Detailed Description
Reference will now be made in detail to exemplary embodiments, examples of which are illustrated in the accompanying drawings. When the following description refers to the accompanying drawings, the same numbers in different drawings refer to the same or similar elements, unless otherwise indicated. The implementations described in the following exemplary examples do not represent all implementations consistent with the application. Rather, they are merely examples of apparatus and methods consistent with aspects of the application as detailed in the accompanying summary.
Example 1
As shown in fig. 1, a flowchart of a method for publishing a network marketing campaign provided in an embodiment of the present disclosure includes:
s11: and receiving marketing activity configuration information sent by the user terminal, wherein the marketing activity configuration information comprises activity template information, activity online time and activity offline time.
S12: based on the marketing campaign configuration information, at least one target marketing strategy is acquired by utilizing a preset dynamic strategy recommendation model, and the preset dynamic strategy recommendation model is realized based on a GMF model and an NCF model.
S13: and sending the target marketing strategy to the user terminal.
It can be understood that, according to the technical scheme provided by the embodiment, a network data acquisition strategy is generated according to user setting, and after a network data acquisition instruction is received, the operation data of the target network is acquired according to the network data acquisition strategy; generating a network diagnosis report according to a preset fault detection strategy based on the operation data of the target network; if the network fault information exists in the network diagnosis report, generating a network optimization scheme according to the network fault information, and optimizing the target network according to the network optimization scheme; generating a network optimization report, automatically completing the network optimization process without manual participation, greatly improving the efficiency of network optimization, saving the cost and effectively ensuring the optimization effect.
Example 2
As shown in fig. 2, another method for publishing a network marketing campaign provided by an embodiment of the present disclosure includes:
s21: training a preset dynamic strategy recommendation model.
S22: and receiving marketing activity configuration information sent by the user terminal, wherein the marketing activity configuration information comprises activity template information, activity online time and activity offline time.
S23: based on the marketing campaign configuration information, at least one target marketing strategy is acquired by utilizing a preset dynamic strategy recommendation model, and the preset dynamic strategy recommendation model is realized based on a GMF model and an NCF model.
S24: and sending the target marketing strategy to the user terminal.
In some alternative implementations, the preset dynamic policy recommendation model includes an input layer and a feature interaction layer, the marketing campaign configuration information includes marketing campaign features, and the S21 training preset dynamic policy recommendation model includes (not shown in the figures):
s211: the input layer acquires the marketing campaign characteristics, and processes the marketing campaign characteristics through the neural network to generate a characteristic matrix.
S212: the feature interaction layer changes the feature matrix into a one-dimensional matrix according to a model algorithm and the full link layer.
S213: the input layer performs normalization processing on the one-dimensional matrix, calculates the matching degree of each marketing strategy to be processed and marketing activity configuration, and takes the marketing strategy to be processed, the configuration degree of which meets the conditions, as a target marketing strategy.
In some alternative implementations, S211 processes the marketing campaign feature via a neural network, and generating the feature matrix includes (not shown):
s211: a denoised vectorized data format is generated based on the marketing campaign characteristics.
S212: the vector length is fixed by the single-hot encoding and densification processes.
In some alternative implementations, S212 alters the feature matrix to a one-dimensional matrix according to a model algorithm and a full link layer includes (not shown):
s2121: the feature combination value is generated by the inner product.
S2122: and respectively adding assigned weights to the values of the feature combinations.
S2123: and multiplying the feature combinations with the added assigned weights.
S2124: and combining the dense vectors at the full link layer to obtain a feature matrix.
In some alternative implementations, the preset dynamic policy recommendation model is trained by the following formula:
the L2 calculating method comprises the following steps:
，/>for loss function->A dot product representing two vectors of size k, k being the degree of the FM algorithm, i, j being the i sample order, d being the end condition, x being the sample data,is the least squares error +.>For target value, & lt + & gt>For the estimated value, W is the network weight, b is the offset, m is the number of samples, +.>For the network layer number->Is a super parameter.
In some alternative implementations, S23 using the preset dynamic policy recommendation model to obtain at least one target marketing strategy includes (not shown):
s231: and inputting user behavior log information, and obtaining the 1 x 4096 single-heat coding feature matrix through an embedding (embedding) layer.
S231: and converting the single-heat coding feature matrix into a feature matrix 4096 which is fixed in dense dimension through a sparse matrix, and transmitting the feature matrix to a feature interaction layer.
S231: and combining through an inner product plate in the feature interaction layer to generate 4096 feature combination values, and distributing the percentages of different marketing features.
S231: and adding two full-connection layers to the output of the feature interaction layer, and densely combining to obtain 1 x 4096 high-order features.
S231: and normalizing all feature extraction matrixes to form a matrix with probability value of 0-1 for each element, and calculating to obtain the target marketing strategy.
It can be understood that, according to the technical scheme provided by the embodiment, a network data acquisition strategy is generated according to user setting, and after a network data acquisition instruction is received, the operation data of the target network is acquired according to the network data acquisition strategy; generating a network diagnosis report according to a preset fault detection strategy based on the operation data of the target network; if the network fault information exists in the network diagnosis report, generating a network optimization scheme according to the network fault information, and optimizing the target network according to the network optimization scheme; generating a network optimization report, automatically completing the network optimization process without manual participation, greatly improving the efficiency of network optimization, saving the cost and effectively ensuring the optimization effect.
Example 3
As shown in fig. 3, the embodiment of the present application further provides a device for publishing a network marketing campaign, which includes:
the configuration information receiving module 31 is configured to receive marketing campaign configuration information sent by the user terminal, where the marketing campaign configuration information includes campaign template information, campaign online time and campaign offline time.
The strategy acquisition module 32 is configured to acquire at least one target marketing strategy by using a preset dynamic strategy recommendation model based on the marketing campaign configuration information, where the preset dynamic strategy recommendation model is implemented based on the GMF model and the NCF model.
And the strategy sending module 33 is used for sending the target marketing strategy to the user terminal.
In some alternative embodiments, the apparatus further comprises:
model training module 34 is used to train a preset dynamic policy recommendation model.
In some alternative embodiments, the preset dynamic policy recommendation model includes an input layer and a feature interaction layer, the marketing campaign configuration information includes marketing campaign features, and the model training module 34 includes:
the campaign feature acquisition sub-module 341 is configured to acquire a marketing campaign feature from the input layer, process the marketing campaign feature through the neural network, and generate a feature matrix.
The feature matrix changing sub-module 342 is configured to change the feature matrix to a one-dimensional matrix according to the model algorithm and the full link layer by the feature interaction layer.
The matrix normalization processing sub-module 343 is configured to normalize the one-dimensional matrix by the input layer, calculate the matching degree of each pending marketing strategy and the marketing campaign configuration, and take the pending marketing strategy with the configuration degree meeting the condition as the target marketing strategy.
In some alternative embodiments, activity feature acquisition sub-module 341 includes:
the data generating unit 3411 is configured to generate a denoised vectorized data format based on the marketing campaign characteristics.
The vector length fixing unit 3412 is configured to fix the vector length by the one-hot encoding and densification process.
In some alternative embodiments, the feature matrix modification sub-module 342 includes:
a feature combination unit 3421 for generating a feature combination value by the inner product.
The weight adjustment unit 3422 is configured to add a specified weight to the values of the feature combinations.
And a feature multiplication unit 3423 for multiplying the feature combinations added with the specified weights by two.
And a vector combining unit 3424, configured to combine the dense vectors at the full link layer to obtain a feature matrix.
In some alternative embodiments, the preset dynamic policy recommendation model is trained by the following formula:
the L2 calculating method comprises the following steps:
，/>for loss function->Representing the dot product of two vectors of size k, k being the degree of the FM algorithm, i, j being the i sample order, d beingThe end condition, x is the sample data,is the least squares error +.>For target value, & lt + & gt>For the estimated value, W is the network weight, b is the offset, m is the number of samples, +.>For the network layer number->Is a super parameter.
In some alternative embodiments, the policy acquisition module 32 includes:
the information obtaining sub-module 321 is configured to input user behavior log information, and obtain a1 x 4096 unique thermal coding feature matrix through the embedding layer.
The feature matrix conversion sub-module 322 is configured to convert the single-hot encoded feature matrix into a feature matrix 4096×256 with a fixed dense dimension through a sparse matrix, and transmit the feature matrix to the feature interaction layer.
The duty allocation sub-module 323 is configured to generate 4096 feature combination values by combining the inner integration plates at the feature interaction layer, and allocate percentages of different marketing features.
The high-order feature acquisition sub-module 324 adds the output of the feature interaction layer to the two full connection layers, and densely combines the two full connection layers to obtain 1×4096 high-order features.
The normalization processing sub-module 325 is configured to normalize the total feature extraction matrix to form a matrix with probability value of 0-1 for each element, and calculate the target marketing strategy.
It can be understood that, according to the technical scheme provided by the embodiment, a network data acquisition strategy is generated according to user setting, and after a network data acquisition instruction is received, the operation data of the target network is acquired according to the network data acquisition strategy; generating a network diagnosis report according to a preset fault detection strategy based on the operation data of the target network; if the network fault information exists in the network diagnosis report, generating a network optimization scheme according to the network fault information, and optimizing the target network according to the network optimization scheme; generating a network optimization report, automatically completing the network optimization process without manual participation, greatly improving the efficiency of network optimization, saving the cost and effectively ensuring the optimization effect.
Example 4
Based on the same technical concept, the embodiment of the application also provides a computer device, which comprises a memory 1 and a processor 2, as shown in fig. 4, the memory 1 stores a computer program, and the processor 2 implements the network marketing campaign issuing method of any one of the above when executing the computer program.
The memory 1 includes at least one type of readable storage medium including flash memory, a hard disk, a multimedia card, a card type memory (e.g., SD or DX memory, etc.), a magnetic memory, a magnetic disk, an optical disk, etc. The memory 1 may in some embodiments be an internal storage unit of a network marketing campaign distribution system, such as a hard disk. The memory 1 may in other embodiments also be an external storage device of a network marketing campaign distribution system, such as a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) Card, a Flash Card (Flash Card), etc. Further, the memory 1 may also include both an internal storage unit and an external storage device of the networked marketing campaign distribution system. The memory 1 may be used not only for storing application software installed in the network marketing campaign distribution system and various types of data, such as codes of the network marketing campaign distribution program, etc., but also for temporarily storing data that has been output or is to be output.
The processor 2 may in some embodiments be a central processing unit (Central Processing Unit, CPU), controller, microcontroller, microprocessor or other data processing chip for running program code or processing data stored in the memory 1, for example executing a web marketing campaign distribution program or the like.
It can be understood that, according to the technical scheme provided by the embodiment, a network data acquisition strategy is generated according to user setting, and after a network data acquisition instruction is received, the operation data of the target network is acquired according to the network data acquisition strategy; generating a network diagnosis report according to a preset fault detection strategy based on the operation data of the target network; if the network fault information exists in the network diagnosis report, generating a network optimization scheme according to the network fault information, and optimizing the target network according to the network optimization scheme; generating a network optimization report, automatically completing the network optimization process without manual participation, greatly improving the efficiency of network optimization, saving the cost and effectively ensuring the optimization effect.
In order to facilitate the reader to understand the technical scheme of the embodiment of the application, the technical details in the scheme are described in detail below through specific examples.
According to the network marketing campaign issuing method and device provided by the embodiment of the application, the dynamic strategy recommendation model is constructed, in the marketing campaign configuration process, the optimal campaign case recommendation meeting the current known parameters is obtained through calculation, and operators can locally adjust or directly and rapidly issue the campaign to channel contacts after selecting the optimal campaign case recommendation. The actual activity page is provided with a buried point program, the server collects index data such as actual click rate, page stay, conversion rate and the like in real time, and feeds the index data back to the model for parameter optimization to form a complete closed loop, so that a more accurate calculation method is provided for next activity case recommendation.
The operator logs in the marketing activity system, starts specific parameter configuration after selecting a template, dynamically calculates known information and performs excellent case recommendation from the first step of configuration, and can select case ending configuration at any node, display complete activity overview according to information built in the excellent case, and further adjust individual parameters to release or directly release the activity to a channel.
Step 1: the operator logs in the marketing activity system, and the template configuration is selected to be published activities, which concretely comprises the following steps:
marketing campaign system: a large number of movable templates are integrated, marketing activities can be rapidly configured and released to target channel contacts, a statistical analysis function is provided, and the activity effect is mastered in real time.
And (3) a movable template: the core competitiveness in the marketing campaign system is divided into lottery drawing class, game class, assistance class, sign-in class, answer class, questionnaire class and the like according to templates commonly used in the Internet, and each class comprises a plurality of movable templates.
Dynamic policy recommendation model: and recommending the excellent case in real time according to parameters configured by operators in the marketing campaign system.
Step 2: after the online time of the first step of activities is configured, excellent activity cases recommended according to known templates and online time information are displayed, the online time of the activities is a first item of content configured after the templates are selected, a dynamic strategy recommendation model recommends excellent cases corresponding to a theme for operators according to whether holidays are covered or not, and optimal recommended content can be changed at any time by combining other seasons, months, chinese zodiac and other information.
Step 3: and (3) subsequently configuring parameters such as activity awards, popularization channels, participation times, winning probability and the like, and recalculating the excellent case arrangement sequence through a dynamic strategy recommendation model and refreshing and displaying every time a new parameter is configured. After the space part of the form of the configuration page is changed, the space part is submitted to a dynamic strategy recommendation model in real time, and other configured contents are synthesized to recommend the current optimal activity case, for example: after the template is selected, the recommended ranking list of the activities displayed on the online date is configured to be TOPA1 to TOPA10, the recommended ranking list of the activities displayed after the distribution channel is continuously configured to be TOPB1 to TOPB10, and the recommended case displayed in the first position is used as a model to calculate the case which is best adapted to the known information and expected to obtain the best marketing effect.
Step 4: and selecting the recommended excellent case, and previewing the overall activity information, and locally adjusting or directly publishing the overall activity information. The selected excellent case is not necessarily ranked first, and the process jumps to the activity strategy preview interface, and at this time, operators can perform partial parameter fine adjustment according to actual marketing requirements, but the process will not continue to recommend.
Fig. 5 shows a functional structure diagram of the network marketing campaign publishing device provided in this embodiment, including configuration of front-end campaign basis, rules and documents, interacting with a dynamic policy recommendation model, performing bottom-layer service support by a campaign case management and core function module of a server, and displaying the final calculation result on a client contact through a ranking list, so that an operator does not need to push back the campaign parameter configuration through marketing policy analysis, thereby effectively reducing an operation threshold, and providing powerful support tools for people with different experience degrees.
Marketing campaign system front-end: the front end comprises an activity foundation, activity rules and activity document settings, the configuration is completed through sequential circulation, and after the configuration is completed, the result can be saved and the next configuration can be performed.
Activity base setting: the method comprises the steps of event online time and event offline time, automatic event offline options of prizes without residual events, event daily frequency limitation, event prompt participation, event popularization channels and release ranges, event users participation, title sharing configuration and the like.
Active rule configuration: including setting the number of winnings, different users can limit the number of winnings, prize release time setting, prize winning prompt setting, information setting required for exchanging the prize, etc.
The movable document setting: activity titles, activity rules, and winning prompts can be edited; bold, italics, color, underline, etc. of the font can be edited.
Dynamic policy recommendation model: based on a plurality of dimensions of activity foundations, rules and document settings, the matching relation between the current strategy parameters and the activity cases is calculated based on a GMF (generalized matrix decomposition algorithm, generalized matrix factorization model) +NCF (a telescopic network calculation model, neural Collaborative Filtering) model, the activity cases with poor effects are filtered continuously, and the improvement of the current configuration is realized from the aspect of activity effects.
The activity management module: the method comprises the steps of historical case management, case combination and comprehensive sequencing, wherein the marketing strategy and result information of the historical cases are maintained, the optimal strategy is calculated through a dynamic recommendation model when an actual operator configures the historical cases, the best marketing expectation is obtained, and the calculated results are arranged and displayed according to the scores.
Core function module: including flow management, campaign rules management, channel management, topic management, contact management, campaign data management, provide a basic capability support for marketing campaign systems.
And (3) flow control: and carrying out management and control on the whole activity configuration flow in order, wherein the activity configuration step comprises basic configuration, rule configuration and document configuration.
Activity rule management: the options and selectable options contained in the activity rules are defined, including standardized indexes such as number of wins, time of prize issuer, prize prompt, etc.
Channel management: and defining optional transmission channels in marketing configuration, including APP, H5, web end and the like.
Theme management: and defining marketing page topics, such as topics of mid-autumn festival, national celebration festival, spring festival, four seasons of spring, summer, autumn and winter, and the like, and replacing the whole set of original picture background after selection.
And (3) contact management: the floor set by the channel of marketing activity release is managed, and the actual position of activity transmission is defined.
Activity data management: burying points on the marketing page, and collecting effect data in real time as a dynamic strategy recommendation model adjustment standard.
Fig. 6 shows a schematic diagram of an optimization flow of a preset dynamic policy recommendation model in an embodiment of the present application, from multiple dimensions of activity basis, rules and document settings, the matching relationship between the current policy parameters and the activity cases is calculated based on the gmf+ncf model, the activity cases with poor effects are continuously filtered, and the current marketing policy is completed by triggering from the perspective of activity effects.
According to the scene demand of excellent case recommendation, a feature matrix is generated through a neural network by extracting and processing the marketing strategy features of an input layer to achieve a feature interaction layer, the multi-dimensional feature matrix is further changed into one dimension according to an algorithm and a full-link layer, the marketing strategy with low adaptation degree is continuously filtered out in the process, the one-dimensional matrix is normalized through a normalization exponential function (softmax), a probability value of the matching degree of the excellent case and the current configuration is generated, and the neural network can adopt FM (Factorization Machine, a machine learning algorithm).
Step 1: and preprocessing marketing strategy information at an input layer to generate a denoised vectorized data format, wherein the vectorized data format has difference in length, fixing the length of the vector through independent thermal coding and densification processing, and transmitting the vector with equal length to a characteristic interaction layer as a characteristic value identifiable by a neural network.
Step 2: the feature interaction layer comprises a model algorithm and a full link layer, feature combination values are generated through inner products, weights a1-an are added to the feature combination values, multiplication processing is carried out on the feature combination values, and dense vectors are combined in the full link layer to obtain high-order features.
Step 3: mse (mean square error) (Microsoft service management engine, microsoft Service Engine) is used as a loss function, a model of DNN is trained through a deep learning optimizer, so that the loss function is gradually reduced, model training is finished when the loss function is stable or is reduced to the lowest point, and the model is saved.
Step 4: the final result is calculated.
The L2 calculating method comprises the following steps:
，/>for loss function->A dot product representing two vectors of size k, k being the degree of the FM algorithm, i, j being the i sample order, d being the end condition, x being the sample data,is the least squares error +.>For target value, & lt + & gt>For the estimated value, W is the network weight, b is the offset, m is the number of samples, +.>For the network layer number->Is a super parameter. So far, the model training process is ended.
Model application flow:
step 1: and inputting log information of user behaviors, and obtaining the 1 x 4096 single-hot coding feature matrix through an embedding layer.
Step 2: and changing the single-heat coding feature matrix into a feature matrix 4096 with fixed dense dimensions through a sparse matrix, and transmitting the feature matrix to a feature interaction layer.
Step 3: in the feature interaction layer, 4096 feature combination values are generated through combination of the inner product plates, and the percentages of different marketing features are distributed and are integrated to be 100%.
And step 4, adding two full-connection layers to the output of the feature interaction layer to obtain 1 x 4096 high-order features by dense combination.
And 5, finally, normalizing all feature extraction matrixes to form a matrix with probability value of 0-1 for each element, and calculating to obtain the marketing strategy conforming to the user.
According to the scheme, a dynamic strategy recommendation model is constructed, the most suitable activity strategy is changed in real time according to configuration parameters of operators, the priority of an adaptive activity is displayed in a ranking list mode, the expected activity effect is calculated, the marketing effect is recorded, the actual data is optimized through the model, the better recommendation effect is ensured, in the marketing activity configuration process, the best activity case recommendation meeting the current known parameters is obtained through calculation, and the operators can locally adjust or directly and rapidly issue the activity to channel contacts after selecting the optimal activity case recommendation. The actual activity page is provided with a buried point program, the server collects index data such as actual click rate, page stay, conversion rate and the like in real time, and feeds the index data back to the model for parameter optimization to form a complete closed loop, so that a more accurate calculation method is provided for next activity case recommendation.
The disclosed embodiments also provide a computer readable storage medium having a computer program stored thereon, which when executed by a processor performs the steps of the network marketing campaign distribution method of the method embodiments described above. Wherein the storage medium may be a volatile or nonvolatile computer readable storage medium.
The computer program product of the network marketing campaign issuing method provided by the embodiment of the present application includes a computer readable storage medium storing program code, and the instructions included in the program code may be used to execute the steps of the network marketing campaign issuing method in the method embodiment, and specifically, refer to the method embodiment and are not repeated herein.
The disclosed embodiments also provide a computer program which, when executed by a processor, implements any of the methods of the previous embodiments. The computer program product may be realized in particular by means of hardware, software or a combination thereof. In an alternative embodiment, the computer program product is embodied as a computer storage medium, and in another alternative embodiment, the computer program product is embodied as a software product, such as a software development kit (Software Development Kit, SDK), or the like.
It is to be understood that the same or similar parts in the above embodiments may be referred to each other, and that in some embodiments, the same or similar parts in other embodiments may be referred to.
It should be noted that in the description of the present application, the terms "first," "second," and the like are used for descriptive purposes only and are not to be construed as indicating or implying relative importance. Furthermore, in the description of the present application, unless otherwise indicated, the meaning of "plurality" means at least two.
Any process or method descriptions in flow charts or otherwise described herein may be understood as representing modules, segments, or portions of code which include one or more executable instructions for implementing specific logical functions or steps of the process, and further implementations are included within the scope of the preferred embodiment of the present application in which functions may be executed out of order from that shown or discussed, including substantially concurrently or in reverse order, depending on the functionality involved, as would be understood by those reasonably skilled in the art of the present application.
It is to be understood that portions of the present application may be implemented in hardware, software, firmware, or a combination thereof. In the above-described embodiments, the various steps or methods may be implemented in software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, may be implemented using any one or combination of the following techniques, as is well known in the art: discrete logic circuits having logic gates for implementing logic functions on data signals, application specific integrated circuits having suitable combinational logic gates, programmable Gate Arrays (PGAs), field Programmable Gate Arrays (FPGAs), and the like.
Those of ordinary skill in the art will appreciate that all or a portion of the steps carried out in the method of the above-described embodiments may be implemented by a program to instruct related hardware, where the program may be stored in a computer readable storage medium, and where the program, when executed, includes one or a combination of the steps of the method embodiments.
In addition, each functional unit in the embodiments of the present application may be integrated in one processing module, or each unit may exist alone physically, or two or more units may be integrated in one module. The integrated modules may be implemented in hardware or in software functional modules. The integrated modules may also be stored in a computer readable storage medium if implemented in the form of software functional modules and sold or used as a stand-alone product.
The above-mentioned storage medium may be a read-only memory, a magnetic disk or an optical disk, or the like.
In the description of the present specification, a description referring to terms "one embodiment," "some embodiments," "examples," "specific examples," or "some examples," etc., means that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the present application. In this specification, schematic representations of the above terms do not necessarily refer to the same embodiments or examples. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.
While embodiments of the present application have been shown and described above, it will be understood that the above embodiments are illustrative and not to be construed as limiting the application, and that variations, modifications, alternatives and variations may be made to the above embodiments by one of ordinary skill in the art within the scope of the application.