.
According to an aspect of the embodiments of the present application, there is provided a pushing method of a material, including: acquiring historical behavior information of a target user aiming at a plurality of entities; according to the historical behavior information, a trend prediction model is adopted to output trend values of the target user for a plurality of entity types, wherein the trend prediction model comprises a multiplexer; and selecting a target type with the highest tendency value from the entity types, and pushing the promotion materials of the target type to the target user.
Further, before outputting the tendency values of the target user to the plurality of entity types by using the tendency prediction model according to the attribute information, the method further comprises: acquiring first sample data of the target user, wherein the sample data comprises a material click record; and training an initial model by adopting the first sample data to obtain the trend prediction model.
Further, training an initial model using the first sample data to obtain the trend prediction model, including: extracting a plurality of types of ID information in the sample data, wherein each type of ID information corresponds to a sample feature of one dimension; encoding the plurality of types of ID information by adopting a plurality of channels of transformers, wherein each channel of transformers corresponds to an ID sequence formed by one type of ID information; and aggregating the coding results of the multipath transformers, and training to obtain the trend prediction model based on the coding results and label data, wherein the label data is used for representing whether a sample user purchases a corresponding sample material.
Further, encoding the plurality of types of ID information using a plurality of transgenes includes: iteratively encoding the ID information by adopting a plurality of conversion layers for each of the plurality of paths of convertors, and outputting hidden variables of the ID information, wherein the conversion layers comprise an encoder and a decoder; and each Layer of conversion Layer takes the hidden variable output by the previous Layer as input, obtains the hidden variable of the current Layer after passing through a Multi-head self-attention network Multi-Head Self Attention and a Layer normalization network Layer Norm, and inputs the hidden variable into the next Layer of conversion Layer until the last Layer of conversion Layer.
Further, aggregating the encoded results of the multipath transformers comprises: carrying out maximum pooling treatment and average pooling treatment on the coding result of each path of transformer to respectively obtain a first pooling result and a second pooling result; and splicing the first pooling result and the second pooling result which are respectively output by the plurality of paths of transformers to obtain the coding result.
Further, training the trend prediction model based on the encoding results and tag data includes: the coding result is processed by adopting a full link layer Fully Connected Layer, a batch normalization layer Batch Normal Layer and a softmax logistic regression function in sequence, so that an output result is obtained; fitting the output result pred by using the following loss function loss, and training to obtain the tendency prediction model: loss= -labels x log (pred) - (1-labels) x log (1-pred); and labels is the tag data.
Further, after pushing the targeted type of promotional material to the targeted user, the method further comprises: the current time is taken as the starting time to collect second sample data generated by the target user in a preset time period; updating the trend prediction model based on the second sample data.
According to another aspect of the embodiments of the present application, there is also provided a pushing device for materials, including: the acquisition module is used for acquiring historical behavior information of a target user for a plurality of entities; the output module is used for outputting the tendency values of the target user to a plurality of entity types by adopting a tendency prediction model according to the historical behavior information, wherein the tendency prediction model comprises a multiplexer; and the pushing module is used for selecting a target type with the highest tendency value from the plurality of entity types and pushing the popularization materials of the target type to the target user.
Further, the apparatus further comprises: the acquisition module is used for acquiring first sample data of the target user before the output module outputs the tendency values of the target user to a plurality of entity types by adopting a tendency prediction model according to attribute information, wherein the sample data comprises a material click record; and the training module is used for training an initial model by adopting the first sample data to obtain the trend prediction model.
Further, the training module includes: an extracting unit, configured to extract a plurality of types of ID information in the sample data, where each type of ID information corresponds to a sample feature of one dimension; the coding unit is used for coding the plurality of types of ID information by adopting a plurality of channels of transformers, wherein each channel of transformers corresponds to an ID sequence formed by one type of ID information; and the training unit is used for aggregating the coding results of the multipath transformers, and training to obtain the trend prediction model based on the coding results and label data, wherein the label data is used for representing whether a sample user purchases a corresponding sample material or not.
Further, the encoding unit includes: an encoding subunit, configured to iteratively encode the ID information by using a plurality of conversion layers for each of the multiple paths of convertors, and output hidden variables of the ID information, where the conversion layers include an encoder and a decoder; and each Layer of conversion Layer takes the hidden variable output by the previous Layer as input, obtains the hidden variable of the current Layer after passing through a Multi-head self-attention network Multi-Head Self Attention and a Layer normalization network Layer Norm, and inputs the hidden variable into the next Layer of conversion Layer until the last Layer of conversion Layer.
Further, the training unit includes: the first processing subunit is used for carrying out maximum pooling processing and average pooling processing on the coding result of each path of transformation former to respectively obtain a first pooling result and a second pooling result; and the splicing subunit is used for splicing the first pooling result and the second pooling result which are respectively output by the multipath transformers to obtain the coding result.
Further, the training unit includes: the second processing subunit is configured to process the encoding result by sequentially adopting a full link layer Fully Connected Layer, a batch normalization layer Batch Normal Layer and a softmax logistic regression function to obtain an output result; the training subunit is used for fitting the output result pred by adopting the following loss function loss, and training to obtain the tendency prediction model: loss= -labels x log (pred) - (1-labels) x log (1-pred); and labels is the tag data.
Further, the apparatus further comprises: the collection module is used for collecting second sample data generated by the target user in a preset duration forwards by taking the current time as the starting time after the pushing module pushes the promotion materials of the target type to the target user; and the updating module is used for updating the trend prediction model based on the second sample data.
According to another aspect of the embodiments of the present application, there is also provided a storage medium including a stored program that performs the steps described above when running.
According to another aspect of the embodiments of the present application, there is also provided an electronic device, including a processor, a communication interface, a memory, and a communication bus, where the processor, the communication interface, and the memory complete communication with each other through the communication bus; wherein: a memory for storing a computer program; and a processor for executing the steps of the method by running a program stored on the memory.
Embodiments of the present application also provide a computer program product comprising instructions which, when run on a computer, cause the computer to perform the steps of the above method.
According to the invention, the historical behavior information of the target user for a plurality of entities is obtained, the trend values of the target user for a plurality of entity types are output by adopting the trend prediction model according to the historical behavior information, wherein the trend prediction model comprises a multiplexer transducer, the target type with the highest trend value is selected from the plurality of entity types, the promotion materials of the target type are pushed to the target user, the training time and the prediction speed of the model can be reduced by using the multiplexer transducer to model the behavior information of the user, the technical problem of low efficiency of predicting the behavior trend of the user in the related technology is solved, the rapid and accurate prediction of the user on the trend of each entity type is realized, the accurate pushing of the materials is realized, and the pushing quantity of the garbage materials is reduced.
Drawings
The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this application, illustrate embodiments of the invention and together with the description serve to explain the invention and do not constitute a limitation on the invention. In the drawings:
FIG. 1 is a block diagram of the hardware architecture of a server according to an embodiment of the present invention;
FIG. 2 is a flow chart of a method of pushing material according to an embodiment of the invention;
FIG. 3 is a schematic diagram of pre-training of the emmbedding in an embodiment of the invention;
FIG. 4 is a diagram of a network architecture of a multipath transducer in accordance with an embodiment of the present invention;
FIG. 5 is a block diagram of a pushing device for material according to an embodiment of the present invention;
fig. 6 is a block diagram of an electronic device embodying an embodiment of the present invention.
Detailed Description
In order to make the present application solution better understood by those skilled in the art, the following description will be made in detail and with reference to the accompanying drawings in the embodiments of the present application, it is apparent that the described embodiments are only some embodiments of the present application, not all embodiments. All other embodiments, which can be made by one of ordinary skill in the art based on the embodiments herein without making any inventive effort, shall fall within the scope of the present application. It should be noted that, in the case of no conflict, the embodiments and features in the embodiments may be combined with each other.
It should be noted that the terms "first," "second," and the like in the description and claims of the present application and the above figures are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used may be interchanged where appropriate such that embodiments of the present application described herein may be implemented in sequences other than those illustrated or otherwise described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.
Example 1
The method embodiment provided in the first embodiment of the present application may be executed in a server, a computer, or a similar computing device. Taking the operation on a server as an example, fig. 1 is a block diagram of a hardware structure of a server according to an embodiment of the present invention. As shown in fig. 1, the server 10 may include one or more (only one is shown in fig. 1) processors 102 (the processor 102 may include, but is not limited to, a microprocessor MCU or a processing device such as a programmable logic device FPGA) and a memory 104 for storing data, and optionally, a transmission device 106 for communication functions and an input-output device 108. It will be appreciated by those skilled in the art that the structure shown in fig. 1 is merely illustrative, and is not intended to limit the structure of the server described above. For example, the server 10 may also include more or fewer components than shown in FIG. 1, or have a different configuration than shown in FIG. 1.
The memory 104 may be used to store server programs, such as software programs and modules of application software, such as a server program corresponding to a pushing method of a material in an embodiment of the present invention, and the processor 102 executes the server program stored in the memory 104, thereby executing various functional applications and data processing, that is, implementing the method described above. Memory 104 may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory 104 may further include memory remotely located with respect to the processor 102, which may be connected to the server 10 via a network. Examples of such networks include, but are not limited to, the internet, intranets, local area networks, mobile communication networks, and combinations thereof.
The transmission means 106 is arranged to receive or transmit data via a network. The specific example of the network described above may include a wireless network provided by a communication provider of the server 10. In one example, the transmission device 106 includes a network adapter (Network Interface Controller, simply referred to as NIC) that can connect to other network devices through a base station to communicate with the internet. In one example, the transmission device 106 may be a Radio Frequency (RF) module, which is used to communicate with the internet wirelessly.
In this embodiment, a method for pushing a material is provided, and fig. 2 is a flowchart of a method for pushing a material according to an embodiment of the present invention, as shown in fig. 2, where the flowchart includes the following steps:
step S202, historical behavior information of a target user for a plurality of entities is obtained;
in this embodiment, the entity may be any object that a user may click on a client such as a computer, a mobile phone, etc., for example, online merchandise, news, video, etc., where the entity is illustrated as an online merchandise, and the historical behavior information may be behavior information of the user clicking on the entity, searching for the entity, collecting, sharing the entity, etc.
Step S204, a trend prediction model is adopted to output trend values of a target user to a plurality of entity types according to the historical behavior information, wherein the trend prediction model comprises a multiplexer;
the entity types in this embodiment may be classified according to brands, sizes, audience groups, providers, platforms, etc. of the entity, for example, online commodities may be classified into multiple commodity types according to brands, each commodity type corresponds to a brand, news may be classified into multiple news types according to publishers, each news type corresponds to a news publishing organization, etc. In this embodiment, a brand of a commodity is taken as an entity type as an example.
In this embodiment, the multiplexer transformers correspond to characteristic information of historical behavior information in multiple dimensions, and each channel of transformers corresponds to characteristic information in one dimension in the processing process.
Step S206, selecting a target type with the highest tendency value from a plurality of entity types, and pushing promotion materials of the target type to a target user;
optionally, the promotion materials can be promotion links, pictures, information and the like, such as display materials of software open screens, popup materials, and inserting materials (such as advertisements and the like) in the middle of videos.
Through the steps, the historical behavior information of the target user aiming at a plurality of entities is obtained, the trend values of the target user on a plurality of entity types are output according to the historical behavior information by adopting a trend prediction model, wherein the trend prediction model comprises a multiplexer transducer, the target type with the highest trend value is selected from the plurality of entity types, the promotion materials of the target type are pushed to the target user, the training time and the prediction speed of the model can be reduced by using the multiplexer transducer to model the behavior information of the user, the technical problem that the efficiency of predicting the behavior tendency of the user in the related technology is low is solved, the rapid and accurate prediction of the user on the tendency of each entity type is realized, the accurate pushing of the materials is further realized, and the pushing quantity of the garbage materials is reduced.
In one implementation manner of the present embodiment, before the trend prediction model is used to output the trend values of the target user for the plurality of entity types according to the attribute information, the method further includes:
s11, acquiring first sample data of the target user, wherein the sample data comprises a material click record;
the target user has purchasing behavior for different brands, the advertisement click history record of the user in a period window is used as a training data set, and the clicked advertisement information comprises material id, advertisement id, product category id, advertiser industry id and the like, and the times of clicking the advertisement by the user in the same day (taking day as a sample period as an example). Training a multipath converter brand purchase estimation model under the scene, and predicting the brand purchase probability of the users according to the advertisement click records of the users. The first sample data may be converted into ID type data, including: the creativity_id belongs to the advertising creative id to which the material belongs; ad_id, the id of the advertisement to which the material belongs; product_id, the id of the product advertised in the advertisement; product_category, category id of the product advertised in the advertisement; an advertisement_id, an advertiser id; an industry, an id of the industry to which the advertiser belongs.
Optionally, in addition to the material click record, the data in the material exposure record, the material transaction record, the material search record and the material browse record can be used as samples.
And S12, training an initial model by adopting the first sample data to obtain the trend prediction model.
In an example based on the present embodiment, training an initial model using the first sample data to obtain the trend prediction model includes: extracting a plurality of types of ID information in the sample data, wherein each type of ID information corresponds to a sample feature of one dimension; encoding the plurality of types of ID information by adopting a plurality of channels of transformers, wherein each channel of transformers corresponds to an ID sequence formed by one type of ID information; and aggregating the coding results of the multipath transformers, and training to obtain the trend prediction model based on the coding results and label data, wherein the label data is used for representing whether a sample user purchases a corresponding sample material.
Fig. 3 is a schematic diagram of pre-training the text in the embodiment of the present invention, where the text is pre-trained by the training layer, and the ids (e.g., the create_id, the ad_id, and the product_id) related to the advertisements in each text are pre-trained by using a transducer, and the first sample data is used as a text coding feature vector.
In some embodiments, encoding the plurality of types of ID information using a multipath transducer comprises: iteratively encoding the ID information by adopting a plurality of conversion layers for each of the plurality of paths of convertors, and outputting hidden variables of the ID information, wherein the conversion layers comprise an encoder and a decoder; and each Layer of conversion Layer takes the hidden variable output by the previous Layer as input, obtains the hidden variable of the current Layer after passing through a Multi-head self-attention network Multi-Head Self Attention and a Layer normalization network Layer Norm, and inputs the hidden variable into the next Layer of conversion Layer until the last Layer of conversion Layer.
The transducer of this embodiment is composed of 6 encoders (encoders) and 6 decoders (decoders), the encoders being composed of n=6 identical layers, each Layer being composed of two sub-layers, multi-Head Self Attention and Layer Norm, respectively. Wherein residual connection and normalization are added to each sub-layer. The structure of the Decoder and the Encoder is almost similar, but the encoding can be calculated in parallel by one more sub-layer of the position, and all the encodings can be done at a time, but the decoding does not solve all the sequences at a time, but one by one like rnn, because the input of the last position is used as the query of the position. In the transducer, the relevant id of each advertisement is assumed to be a word, the user click sequence is taken as a sentence, the problem is converted into the text classification problem of NLP, and the 12-layer transducer is used for encoding the advertisement id sequence in the embodiment. Each Layer of transformer takes hidden variables of the upper Layer as input, and finally outputs hidden variables with higher interaction through Multi-Head Self Attention, layer Norm and the like.
In some embodiments, aggregating the encoded results of the multipath transformers comprises: carrying out maximum pooling treatment and average pooling treatment on the coding result of each path of transformer to respectively obtain a first pooling result and a second pooling result; and splicing the first pooling result and the second pooling result which are respectively output by the plurality of paths of transformers to obtain the coding result.
In some embodiments, training the trend prediction model based on the encoding results and tag data includes: the coding result is processed by adopting a full link layer Fully Connected Layer, a batch normalization layer Batch Normal Layer and a softmax logistic regression function in sequence, so that an output result is obtained; fitting the output result pred by using the following loss function loss, and training to obtain the tendency prediction model: loss= -labels x log (pred) - (1-labels) x log (1-pred); and labels is the tag data.
Fig. 4 is a network structure diagram of a multipath Transformer according to an embodiment of the present invention, and a flow based on the network structure includes:
passing ids (for example, a create_id, an ad_id, and a product_id) through a pre-trained transducer layer;
performing maxpooling and avgpooling on the output of the transducer layer of each id sequence;
splicing the pooling results of each id together and inputting the pooling results to a full link layer;
through one batch normalization layer;
finally, through the full connection layer, the softmax is connected as the output of the model;
the cross entropy is used as a loss function of the model as follows:
pred＝sigmoid(y)；
loss＝-labels*log(pred)-(1-labels)*log(1-pred)；
wherein pred is the output of the model, labels is the label of whether the training set user really purchases on each brand, specific labels are one-dimensional vectors, the length is the number of brands, each bit has a value of {0,1},1 indicates that the product is purchased, and 0 is not purchased.
In one implementation manner of this embodiment, after pushing the promotion materials of the target type to the target user, the method further includes: the current time is taken as the starting time to collect second sample data generated by the target user in a preset time period; updating the trend prediction model based on the second sample data.
The sample data is updated based on the keep-alive period, so that the trained trend prediction model can be adaptively adjusted along with the habit period of the target user, and the finally predicted and pushed popularization materials are fresh and accurate, and aesthetic fatigue is avoided.
And modeling consumer behaviors by using multiple paths of transformers, expressing purchase predictions of different brands by using multiple classification models, constructing a unified model, reducing model training time and sharing sample data. Training a multipath converter brand purchase estimation model under the scene, and predicting the brand purchase probability of users according to advertisement click records of the users.
From the description of the above embodiments, it will be clear to a person skilled in the art that the method according to the above embodiments may be implemented by means of software plus the necessary general hardware platform, but of course also by means of hardware, but in many cases the former is a preferred embodiment. Based on such understanding, the technical solution of the present invention may be embodied essentially or in a part contributing to the prior art in the form of a software product stored in a storage medium (e.g. ROM/RAM, magnetic disk, optical disk) comprising instructions for causing a terminal device (which may be a mobile phone, a computer, a server, or a network device, etc.) to perform the method according to the embodiments of the present invention.
Example 2
In this embodiment, a material pushing device is further provided, which is used to implement the foregoing embodiments and preferred embodiments, and the description is omitted herein. As used below, the term "module" may be a combination of software and/or hardware that implements a predetermined function. While the means described in the following embodiments are preferably implemented in software, implementation in hardware, or a combination of software and hardware, is also possible and contemplated.
Fig. 5 is a block diagram of a pushing device for materials according to an embodiment of the present invention, as shown in fig. 5, the device includes: an acquisition module 50, an output module 52, a push module 54, wherein,
an obtaining module 50, configured to obtain historical behavior information of a target user for a plurality of entities;
an output module 52, configured to output, according to the historical behavior information, tendency values of the target user for a plurality of entity types using a tendency prediction model, where the tendency prediction model includes a multiplexer;
and the pushing module 54 is configured to select a target type with the highest tendency value from the plurality of entity types, and push the promotion materials of the target type to the target user.
Optionally, the apparatus further includes: the acquisition module is used for acquiring first sample data of the target user before the output module outputs the tendency values of the target user to a plurality of entity types by adopting a tendency prediction model according to attribute information, wherein the sample data comprises a material click record; and the training module is used for training an initial model by adopting the first sample data to obtain the trend prediction model.
Optionally, the training module includes: an extracting unit, configured to extract a plurality of types of ID information in the sample data, where each type of ID information corresponds to a sample feature of one dimension; the coding unit is used for coding the plurality of types of ID information by adopting a plurality of channels of transformers, wherein each channel of transformers corresponds to an ID sequence formed by one type of ID information; and the training unit is used for aggregating the coding results of the multipath transformers, and training to obtain the trend prediction model based on the coding results and label data, wherein the label data is used for representing whether a sample user purchases a corresponding sample material or not.
Optionally, the encoding unit includes: an encoding subunit, configured to iteratively encode the ID information by using a plurality of conversion layers for each of the multiple paths of convertors, and output hidden variables of the ID information, where the conversion layers include an encoder and a decoder; and each Layer of conversion Layer takes the hidden variable output by the previous Layer as input, obtains the hidden variable of the current Layer after passing through a Multi-head self-attention network Multi-Head Self Attention and a Layer normalization network Layer Norm, and inputs the hidden variable into the next Layer of conversion Layer until the last Layer of conversion Layer.
Optionally, the training unit includes: the first processing subunit is used for carrying out maximum pooling processing and average pooling processing on the coding result of each path of transformation former to respectively obtain a first pooling result and a second pooling result; and the splicing subunit is used for splicing the first pooling result and the second pooling result which are respectively output by the multipath transformers to obtain the coding result.
Optionally, the training unit includes: the second processing subunit is configured to process the encoding result by sequentially adopting a full link layer Fully Connected Layer, a batch normalization layer Batch Normal Layer and a softmax logistic regression function to obtain an output result; the training subunit is used for fitting the output result pred by adopting the following loss function loss, and training to obtain the tendency prediction model: loss= -labels x log (pred) - (1-labels) x log (1-pred); and labels is the tag data.
Optionally, the apparatus further includes: the collection module is used for collecting second sample data generated by the target user in a preset duration forwards by taking the current time as the starting time after the pushing module pushes the promotion materials of the target type to the target user; and the updating module is used for updating the trend prediction model based on the second sample data.
It should be noted that each of the above modules may be implemented by software or hardware, and for the latter, it may be implemented by, but not limited to: the modules are all located in the same processor; alternatively, the above modules may be located in different processors in any combination.
Example 3
An embodiment of the invention also provides a storage medium having a computer program stored therein, wherein the computer program is arranged to perform the steps of any of the method embodiments described above when run.
Alternatively, in the present embodiment, the above-described storage medium may be configured to store a computer program for performing the steps of:
s1, acquiring historical behavior information of a target user for a plurality of entities;
s2, outputting tendency values of the target user to a plurality of entity types by adopting a tendency prediction model according to the historical behavior information, wherein the tendency prediction model comprises a multiplexer;
s3, selecting a target type with the highest tendency value from the entity types, and pushing popularization materials of the target type to the target user.
Alternatively, in the present embodiment, the storage medium may include, but is not limited to: a usb disk, a Read-Only Memory (ROM), a random access Memory (Random Access Memory, RAM), a removable hard disk, a magnetic disk, or an optical disk, or other various media capable of storing a computer program.
An embodiment of the invention also provides an electronic device comprising a memory having stored therein a computer program and a processor arranged to run the computer program to perform the steps of any of the method embodiments described above.
Optionally, the electronic device may further include a transmission device and an input/output device, where the transmission device is connected to the processor, and the input/output device is connected to the processor.
Alternatively, in the present embodiment, the above-described processor may be configured to execute the following steps by a computer program:
s1, acquiring historical behavior information of a target user for a plurality of entities;
s2, outputting tendency values of the target user to a plurality of entity types by adopting a tendency prediction model according to the historical behavior information, wherein the tendency prediction model comprises a multiplexer;
s3, selecting a target type with the highest tendency value from the entity types, and pushing popularization materials of the target type to the target user.
Alternatively, specific examples in this embodiment may refer to examples described in the foregoing embodiments and optional implementations, and this embodiment is not described herein.
Fig. 6 is a block diagram of an electronic device according to an embodiment of the present invention, as shown in fig. 6, including a processor 61, a communication interface 62, a memory 63, and a communication bus 64, where the processor 61, the communication interface 62, and the memory 63 communicate with each other through the communication bus 64, and the memory 63 is used for storing a computer program; a processor 61 for executing programs stored on a memory 63.
The foregoing embodiment numbers of the present application are merely for describing, and do not represent advantages or disadvantages of the embodiments.
In the foregoing embodiments of the present application, the descriptions of the embodiments are emphasized, and for a portion of this disclosure that is not described in detail in this embodiment, reference is made to the related descriptions of other embodiments.
In the several embodiments provided in the present application, it should be understood that the disclosed technology content may be implemented in other manners. The above-described embodiments of the apparatus are merely exemplary, and the division of the units, such as the division of the units, is merely a logical function division, and may be implemented in another manner, for example, multiple units or components may be combined or may be integrated into another system, or some features may be omitted, or not performed. Alternatively, the coupling or direct coupling or communication connection shown or discussed with each other may be through some interfaces, units or modules, or may be in electrical or other forms.
The units described as separate units may or may not be physically separate, and units shown as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
In addition, each functional unit in each embodiment of the present application may be integrated in one processing unit, or each unit may exist alone physically, or two or more units may be integrated in one unit. The integrated units may be implemented in hardware or in software functional units.
The integrated units, if implemented in the form of software functional units and sold or used as stand-alone products, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present application may be embodied in essence or a part contributing to the prior art or all or part of the technical solution in the form of a software product stored in a storage medium, including several instructions to cause a computer device (which may be a personal computer, a server or a network device, etc.) to perform all or part of the steps of the methods described in the embodiments of the present application. And the aforementioned storage medium includes: a U-disk, a Read-Only Memory (ROM), a random access Memory (RAM, random Access Memory), a removable hard disk, a magnetic disk, or an optical disk, or other various media capable of storing program codes.
The foregoing is merely a preferred embodiment of the present application and it should be noted that modifications and adaptations to those skilled in the art may be made without departing from the principles of the present application and are intended to be comprehended within the scope of the present application.
Claims (7)
1. The pushing method of the material is characterized by comprising the following steps:
acquiring historical behavior information of a target user aiming at a plurality of entities;
according to the historical behavior information, a trend prediction model is adopted to output trend values of the target user for a plurality of entity types, wherein the trend prediction model comprises a multiplexer;
selecting a target type with the highest tendency value from the plurality of entity types, and pushing popularization materials of the target type to the target user;
wherein before the trend prediction model is adopted to output the trend values of the target user to a plurality of entity types according to the historical behavior information, the method further comprises: acquiring first sample data of the target user, wherein the sample data comprises a material click record; training an initial model by adopting the first sample data to obtain the trend prediction model; wherein training an initial model using the first sample data to obtain the trend prediction model includes: extracting a plurality of types of ID information in the sample data, wherein each type of ID information corresponds to a sample feature of one dimension; encoding the plurality of types of ID information by adopting a plurality of channels of transformers, wherein each channel of transformers corresponds to an ID sequence formed by one type of ID information; aggregating the coding results of the multipath transformers, and training to obtain the trend prediction model based on the coding results and tag data, wherein the tag data is used for representing whether a sample user purchases a corresponding sample material or not; wherein training the trend prediction model based on the encoding result and tag data comprises: the coding result is processed by adopting a full link layer Fully Connected Layer, a batch normalization layer Batch Normal Layer and a softmax logistic regression function in sequence, so that an output result is obtained; fitting the output result using the following loss function lossTraining to obtain the trend prediction model：/>The method comprises the steps of carrying out a first treatment on the surface of the Wherein (1)>Is the tag data.
2. The method of claim 1, wherein encoding the plurality of types of ID information using a plurality of transgers comprises:
iteratively encoding the ID information by adopting a plurality of conversion layers for each of the plurality of paths of convertors, and outputting hidden variables of the ID information, wherein the conversion layers comprise an encoder and a decoder;
and each Layer of conversion Layer takes the hidden variable output by the previous Layer as input, obtains the hidden variable of the current Layer after passing through a Multi-head self-attention network Multi-Head Self Attention and a Layer normalization network Layer Norm, and inputs the hidden variable into the next Layer of conversion Layer until the last Layer of conversion Layer.
3. The method of claim 1, wherein aggregating the encoded results of the multipath transformers comprises:
carrying out maximum pooling treatment and average pooling treatment on the coding result of each path of transformer to respectively obtain a first pooling result and a second pooling result;
and splicing the first pooling result and the second pooling result which are respectively output by the plurality of paths of transformers to obtain the coding result.
4. The method of claim 1, wherein after pushing the targeted type of promotional material to the targeted user, the method further comprises:
the current time is taken as the starting time to collect second sample data generated by the target user in a preset time period;
updating the trend prediction model based on the second sample data.
5. A pushing device for a material, comprising:
the acquisition module is used for acquiring historical behavior information of a target user for a plurality of entities;
the output module is used for outputting the tendency values of the target user to a plurality of entity types by adopting a tendency prediction model according to the historical behavior information, wherein the tendency prediction model comprises a multiplexer;
the pushing module is used for selecting a target type with the highest tendency value from the plurality of entity types and pushing popularization materials of the target type to the target user;
wherein the apparatus further comprises: the acquisition module is used for acquiring first sample data of the target user before the output module outputs tendency values of the target user to a plurality of entity types by adopting a tendency prediction model according to the historical behavior information, wherein the sample data comprises a material click record; the training module is used for training an initial model by adopting the first sample data to obtain the trend prediction model; wherein, training module includes: an extracting unit, configured to extract a plurality of types of ID information in the sample data, where each type of ID information corresponds to a sample feature of one dimension; the coding unit is used for coding the plurality of types of ID information by adopting a plurality of channels of transformers, wherein each channel of transformers corresponds to an ID sequence formed by one type of ID information; the training unit is used for aggregating the coding results of the multipath transformers, and training the trend prediction model based on the coding results and label data, wherein the label data is used for representing whether a sample user purchases a corresponding sample material or not; wherein the training unit comprises: the second processing subunit is configured to process the encoding result by sequentially adopting a full link layer Fully Connected Layer, a batch normalization layer Batch Normal Layer and a softmax logistic regression function to obtain an output result; training subunit for fitting the output result using the following loss function lossTraining to obtain the trend prediction model:
the method comprises the steps of carrying out a first treatment on the surface of the Wherein (1)>Is the tag data.
6. A storage medium comprising a stored program, wherein the program when run performs the method steps of any of the preceding claims 1 to 4.
7. An electronic device comprises a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory are communicated with each other through the communication bus; wherein:
a memory for storing a computer program;
a processor for executing the method steps of any one of claims 1 to 4 by running a program stored on a memory.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011519601.5A CN112507234B (en) | 2020-12-21 | 2020-12-21 | Material pushing method and device, storage medium and electronic equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011519601.5A CN112507234B (en) | 2020-12-21 | 2020-12-21 | Material pushing method and device, storage medium and electronic equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112507234A CN112507234A (en) | 2021-03-16 |
| CN112507234B true CN112507234B (en) | 2024-03-01 |
Family
ID=74922810
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011519601.5A Active CN112507234B (en) | 2020-12-21 | 2020-12-21 | Material pushing method and device, storage medium and electronic equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112507234B (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| AU2016222401B1 (en) * | 2015-08-31 | 2017-02-23 | Accenture Global Solutions Limited | Intelligent data munging |
| CN110502627A (en) * | 2019-08-28 | 2019-11-26 | 上海海事大学 | A kind of answer generation method based on multilayer Transformer polymerization encoder |
| CN111461841A (en) * | 2020-04-07 | 2020-07-28 | 腾讯云计算（北京）有限责任公司 | Article recommendation method, device, server and storage medium |
| CN111881343A (en) * | 2020-07-07 | 2020-11-03 | Oppo广东移动通信有限公司 | Information pushing method and device, electronic equipment and computer readable storage medium |
| CN111897929A (en) * | 2020-08-04 | 2020-11-06 | 腾讯科技（深圳）有限公司 | Method and device for processing multiple rounds of questions, storage medium and electronic equipment |
-
2020
- 2020-12-21 CN CN202011519601.5A patent/CN112507234B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| AU2016222401B1 (en) * | 2015-08-31 | 2017-02-23 | Accenture Global Solutions Limited | Intelligent data munging |
| CN110502627A (en) * | 2019-08-28 | 2019-11-26 | 上海海事大学 | A kind of answer generation method based on multilayer Transformer polymerization encoder |
| CN111461841A (en) * | 2020-04-07 | 2020-07-28 | 腾讯云计算（北京）有限责任公司 | Article recommendation method, device, server and storage medium |
| CN111881343A (en) * | 2020-07-07 | 2020-11-03 | Oppo广东移动通信有限公司 | Information pushing method and device, electronic equipment and computer readable storage medium |
| CN111897929A (en) * | 2020-08-04 | 2020-11-06 | 腾讯科技（深圳）有限公司 | Method and device for processing multiple rounds of questions, storage medium and electronic equipment |
Non-Patent Citations (1)
| Title |
|---|
| 基于多头自注意力神经网络的购物篮推荐方法;倪维健;郭浩宇;刘彤;曾庆田;;数据分析与知识发现(第Z1期);第68-77页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112507234A (en) | 2021-03-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US20200195737A1 (en) | Clickstream analysis methods and systems related to determining actionable insights relating to a path to purchase | |
| CN110717098B (en) | Meta-path-based context-aware user modeling method and sequence recommendation method | |
| CN109903117B (en) | Knowledge graph processing method and device for commodity recommendation | |
| CN110008408A (en) | A kind of session recommended method, system, equipment and medium | |
| CN112765373B (en) | Resource recommendation method and device, electronic equipment and storage medium | |
| CN112669096B (en) | Object recommendation model training method and device | |
| CN101271559A (en) | Cooperation recommending system based on user partial interest digging | |
| CN112287238B (en) | User characteristic determination method and device, storage medium and electronic equipment | |
| CN112712418B (en) | Method and device for determining recommended commodity information, storage medium and electronic equipment | |
| CN112365283B (en) | Coupon issuing method and device, terminal equipment and storage medium | |
| CN113077276A (en) | Behavior data processing method and system, storage medium and processor | |
| CN106445954A (en) | Business object display method and apparatus | |
| CN112288554B (en) | Commodity recommendation method and device, storage medium and electronic device | |
| CN110189163A (en) | Evaluation method, device, electronic equipment and the storage medium of promotional content | |
| CN111429161B (en) | Feature extraction method, feature extraction device, storage medium and electronic equipment | |
| CN116127184A (en) | Product recommendation method and device, nonvolatile storage medium and electronic equipment | |
| CN112307352A (en) | Content recommendation method, system, device and storage medium | |
| CN116821516B (en) | Resource recommendation method, device, equipment and storage medium | |
| CN108563648B (en) | Data display method and device, storage medium and electronic device | |
| CN112507234B (en) | Material pushing method and device, storage medium and electronic equipment | |
| CN116911926A (en) | Advertisement marketing recommendation method based on data analysis | |
| CN115564532A (en) | Training method and device of sequence recommendation model | |
| CN115618079A (en) | Session recommendation method, device, electronic equipment and storage medium | |
| CN112464087B (en) | Recommendation probability output method and device, storage medium and electronic equipment | |
| CN110418171B (en) | Media resource pushing method and device, storage medium and electronic device |
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 |