CN112348600A - Service promotion object maintenance method and device, computer equipment and storage medium - Google Patents
Service promotion object maintenance method and device, computer equipment and storage medium Download PDFInfo
- Publication number
- CN112348600A CN112348600A CN202011379503.6A CN202011379503A CN112348600A CN 112348600 A CN112348600 A CN 112348600A CN 202011379503 A CN202011379503 A CN 202011379503A CN 112348600 A CN112348600 A CN 112348600A
- Authority
- CN
- China
- Prior art keywords
- promotion
- decision
- model
- decision tree
- sample
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 53
- 238000012423 maintenance Methods 0.000 title claims abstract description 39
- 238000003066 decision tree Methods 0.000 claims abstract description 138
- 238000012549 training Methods 0.000 claims description 37
- 230000006870 function Effects 0.000 claims description 28
- 230000015654 memory Effects 0.000 claims description 17
- 238000004590 computer program Methods 0.000 claims description 15
- 230000000694 effects Effects 0.000 claims description 13
- 238000012545 processing Methods 0.000 claims description 9
- 238000004422 calculation algorithm Methods 0.000 claims description 8
- 230000011218 segmentation Effects 0.000 claims description 4
- 230000002829 reductive effect Effects 0.000 abstract description 4
- 230000008569 process Effects 0.000 description 11
- 238000004364 calculation method Methods 0.000 description 7
- 238000010586 diagram Methods 0.000 description 7
- 238000004891 communication Methods 0.000 description 6
- 230000008901 benefit Effects 0.000 description 4
- 230000036961 partial effect Effects 0.000 description 4
- 230000001737 promoting effect Effects 0.000 description 3
- 238000007405 data analysis Methods 0.000 description 2
- 238000013480 data collection Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 238000012795 verification Methods 0.000 description 2
- 230000008859 change Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000003776 cleavage reaction Methods 0.000 description 1
- 230000001186 cumulative effect Effects 0.000 description 1
- 238000013075 data extraction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 230000000670 limiting effect Effects 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003032 molecular docking Methods 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000003252 repetitive effect Effects 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 230000000717 retained effect Effects 0.000 description 1
- 230000007017 scission Effects 0.000 description 1
- 230000006403 short-term memory Effects 0.000 description 1
- 238000004904 shortening Methods 0.000 description 1
- 238000000638 solvent extraction Methods 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 230000001360 synchronised effect Effects 0.000 description 1
- 230000017105 transposition Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0241—Advertisements
- G06Q30/0242—Determining effectiveness of advertisements
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N5/00—Computing arrangements using knowledge-based models
- G06N5/01—Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound
-
- 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
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/067—Enterprise or organisation modelling
Abstract
The application relates to a method and a device for maintaining a service promotion object, computer equipment and a storage medium, wherein the method comprises the following steps: acquiring release data of the business promotion object in a corresponding promotion platform; inputting the delivery data into a pre-trained decision model, so as to obtain support probabilities for continuing delivering the service promotion objects in a corresponding promotion platform according to the delivery data and a predefined objective function through a plurality of decision trees in the decision maintenance model, and fusing the support probabilities corresponding to the decision trees to output corresponding decision results; and determining whether to continue to put the service promotion object in the corresponding promotion platform according to a decision result output by the decision model, so that the automatic and timely maintenance of the subsequent advertisement putting is realized, a manual and empirical maintenance method is eliminated, and the advertisement putting cost is effectively reduced.
Description
Technical Field
The present application relates to the field of computer technologies, and in particular, to a method and an apparatus for maintaining a service promotion object, a computer device, and a storage medium.
Background
In the prior art, after the advertisement is put, the advertisement is maintained subsequently by an advertisement optimizer, and the advertisement is closed or kept for continuous propaganda and promotion according to the personal experience of the advertisement optimizer.
However, the follow-up maintenance mode of the advertisement is determined by depending on personal experience, which easily causes untimely advertisement maintenance, for example, the advertisement with low promotion efficiency is not closed in time, and it is difficult to effectively reduce the advertisement putting cost.
Disclosure of Invention
In view of the foregoing, it is necessary to provide a service promotion object maintenance method, apparatus, computer device, and storage medium for solving the above technical problem.
A method for maintaining a business promotion object, the method comprising:
acquiring release data of the business promotion object in a corresponding promotion platform; the release data comprises data of a plurality of preset indexes, and the preset indexes reflect the promotion effect of the service promotion objects in the corresponding promotion platforms;
inputting the delivery data into a pre-trained decision model, so as to obtain support probabilities for continuing delivering the service promotion objects in a corresponding promotion platform according to the delivery data and a predefined objective function through a plurality of decision trees in the decision maintenance model, and fusing the support probabilities corresponding to the decision trees to output corresponding decision results;
and determining whether to continue to put the service promotion object in the corresponding promotion platform according to a decision result output by the decision model.
Optionally, the decision result is a sum of a plurality of support probabilities; the determining whether to continue to put the service promotion object in the corresponding promotion platform according to the decision result output by the decision model includes:
when the sum of the support probabilities is greater than or equal to a preset threshold value, determining to continuously release the service promotion object on the corresponding promotion platform;
and when the sum of the support probabilities is smaller than the preset threshold value, generating information for stopping continuous delivery in the corresponding promotion platform aiming at the service promotion object.
Optionally, the method further comprises:
acquiring sample putting data of a sample service promotion object in a corresponding promotion platform and initializing the current support probability of a gradient promotion decision tree model;
calculating derivative information of a predefined objective function for the current support probability; newly building a decision tree in the gradient lifting decision tree model according to the derivative information, inputting the sample delivery data into the adjusted gradient lifting decision tree model, and acquiring a new current support probability output by the gradient lifting decision tree model;
and returning to the step of calculating the derivative information of the predefined objective function to the current support probability, repeatedly establishing a new decision tree, carrying out iterative training on the gradient lifting decision tree model until a training end condition is met, and taking the current gradient lifting decision tree model as the decision model.
Optionally, the creating a decision tree in the gradient boosting decision tree model according to the derivative information includes:
in the gradient lifting decision tree model, an initial decision tree is newly built, leaf node splitting is carried out on the newly built initial decision tree according to the derivative information and the optimal segmentation point division algorithm, and when the leaf node splitting meets the splitting stopping condition, the splitting is stopped, so that the newly built decision tree is obtained;
and taking the gradient lifting decision tree model containing the newly-built decision tree as an adjusted gradient lifting decision tree model.
Optionally, the method further comprises:
acquiring multiple preset indexes of the sample, and acquiring multiple repeated associated indexes from the preset indexes;
for the multiple sample preset indexes, carrying out duplicate removal on the multiple associated indexes to obtain duplicate-removed sample preset indexes;
the inputting the sample delivery data into the adjusted gradient lifting decision tree model to obtain a new current support probability output by the gradient lifting decision tree model includes:
and inputting the sample delivery data corresponding to the sample preset index after the duplication removal into the adjusted gradient lifting decision tree model, and acquiring the new current support probability output by the gradient lifting decision tree model.
Optionally, the preset indexes include at least two of the following: promotion times, promotion object browsed times, promotion starting time, promotion ending time, sales volume, click volume, audience number, click through rate, shopping cart adding amount and shopping cart adding amount;
the obtaining of the multiple repeated correlation indexes includes:
obtaining correlation coefficients corresponding to at least two preset indexes from the preset indexes of the multiple samples;
and when the correlation coefficient exceeds a preset threshold value, determining the obtained at least two preset indexes as a plurality of repeated correlation indexes.
Optionally, the obtaining of sample delivery data of the sample service promotion object in the corresponding promotion platform includes:
acquiring a release rule and release flow information corresponding to the promotion platform;
acquiring sample release data corresponding to the sample service promotion object from the corresponding promotion platform according to the release rule and the release flow information;
the delivery flow information includes at least one of:
audience positioning flow, promotion object display flow, clicking flow, drainage flow, purchase adding flow, order placing flow and order processing flow.
An apparatus for maintaining a business promotion object, the apparatus comprising:
the release data acquisition module is used for acquiring release data of the business promotion objects in the corresponding promotion platform; the release data comprises data of a plurality of preset indexes, and the preset indexes reflect the promotion effect of the service promotion objects in the corresponding promotion platforms;
the input module is used for inputting the delivery data into a pre-trained decision model, so as to obtain support probabilities for continuously delivering the service promotion objects in a corresponding promotion platform according to the delivery data and a predefined objective function through a plurality of decision trees in the decision maintenance model, fuse the support probabilities corresponding to the decision trees and output corresponding decision results;
and the decision module is used for determining whether to continuously put the service promotion object in the corresponding promotion platform according to the decision result output by the decision model.
A computer device comprising a memory storing a computer program and a processor implementing the steps of the method as described above when executing the computer program.
A computer-readable storage medium, on which a computer program is stored which, when being executed by a processor, carries out the steps of the method as set forth above.
According to the maintenance method, the maintenance device, the computer equipment and the storage medium for the service popularization object, the releasing data of the service popularization object in the corresponding popularization platform is obtained and is input into the pre-trained decision model, so that the support probability of continuing releasing the service popularization object in the corresponding popularization platform is obtained through a plurality of decision trees in the decision maintenance model according to the releasing data and the pre-defined objective function respectively, the support probabilities corresponding to the decision trees are fused, and the corresponding decision result is output; and determining whether to continue to put the service promotion object in the corresponding promotion platform according to a decision result output by the decision model, so that the automatic and timely maintenance of the subsequent advertisement putting is realized, a manual and empirical maintenance method is eliminated, and the advertisement putting cost is effectively reduced.
Drawings
FIG. 1 is a diagram of an application environment of a method for maintaining a business promotion object in an embodiment;
fig. 2 is a schematic flow chart of a method for maintaining a service promotion object in an embodiment;
FIG. 3 is a schematic flow chart diagram illustrating the training steps of the gradient boosting decision tree model in one embodiment;
FIG. 4 is a schematic flow chart diagram illustrating the decision model building step in one embodiment;
fig. 5 is a block diagram illustrating a structure of a service promotion object maintenance apparatus according to an embodiment;
FIG. 6 is a diagram illustrating an internal structure of a computer device according to an embodiment.
Detailed Description
In order to make the objects, technical solutions and advantages of the present application more apparent, the present application is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application.
The service promotion object maintenance method provided by the application can be applied to the application environment shown in fig. 1. The terminal 102 may communicate with the popularization platform 104 through a network, the terminal 102 may be but is not limited to various personal computers, notebook computers, smart phones, tablet computers, and portable wearable devices, and the popularization platform 104 may be implemented by an independent server or a server cluster formed by a plurality of servers.
In an embodiment, as shown in fig. 2, a method for maintaining a service promotion object is provided, and this embodiment is illustrated by applying the method to a terminal, it can be understood that the method may also be applied to a server, may also be applied to a system including a terminal and a server, and is implemented by interaction between the terminal and the server. In this embodiment, the method includes the steps of:
As an example, the service promotion object may refer to an object having a function of publishing service promotion information, where the service promotion information is information used for promoting, introducing, and promoting relevant contents of a product, such as information about characteristics, functions, and effects of the product; specifically, the service promotion object may include a promotion advertisement, and the corresponding promotion platform may be a platform for delivering the service promotion object, for example, when the promotion advertisement is delivered by means of electronic commerce, the corresponding promotion platform may be an e-commerce platform; when the propaganda advertisement is released in the modes of short messages, telephones and the like, the corresponding promotion platform can be a communication platform; when the propaganda advertisement is released in a self-media mode, the corresponding promotion platform can be a social platform; a corresponding promotional platform may also refer to a platform that has a large data traffic (e.g., exceeds a data traffic threshold or the number of active users exceeds a number threshold). Certainly, the promotion platform may be a platform selected by a person in the field according to an actual situation, and the promotion platform is not limited in the application.
The preset index may include at least two of: the number of promotions (also referred to as show volume), the number of times the referral object was viewed, promotion start time, promotion end time, sales volume, Click volume, audience number (also referred to as coverage), Click Through Rate (CTR), shopping cart add-on volume, shopping cart add-on amount, days of delivery, landing page view volume, Cost Per Click (CPC), conversion Rate (Click Value Rate, CVR), thousand show price (Cost Per mill, CPM), and Return On Investment (ROI).
In practical application, the release data of the service promotion object in the corresponding promotion platform can be obtained, wherein the release data comprises data corresponding to a plurality of preset indexes, and each preset index can reflect the promotion benefit of the service promotion object in the corresponding promotion platform.
In one example, other indicators, such as parent-level features, related features, ring ratio features, etc., may also be derived based on the preset indicators. The parent characteristic refers to the influence degree of preset indexes of other samples on the preset characteristic of the sample in the same group of samples; the correlation characteristics refer to the correlation degree of the correlation indexes; the ring ratio feature is to preset an index for the same sample and calculate the ring ratio according to the time dimension.
In a specific implementation, a pre-trained decision model may be set, and after the delivery data is obtained, the delivery data may be input to the decision model. The decision model comprises a plurality of decision trees, the decision trees in the decision model can respectively determine the support probability of the business promotion object continuing to release in the corresponding promotion platform according to the release data and the predefined objective function, the support probabilities output by the decision trees are fused, and the decision model can output the decision result.
After the decision result output by the decision model is obtained, whether to continue to release the service promotion object in the corresponding promotion platform can be determined according to the decision result, specifically, whether the decision result meets the condition of continuing to release the service promotion object can be judged, and when the condition of continuing to release the service promotion object is met, the service promotion object can be continuously released.
In the embodiment of the application, the releasing data of the service popularization object in the corresponding popularization platform is acquired and is input into a pre-trained decision model, so that the support probability of continuing releasing the service popularization object in the corresponding popularization platform is obtained through a plurality of decision trees in a decision maintenance model according to the releasing data and a predefined objective function respectively, the support probabilities corresponding to the decision trees are fused, and the corresponding decision result is output; and determining whether to continue to put the service promotion object in the corresponding promotion platform according to a decision result output by the decision model, so that the automatic and timely maintenance of the subsequent advertisement putting is realized, a manual and empirical maintenance method is eliminated, and the advertisement putting cost is effectively reduced.
Furthermore, the decision model can objectively determine the maintenance mode according to the delivery data, so that the interference caused by subjective judgment is avoided, meanwhile, the subsequent maintenance mode of the advertisement is automatically determined through the decision model, the batch advertisement maintenance mode judgment can be supported, the labor cost is not required to be increased along with the increase of the advertisement delivery quantity, and the maintenance efficiency is greatly improved.
In an embodiment, the decision result may be a sum of a plurality of support probabilities, and the determining whether to continue to deliver the service promotion object in the corresponding promotion platform according to the decision result output by the decision model may include the steps of:
when the sum of the support probabilities is greater than or equal to a preset threshold value, determining to continuously release the service promotion object on the corresponding promotion platform; and when the sum of the support probabilities is smaller than the preset threshold value, generating information for stopping continuous delivery in the corresponding promotion platform aiming at the service promotion object.
In practical application, since the decision result may be a sum of a plurality of quantized support probabilities, when determining whether to continue to release the service promotion object, the sum of the plurality of support probabilities may be compared with a preset threshold, and when the sum of the plurality of support probabilities is greater than or equal to the preset threshold, it may be determined that the service promotion object is retained, and the service promotion object continues to be released in the corresponding promotion platform. When the sum of the support probabilities is smaller than a preset threshold, information which stops being continuously released on the corresponding promotion platform can be generated for the business promotion object, and the information is fed back to the relevant client side so as to suggest the user to close the business promotion object. The preset threshold may be set to 0.5, or may also be set according to actual requirements, for example, according to the importance of budget funds or business promotion objects.
In this embodiment, when the sum of the support probabilities is greater than or equal to a preset threshold, it is determined that the service promotion object continues to be launched on the corresponding promotion platform; when the sum of the support probabilities is smaller than the preset threshold value, information for stopping continuous delivery in the corresponding promotion platform is generated aiming at the service promotion object, so that the advertisement delivery mode is determined through the quantized probability value, and a high-reliability advertisement delivery decision basis is provided.
In one embodiment, as shown in fig. 3, the method may further include the steps of:
As an example, a sample business promotion object may refer to a business promotion object that is a model training sample.
In practical application, sample launching data of a sample service promotion object in a corresponding promotion platform can be obtained, and a current support probability of a gradient decision model for the sample launching data can be initialized, specifically, when training is just started, the current support probability can be initialized, for example, the current support probability can be initialized to 0.
In practical applications, the objective function may be predefined, and after obtaining the current support probability, the derivative information of the predefined objective function with respect to the current support probability may be calculated. After the derivative information is obtained, a decision tree can be newly built in the gradient lifting decision tree model according to the derivative information, and the sample delivery data is input into the adjusted gradient lifting decision tree model to obtain a new current support probability output by the model. Specifically, when the sample delivery data is input into the adjusted gradient boosting decision tree model, the support probability can be predicted by using a newly-built decision tree in the model, and is accumulated into the original support probability, and finally the model can output a new current probability value.
The predefined objective function is composed of a loss function l and a regular term Ω for suppressing the complexity of the model, and can be represented by the following formula:
the current support probability value may be expressed using the following formula:
wherein, y'iCurrent support probability, f, corresponding to the ith sample service promotion objecttFor the t-th decision tree,
for the support probabilities predicted by the tth decision tree model,
is a loss function consisting of a predictor y'iAnd true value yiIt is shown that,
the regularization term of the decision tree model is lifted for the gradient.
After obtaining the new current support probability, the process may return to step 302, calculate the derivative information of the target function for the new current support probability again, create a new decision tree according to the derivative information, continuously circulate, repeat the steps of creating the new decision tree in the gradient lifting decision tree model, and obtain the new current support probability output by the adjusted gradient lifting decision tree model, perform iterative training on the gradient lifting decision tree model until the training end condition is met, may use the current gradient lifting decision tree model as the decision model, and may normalize the finally predicted support probability by a preset formula, so that the support probability is in the range of (0, 1). The training end condition may be that a preset training frequency is reached, or a value corresponding to the loss function meets a threshold condition, and the like, and a person skilled in the art may select the training end condition according to actual needs, and the normalization of the support probability may be implemented by the following formula:
in a specific implementation, the prediction of the service promotion object maintenance mode needs to solve two problems: the method has the advantages that the time sequence problem and the correlation between the features are solved, and even the correlation problem between the features on the time sequence can be combined through the feature combination of the finally trained decision model.
If other models are used, the effect of the present application cannot be achieved, for example, if "double neural network-MLP (Multi-Layer Perceptron)" is used, although any nonlinear sample features can be fitted during the repetitive training process, the model cannot be explained, and the model does not have the capability of combining time series and correlation between features, and the model also has a risk of overfitting.
For another example, if a "Bi-directional Long Short-Term Memory" model of Bi-directional Bi-LSTM is adopted, although the model has a certain time series processing capability, the processing capability for the correlation between features is weak, and the time series processing capability of Bi-LSTM is on a discrete sample rather than a continuous sample, and the sample delivery data corresponding to the sample service promotion object is continuous and numerical, and is difficult to discretize, which is one of the reasons why Bi-LSTM cannot be applied to the numerical modeling prediction project. In addition, numerical discretization can increase the difficulty of model training.
In this embodiment, sample delivery data and current support probability of a sample service promotion object in a corresponding promotion platform are obtained, a gradient boost decision tree model is trained, feature combination is performed on the delivery data by adopting a gradient boost decision tree algorithm, an advertisement maintenance mode is predicted, and prediction reliability of the decision model for the advertisement maintenance mode is improved.
In an embodiment, the obtaining of the sample delivery data of the sample service promotion object in the corresponding promotion platform may include the following steps:
acquiring a release rule and release flow information corresponding to the promotion platform; and acquiring sample release data corresponding to the sample service promotion object from the corresponding promotion platform according to the release rule and the release flow information.
In a specific implementation, different promotion platforms may have different delivery rules and delivery flows, where the delivery rules are API (Application Programming Interface) rules of the delivery platform, the rules provide specifications of data docking for users, and the delivery flows are delivery flows of the business promotion objects. As an example, the delivery flow information may include at least one of: audience positioning flow, promotion object display flow, clicking flow, drainage flow, purchase adding flow, order placing flow and order processing flow.
Based on the method, the releasing rules and releasing flow information corresponding to the promotion platform can be obtained, data collection can be carried out after the releasing rules and the releasing flow information are obtained, sample releasing data corresponding to one or more sample service promotion objects are obtained from the corresponding promotion platform, and the sample releasing data can comprise a plurality of sample preset indexes. Specifically, for the audience positioning process, the number of coverage people corresponding to the promotion object can be obtained; aiming at the promotion object display flow, the display amount corresponding to the promotion object can be determined; aiming at the click process, the number of times that the link corresponding to the promotion object is clicked, namely the click rate, can be obtained; for the drainage flow, the browsing amount of the landing page can be acquired; aiming at the architecture flow, the adding amount of the shopping cart and the adding amount of the shopping cart can be obtained; for the order placing process, the order placing quantity corresponding to the popularization object can be obtained; for the order processing flow, the sales volume and sales amount may be obtained.
Optionally, data touch can be performed, that is, data dimension, data accuracy, data integrity and data timeliness can be verified for a plurality of samples acquired from the corresponding promotion platform, and the data quality of the collected sample input data is guaranteed.
In this embodiment, the sample release data corresponding to the sample service promotion object is acquired from the corresponding promotion platform according to the release rule and the release flow information, and the decision model only needs to acquire the sample release data in response according to the release rule and the release flow information for self-learning, so that the rule change of the promotion platform can be quickly adapted to acquire the corresponding sample release data, the defect that the platform rule cannot be synchronized in time in manual empirical maintenance is effectively avoided, and the additional artificial learning cost is not required.
In one embodiment, the method may further comprise the steps of:
acquiring multiple preset indexes of the sample, and acquiring multiple repeated associated indexes from the preset indexes; and removing the weight of the plurality of associated indexes aiming at the plurality of sample preset indexes to obtain the removed sample preset indexes.
As an example, the multiple association indexes that are mutually repeated may refer to multiple preset indexes that have an association degree higher than a threshold and produce the same training effect in the model training process.
In the advertisement delivery data analysis, two or more sample preset indexes of the service promotion object can have an incidence relation, for example, a commodity sales amount and an advertisement fund investment amount have a direct proportion relation in time. In practice, the service promotion object may have a large amount of preset indexes, and the preset indexes may be screened to avoid redundant computation.
Specifically, data exploration can be performed, multiple preset indexes corresponding to the sample popularization object are obtained, and multiple related indexes which are mutually repeated are obtained from the multiple preset indexes of the sample. After obtaining the multiple repeated correlation indexes, the multiple correlation indexes can be deduplicated for the existing multiple sample preset indexes to obtain the deduplicated sample preset indexes.
The inputting the sample delivery data into the adjusted gradient lifting decision tree model to obtain a new current support probability output by the gradient lifting decision tree model may include the following steps:
and inputting the sample delivery data corresponding to the sample preset index after the duplication removal into the adjusted gradient lifting decision tree model, and acquiring the new current support probability output by the gradient lifting decision tree model.
After the sample preset index after the duplication removal is obtained, the sample delivery data corresponding to the sample preset index after the duplication removal can be input into the adjusted gradient boost decision tree model, a new current support probability output by the model is obtained, and the model training is continued.
In this embodiment, in the training process of the gradient lifting decision tree model, the duplicate removal is performed on the associated indexes which are repeated in the multiple preset sample indexes, and the model training is performed by using the sample release data corresponding to the sample preset indexes after the duplicate removal, so that the redundant calculation can be avoided, and the training time can be shortened.
In an embodiment, the obtaining the plurality of repeated correlation indicators may include:
obtaining correlation coefficients corresponding to at least two preset indexes from the preset indexes of the multiple samples; and when the correlation coefficient exceeds a preset threshold value, determining the obtained at least two preset indexes as a plurality of repeated correlation indexes.
As an example, the correlation coefficient may represent a degree of correlation between a plurality of preset indexes. In an example of the present application, the correlation coefficient may be a non-linear characteristic. Specifically, a linear relationship exists between the preset indexes of every two samples, but the process of solving the correlation coefficient may be nonlinear, and the correlation coefficient may be solved by the following formula:
θ＝(XTX)-1XTy
wherein, X is a matrix formed by sample release data corresponding to a preset index, T represents the transposition operation of the matrix, and y is a row vector.
After the correlation coefficient is obtained, whether the correlation coefficient exceeds a preset threshold value or not can be judged, when the correlation coefficient exceeds the preset threshold value, at least two preset indexes which are obtained can be determined to be high in correlation, training can be performed based on the at least two preset indexes which are obtained when model training is performed, redundant calculation is avoided, and therefore the correlation coefficient can exceed the preset threshold value, at least two preset indexes can be determined to be a plurality of correlation indexes which are mutually repeated.
In this embodiment, when the correlation coefficient exceeds the preset threshold, the obtained at least two preset indexes are determined as multiple repeated correlation indexes, so that the preset indexes are screened through the correlation coefficient, and a basis is provided for avoiding redundant calculation and shortening the model training time.
In one example, the feature input model can be trained as a new feature input model by constructing pairwise index linear correlation coefficients for the features. In the traditional data analysis method, addition and subtraction of features do not generate the features of information, so that new dimension information cannot be extracted from the features of multiple dimensions; division of the features results in a loss of information, for example 4/2 equals 2 and 100/50 also equals 2, but the features 4, 2 contain different information than the features 100, 50. In the present example, the training time of the model can be reduced by the correlation coefficient, and practice shows that the training time corresponding to model training without the correlation coefficient is 18 times of the training time of model training with the correlation coefficient, but the accuracy is lower than that of the model training with the correlation coefficient
Optionally, other calculation processing may be performed on data corresponding to the multiple sample preset indexes, for example, feature fusion may be performed, and a maximum value, a minimum value, a mean value, a variance, and summary information of the data corresponding to each sample preset index are calculated; or, historical characteristics of the sample preset index can be obtained, and the historical characteristics are statistical information obtained by comparing historical data of the sample preset index with other data.
In an embodiment, the creating a decision tree in the gradient boosting decision tree model according to the derivative information may include the following steps:
in the gradient lifting decision tree model, an initial decision tree is newly built, leaf node splitting is carried out on the newly built initial decision tree according to the derivative information and the optimal segmentation point division algorithm, and when the leaf node splitting meets the splitting stopping condition, the splitting is stopped, so that the newly built decision tree is obtained; and taking the gradient lifting decision tree model containing the newly-built decision tree as an adjusted gradient lifting decision tree model.
In practical application, an initial decision tree can be newly established in the gradient lifting decision tree model, the tree depth of the initial decision tree can be 0, after the initial decision tree is obtained, leaf node splitting can be performed on the newly established initial decision tree according to derivative information and an optimal splitting point division algorithm, a left leaf node and a right leaf node can be split during each splitting, the splitting is stopped when the leaf node splitting meets the splitting stopping condition, the newly established decision tree can be obtained, and then the gradient lifting decision tree model containing the newly established decision tree can be used as the adjusted gradient lifting decision tree model.
Specifically, the cleavage stop conditions may include any one or more of the following: stopping the leaf node splitting when the Gain value (Gain) is smaller than the Gain threshold value; or, when the tree depth of the decision tree reaches the preset maximum depth, the splitting is stopped, because the tree depth is too large, the phenomenon of overfitting is easy to occur. When the splitting stop condition is that the gain value is smaller than the gain threshold value, the splitting is stopped, and the gain value can be calculated by the following formula:
wherein G isLIs the accumulated sum of the first partial derivatives, G, of the samples contained in the left leaf nodeRIs the accumulated sum of the first partial derivatives of the samples contained in the right leaf node, HLIs the cumulative sum of the second partial derivatives, H, of the samples contained in the left leaf nodeRThe sum of the second partial derivatives of the samples contained in the right leaf node is accumulated, and gamma and lambda are preset parameters.
In this embodiment, the newly-built decision tree can be subjected to leaf node splitting according to the derivative information and the optimal splitting point partitioning algorithm, an optimal splitting point is found, and the splitting condition of the leaf node is effectively controlled.
In order to enable those skilled in the art to better understand the model building process in the present application, the following description illustrates an embodiment of the present application by way of an example, but it should be understood that the embodiment of the present application is not limited thereto.
As shown in fig. 4, when a decision model is constructed, data collection may be performed in a corresponding promotion platform by using each advertisement as a minimum collection granularity according to a delivery rule and delivery flow information. Data corresponding to multiple 1200 sample preset indexes including advertisement putting days, number of covered people, display amount, click amount, landing page browsing amount, sales volume and sales amount can be subjected to data touch, data acquisition dimensionality, data accuracy, data integrity and data timeliness are verified, data exploration is performed after verification is passed, multiple sample preset indexes for training a model are obtained, and the data are input to a gradient lifting decision tree model for data modeling and model training after data standardization.
After the model training is finished and a decision model is obtained, the prediction effect of the model can be evaluated, specifically, one or more of recall rate, accuracy rate and AUC (Area Under Curve, defined as the Area value enclosed by the coordinate axis Under the Receiver Operating Characteristic (ROC)) can be calculated through the predicted value and the actual value of the verification set, and the model effect can be evaluated.
When the model effect of the decision model is evaluated, the algorithm model can be solidified, specifically, the decision model can be deployed online, task scheduling is received, and subsequent maintenance of the model is performed.
It should be understood that although the various steps in the flow charts of fig. 1-4 are shown in order as indicated by the arrows, the steps are not necessarily performed in order as indicated by the arrows. The steps are not performed in the exact order shown and described, and may be performed in other orders, unless explicitly stated otherwise. Moreover, at least some of the steps in fig. 1-4 may include multiple steps or multiple stages, which are not necessarily performed at the same time, but may be performed at different times, which are not necessarily performed in sequence, but may be performed in turn or alternately with other steps or at least some of the other steps.
In one embodiment, as shown in fig. 5, there is provided a service promotion object maintenance apparatus, including:
a release data acquisition module 501, configured to acquire release data of the service promotion object in the corresponding promotion platform; the release data comprises data of a plurality of preset indexes, and the preset indexes reflect the promotion effect of the service promotion objects in the corresponding promotion platforms;
an input module 502, configured to input the delivery data into a pre-trained decision model, so as to obtain, through multiple decision trees in the decision maintenance model, support probabilities for continuing to deliver the service promotion object in a corresponding promotion platform according to the delivery data and a predefined objective function, and fuse the support probabilities corresponding to the multiple decision trees to output a corresponding decision result;
and the decision module 503 is configured to determine whether to continue to release the service promotion object in the corresponding promotion platform according to the decision result output by the decision model.
In one embodiment, the decision result is a sum of a plurality of support probabilities; the decision module 403 includes:
a first decision result output submodule, configured to determine to continue to release the service promotion object on the corresponding promotion platform when a sum of the support probabilities is greater than or equal to a preset threshold;
and the second decision result output submodule is used for generating information for stopping continuous delivery in the corresponding promotion platform aiming at the service promotion object when the sum of the support probabilities is smaller than the preset threshold value.
In one embodiment, the apparatus further comprises:
the system comprises a sample release data acquisition module, a gradient promotion decision tree model generation module and a gradient promotion decision tree model generation module, wherein the sample release data acquisition module is used for acquiring sample release data of a sample service promotion object in a corresponding promotion platform and initializing the current support probability of the gradient promotion decision tree model;
the derivative information calculation module is used for calculating derivative information of a predefined objective function for the current support probability; newly building a decision tree in the gradient lifting decision tree model according to the derivative information, inputting the sample delivery data into the adjusted gradient lifting decision tree model, and acquiring a new current support probability output by the gradient lifting decision tree model;
and the training module is used for returning the step of calculating the derivative information of the predefined target function on the current support probability, repeatedly establishing a new decision tree, carrying out iterative training on the gradient lifting decision tree model until a training end condition is met, and taking the current gradient lifting decision tree model as the decision model.
In one embodiment, the derivative information calculation module includes:
the leaf node splitting sub-module is used for newly building an initial decision tree in the gradient lifting decision tree model, performing leaf node splitting on the newly built initial decision tree according to the derivative information and an optimal segmentation point division algorithm, and stopping splitting when the leaf node splitting meets a splitting stopping condition to obtain a newly built decision tree;
and the adjusting submodule is used for taking the gradient lifting decision tree model containing the newly-built decision tree as an adjusted gradient lifting decision tree model.
In one embodiment, the apparatus further comprises:
the system comprises a sample preset index acquisition module, a sample preset index acquisition module and a correlation index acquisition module, wherein the sample preset index acquisition module is used for acquiring multiple sample preset indexes and acquiring multiple correlation indexes which are mutually repeated from the multiple sample preset indexes;
the duplicate removal module is used for removing the duplicate of the plurality of associated indexes aiming at the plurality of sample preset indexes to obtain the sample preset indexes after the duplicate removal;
the derivative information calculation module includes:
and the sample delivery data input submodule is used for inputting the sample delivery data corresponding to the sample preset index after the duplication removal into the adjusted gradient lifting decision tree model and acquiring the new current support probability output by the gradient lifting decision tree model.
In one embodiment, the plurality of preset indicators include at least two of: promotion times, promotion object browsed times, promotion starting time, promotion ending time, sales volume, click volume, audience number, click through rate, shopping cart adding amount and shopping cart adding amount;
the sample preset index obtaining module comprises:
the correlation coefficient determining submodule is used for acquiring correlation coefficients corresponding to at least two preset indexes from the preset indexes of the multiple samples;
and the correlation index determining submodule is used for determining the obtained at least two preset indexes as a plurality of correlation indexes which are mutually repeated when the correlation coefficient exceeds a preset threshold value.
In one embodiment, the sample placement data obtaining module includes:
the release rule acquisition submodule is used for acquiring release rules and release flow information corresponding to the promotion platform;
the data extraction submodule is used for acquiring sample release data corresponding to the sample service promotion object from the corresponding promotion platform according to the release rule and the release flow information;
the delivery flow information includes at least one of:
audience positioning flow, promotion object display flow, clicking flow, drainage flow, purchase adding flow, order placing flow and order processing flow.
For specific limitation of a service promotion object maintenance device, reference may be made to the above limitation on a service promotion object maintenance method, which is not described herein again. All modules in the service promotion object maintenance device can be completely or partially realized through software, hardware and a combination thereof. The modules can be embedded in a hardware form or independent from a processor in the computer device, and can also be stored in a memory in the computer device in a software form, so that the processor can call and execute operations corresponding to the modules.
In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as shown in fig. 5. The computer device includes a processor, a memory, a communication interface, a display screen, and an input device connected by a system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device comprises a nonvolatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of an operating system and computer programs in the non-volatile storage medium. The communication interface of the computer device is used for carrying out wired or wireless communication with an external terminal, and the wireless communication can be realized through WIFI, an operator network, NFC (near field communication) or other technologies. The computer program is executed by a processor to implement a method for maintaining a business promotion object. The display screen of the computer equipment can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer equipment can be a touch layer covered on the display screen, a key, a track ball or a touch pad arranged on the shell of the computer equipment, an external keyboard, a touch pad or a mouse and the like.
Those skilled in the art will appreciate that the architecture shown in fig. 5 is merely a block diagram of some of the structures associated with the disclosed aspects and is not intended to limit the computing devices to which the disclosed aspects apply, as particular computing devices may include more or less components than those shown, or may combine certain components, or have a different arrangement of components.
In one embodiment, a computer device is provided, comprising a memory and a processor, the memory having a computer program stored therein, the processor implementing the following steps when executing the computer program:
acquiring release data of the business promotion object in a corresponding promotion platform; the release data comprises data of a plurality of preset indexes, and the preset indexes reflect the promotion effect of the service promotion objects in the corresponding promotion platforms;
inputting the delivery data into a pre-trained decision model, so as to obtain support probabilities for continuing delivering the service promotion objects in a corresponding promotion platform according to the delivery data and a predefined objective function through a plurality of decision trees in the decision maintenance model, and fusing the support probabilities corresponding to the decision trees to output corresponding decision results;
and determining whether to continue to put the service promotion object in the corresponding promotion platform according to a decision result output by the decision model.
In one embodiment, the processor, when executing the computer program, also performs the other steps in the above embodiments.
In one embodiment, a computer-readable storage medium is provided, having a computer program stored thereon, which when executed by a processor, performs the steps of:
acquiring release data of the business promotion object in a corresponding promotion platform; the release data comprises data of a plurality of preset indexes, and the preset indexes reflect the promotion effect of the service promotion objects in the corresponding promotion platforms;
inputting the delivery data into a pre-trained decision model, so as to obtain support probabilities for continuing delivering the service promotion objects in a corresponding promotion platform according to the delivery data and a predefined objective function through a plurality of decision trees in the decision maintenance model, and fusing the support probabilities corresponding to the decision trees to output corresponding decision results;
and determining whether to continue to put the service promotion object in the corresponding promotion platform according to a decision result output by the decision model.
In one embodiment, the computer program when executed by the processor also performs the other steps in the above embodiments.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by hardware instructions of a computer program, which can be stored in a non-volatile computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. Any reference to memory, storage, database or other medium used in the embodiments provided herein can include at least one of non-volatile and volatile memory. Non-volatile Memory may include Read-Only Memory (ROM), magnetic tape, floppy disk, flash Memory, optical storage, or the like. Volatile Memory can include Random Access Memory (RAM) or external cache Memory. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM), among others.
The technical features of the above embodiments can be arbitrarily combined, and for the sake of brevity, all possible combinations of the technical features in the above embodiments are not described, but should be considered as the scope of the present specification as long as there is no contradiction between the combinations of the technical features.
The above-mentioned embodiments only express several embodiments of the present application, and the description thereof is more specific and detailed, but not construed as limiting the scope of the invention. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the concept of the present application, which falls within the scope of protection of the present application. Therefore, the protection scope of the present patent shall be subject to the appended claims.
Claims (10)
1. A method for maintaining a service promotion object is characterized in that the method comprises the following steps:
acquiring release data of the business promotion object in a corresponding promotion platform; the release data comprises data of a plurality of preset indexes, and the preset indexes reflect the promotion effect of the service promotion objects in the corresponding promotion platforms;
inputting the delivery data into a pre-trained decision model, so as to obtain support probabilities for continuing delivering the service promotion objects in a corresponding promotion platform according to the delivery data and a predefined objective function through a plurality of decision trees in the decision maintenance model, and fusing the support probabilities corresponding to the decision trees to output corresponding decision results;
and determining whether to continue to put the service promotion object in the corresponding promotion platform according to a decision result output by the decision model.
2. The method of claim 1, wherein the decision result is a sum of a plurality of support probabilities; the determining whether to continue to put the service promotion object in the corresponding promotion platform according to the decision result output by the decision model includes:
when the sum of the support probabilities is greater than or equal to a preset threshold value, determining to continuously release the service promotion object on the corresponding promotion platform;
and when the sum of the support probabilities is smaller than the preset threshold value, generating information for stopping continuous delivery in the corresponding promotion platform aiming at the service promotion object.
3. The method of claim 1, further comprising:
acquiring sample putting data of a sample service promotion object in a corresponding promotion platform and initializing the current support probability of a gradient promotion decision tree model;
calculating derivative information of a predefined objective function for the current support probability; newly building a decision tree in the gradient lifting decision tree model according to the derivative information, inputting the sample delivery data into the adjusted gradient lifting decision tree model, and acquiring a new current support probability output by the gradient lifting decision tree model;
and returning to the step of calculating the derivative information of the predefined objective function to the current support probability, repeatedly establishing a new decision tree, carrying out iterative training on the gradient lifting decision tree model until a training end condition is met, and taking the current gradient lifting decision tree model as the decision model.
4. The method of claim 3, wherein building a decision tree in the gradient boosting decision tree model according to the derivative information comprises:
in the gradient lifting decision tree model, an initial decision tree is newly built, leaf node splitting is carried out on the newly built initial decision tree according to the derivative information and the optimal segmentation point division algorithm, and when the leaf node splitting meets the splitting stopping condition, the splitting is stopped, so that the newly built decision tree is obtained;
and taking the gradient lifting decision tree model containing the newly-built decision tree as an adjusted gradient lifting decision tree model.
5. The method of claim 3, further comprising:
acquiring multiple preset indexes of the sample, and acquiring multiple repeated associated indexes from the preset indexes;
for the multiple sample preset indexes, carrying out duplicate removal on the multiple associated indexes to obtain duplicate-removed sample preset indexes;
the inputting the sample delivery data into the adjusted gradient lifting decision tree model to obtain a new current support probability output by the gradient lifting decision tree model includes:
and inputting the sample delivery data corresponding to the sample preset index after the duplication removal into the adjusted gradient lifting decision tree model, and acquiring the new current support probability output by the gradient lifting decision tree model.
6. The method of claim 5, wherein the plurality of preset indicators include at least two of: promotion times, promotion object browsed times, promotion starting time, promotion ending time, sales volume, click volume, audience number, click through rate, shopping cart adding amount and shopping cart adding amount;
the obtaining of the multiple repeated correlation indexes includes:
obtaining correlation coefficients corresponding to at least two preset indexes from the preset indexes of the multiple samples;
and when the correlation coefficient exceeds a preset threshold value, determining the obtained at least two preset indexes as a plurality of repeated correlation indexes.
7. The method of claim 3, wherein the obtaining of the sample delivery data of the sample service promotion object in the corresponding promotion platform comprises:
acquiring a release rule and release flow information corresponding to the promotion platform;
acquiring sample release data corresponding to the sample service promotion object from the corresponding promotion platform according to the release rule and the release flow information;
the delivery flow information includes at least one of:
audience positioning flow, promotion object display flow, clicking flow, drainage flow, purchase adding flow, order placing flow and order processing flow.
8. An apparatus for maintaining a business promotion object, the apparatus comprising:
the release data acquisition module is used for acquiring release data of the business promotion objects in the corresponding promotion platform; the release data comprises data of a plurality of preset indexes, and the preset indexes reflect the promotion effect of the service promotion objects in the corresponding promotion platforms;
the input module is used for inputting the delivery data into a pre-trained decision model, so as to obtain support probabilities for continuously delivering the service promotion objects in a corresponding promotion platform according to the delivery data and a predefined objective function through a plurality of decision trees in the decision maintenance model, fuse the support probabilities corresponding to the decision trees and output corresponding decision results;
and the decision module is used for determining whether to continuously put the service promotion object in the corresponding promotion platform according to the decision result output by the decision model.
9. A computer device comprising a memory and a processor, the memory storing a computer program, wherein the processor implements the steps of the method of any one of claims 1 to 7 when executing the computer program.
10. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps of the method according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011379503.6A CN112348600A (en) | 2020-11-30 | 2020-11-30 | Service promotion object maintenance method and device, computer equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011379503.6A CN112348600A (en) | 2020-11-30 | 2020-11-30 | Service promotion object maintenance method and device, computer equipment and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112348600A true CN112348600A (en) | 2021-02-09 |
Family
ID=74366212
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011379503.6A Pending CN112348600A (en) | 2020-11-30 | 2020-11-30 | Service promotion object maintenance method and device, computer equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112348600A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116701770A (en) * | 2023-08-01 | 2023-09-05 | 北京创智汇聚科技有限公司 | Request response optimization method and system based on decision scene |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102122380A (en) * | 2011-02-25 | 2011-07-13 | 北京开心人信息技术有限公司 | Method and system for realizing network advertisement release |
| CN106651411A (en) * | 2015-10-29 | 2017-05-10 | 北京国双科技有限公司 | Advertising promotion strategy adjusting method and apparatus |
| CN111768251A (en) * | 2020-09-03 | 2020-10-13 | 北京悠易网际科技发展有限公司 | Advertisement putting method and device based on traffic information evaluation and electronic equipment |
-
2020
- 2020-11-30 CN CN202011379503.6A patent/CN112348600A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102122380A (en) * | 2011-02-25 | 2011-07-13 | 北京开心人信息技术有限公司 | Method and system for realizing network advertisement release |
| CN106651411A (en) * | 2015-10-29 | 2017-05-10 | 北京国双科技有限公司 | Advertising promotion strategy adjusting method and apparatus |
| CN111768251A (en) * | 2020-09-03 | 2020-10-13 | 北京悠易网际科技发展有限公司 | Advertisement putting method and device based on traffic information evaluation and electronic equipment |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116701770A (en) * | 2023-08-01 | 2023-09-05 | 北京创智汇聚科技有限公司 | Request response optimization method and system based on decision scene |
| CN116701770B (en) * | 2023-08-01 | 2023-10-27 | 北京创智汇聚科技有限公司 | Request response optimization method and system based on decision scene |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US20220198289A1 (en) | Recommendation model training method, selection probability prediction method, and apparatus | |
| EP4145308A1 (en) | Search recommendation model training method, and search result sorting method and device | |
| CN103502899B (en) | Dynamic prediction Modeling Platform | |
| CN110555714A (en) | method and apparatus for outputting information | |
| WO2022100518A1 (en) | User profile-based object recommendation method and device | |
| EP3152640A1 (en) | Systems and methods for serving product recommendations | |
| CN109117442B (en) | Application recommendation method and device | |
| WO2023016159A1 (en) | Method and system for predicting network traffic of smart city | |
| US11811708B2 (en) | Systems and methods for generating dynamic conversational responses using cluster-level collaborative filtering matrices | |
| CN110889759A (en) | Credit data determination method, device and storage medium | |
| CN111783810A (en) | Method and apparatus for determining attribute information of user | |
| CN112258238A (en) | User life value cycle detection method and device and computer equipment | |
| CN115630221A (en) | Terminal application interface display data processing method and device and computer equipment | |
| CN112348600A (en) | Service promotion object maintenance method and device, computer equipment and storage medium | |
| CN108876422B (en) | Method and device for information popularization, electronic equipment and computer readable medium | |
| US20220091713A1 (en) | Systems and methods for generating dynamic interface options using machine learning models | |
| US20210374520A1 (en) | Personalized recommendations using a transformer neural network | |
| CN111062774B (en) | Activity delivery method and device, electronic equipment and computer readable medium | |
| CN117217284A (en) | Data processing method and device | |
| CN111489196A (en) | Prediction method and device based on deep learning network, electronic equipment and medium | |
| US11049041B2 (en) | Online training and update of factorization machines using alternating least squares optimization | |
| CN108667875A (en) | information updating method and device | |
| US20140324524A1 (en) | Evolving a capped customer linkage model using genetic models | |
| CN112307334A (en) | Information recommendation method, information recommendation device, storage medium and electronic equipment | |
| CN110837596A (en) | Intelligent recommendation method and device, computer equipment and storage medium |
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 |