Object classification method, network model training method and device

Abstract

The present disclosure provides an object classification method, a network model training method, and an apparatus, device, medium, and product, which relate to the field of artificial intelligence, specifically the field of content service technology. The specific implementation scheme includes: dividing the target object set to be classified into a first object subset and a second object subset, wherein the portrait feature distribution of the target objects in the target object set is consistent; determining a first conversion rate associated with the target objects in the first object subset, and determining a second conversion rate associated with the target objects in the second object subset; determining an attribute classification result for the target object set based on the first conversion rate and the second conversion rate, wherein the first conversion rate indicates the positive feedback probability of the corresponding target object based on the content push condition, and the second conversion rate indicates the positive feedback probability of the corresponding target object based on the non-content push condition.

Description

Technical Field

[0001] The present disclosure relates to the field of artificial intelligence, specifically the field of content service technology, and can be applied to scenarios such as object classification. Background Art

[0002] Object classification can determine the object's attribute type, and content push based on the object's attribute type can effectively ensure the effectiveness of content push. However, in some scenarios, the object classification process has low accuracy and high cost. Summary of the Invention

[0003] The present disclosure provides an object classification method, a network model training method, and an apparatus, equipment, medium, and product.

[0004] According to one aspect of the present disclosure, an object classification method is provided, including: dividing a target object set to be classified into a first object subset and a second object subset, wherein the portrait feature distribution of the target objects in the target object set is consistent; determining a first conversion rate associated with the target objects in the first object subset, and determining a second conversion rate associated with the target objects in the second object subset; and determining an attribute classification result for the target object set based on the first conversion rate and the second conversion rate, wherein the first conversion rate indicates a positive feedback probability of the corresponding target object based on a content push condition, and the second conversion rate indicates a positive feedback probability of the corresponding target object based on a non-content push condition.

[0005] According to another aspect of the present disclosure, a method for training a network model is provided, comprising: determining a model to be trained that matches each sample object in a sample object set based on whether each sample object is a content push object; using object portrait data of each sample object as input data of the corresponding model to be trained to obtain a predicted conversion rate for each sample object; and adjusting model parameters of the corresponding model to be trained based on the predicted conversion rate and preset conversion label of each sample object to obtain a trained target network model.

[0006] According to another aspect of the present disclosure, an object classification device is provided, including: a first processing module for dividing a target object set to be classified into a first object subset and a second object subset, wherein the portrait feature distribution of the target objects in the target object set is consistent; a second processing module for determining a first conversion rate associated with the target objects in the first object subset, and determining a second conversion rate associated with the target objects in the second object subset; and a third processing module for determining an attribute classification result for the target object set based on the first conversion rate and the second conversion rate, wherein the first conversion rate indicates a positive feedback probability of the corresponding target object based on a content push condition, and the second conversion rate indicates a positive feedback probability of the corresponding target object based on a non-content push condition.

[0007] According to another aspect of the present disclosure, a network model training device is provided, including: an eighth processing module, used to determine a model to be trained that matches each sample object in a sample object set according to whether each sample object is a content push object; a ninth processing module, used to use the object portrait data of each sample object as input data of the corresponding model to be trained to obtain a predicted conversion rate for each sample object; and a tenth processing module, used to adjust the model parameters of the corresponding model to be trained according to the predicted conversion rate and preset conversion label of each sample object to obtain a trained target network model.

[0008] According to another aspect of the present disclosure, an electronic device is provided, comprising: at least one processor and a memory communicatively connected to the at least one processor. The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the above-described object classification method or network model training method.

[0009] According to another aspect of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to enable the computer to execute the above-mentioned object classification method or network model training method.

[0010] According to another aspect of the present disclosure, a computer program product is provided, including a computer program, which implements the above-mentioned object classification method or network model training method when executed by a processor.

[0011] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present disclosure, nor are they intended to limit the scope of the present disclosure. Other features of the present disclosure will become readily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] The accompanying drawings are provided to facilitate a better understanding of the present invention and do not constitute a limitation of the present disclosure.

[0013] Figure 1 The system architecture of the object classification method and device according to an embodiment of the present disclosure is schematically shown;

[0014] Figure 2 The following schematically shows a flow chart of an object classification method according to an embodiment of the present disclosure;

[0015] Figure 3 Schematically shows a flow chart of an object classification method according to yet another embodiment of the present disclosure;

[0016] Figure 4 The flowchart of the network model training method according to an embodiment of the present disclosure is schematically shown;

[0017] Figure 5 A schematic diagram schematically illustrates a training process of a network model according to an embodiment of the present disclosure;

[0018] Figure 6 A block diagram of an object classification device according to an embodiment of the present disclosure is schematically shown;

[0019] Figure 7 A block diagram schematically illustrates a network model training device according to an embodiment of the present disclosure;

[0020] Figure 8 A block diagram of an electronic device for object classification according to an embodiment of the present disclosure is schematically shown. DETAILED DESCRIPTION

[0021] The following description of exemplary embodiments of the present disclosure is made in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding. These details should be considered as merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0022] The terms used herein are only for describing specific embodiments and are not intended to limit the present disclosure. The terms "comprise," "include," etc. used herein indicate the presence of the features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.

[0023] All terms used herein (including technical and scientific terms) have the meanings commonly understood by those skilled in the art unless otherwise defined. It should be noted that the terms used herein should be interpreted as having a meaning consistent with the context of this specification and should not be interpreted in an idealized or overly rigid manner.

[0024] When expressions such as "at least one of A, B and C, etc." are used, they should generally be interpreted in accordance with the meaning of the expression commonly understood by those skilled in the art (for example, "a system having at least one of A, B and C" should include but is not limited to a system having A alone, B alone, C alone, A and B, A and C, B and C, and / or A, B, C, etc.).

[0025] An embodiment of the present disclosure provides an object classification method. The method of this embodiment includes: dividing the target object set to be classified into a first object subset and a second object subset, wherein the portrait feature distribution of the target objects in the target object set is consistent; determining a first conversion rate associated with the target objects in the first object subset, and determining a second conversion rate associated with the target objects in the second object subset; and determining an attribute classification result for the target object set based on the first conversion rate and the second conversion rate. The first conversion rate indicates the probability of positive feedback of the corresponding target object based on content push conditions, and the second conversion rate indicates the probability of positive feedback of the corresponding target object based on non-content push conditions.

[0026] Figure 1 The system architecture of the object classification method and device according to an embodiment of the present disclosure is schematically shown. It should be noted that, Figure 1 The examples shown are merely examples of system architectures to which the embodiments of the present disclosure may be applied, to help those skilled in the art understand the technical content of the present disclosure, but do not mean that the embodiments of the present disclosure may not be used in other devices, systems, environments or scenarios.

[0027] System architecture 100 according to this embodiment may include a requesting terminal 101, a network 102, and a server 103. Network 102 is used to provide a medium for a communication link between requesting terminal 101 and server 103. Network 102 may include various connection types, such as wired or wireless communication links or fiber optic cables. Server 103 may be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud computing, network services, and middleware services.

[0028] The request terminal 101 interacts with the server 103 via the network 102 to receive or send data, etc. The request terminal 101 is used, for example, to initiate an object classification request to the server 103. The request terminal 101 is also used to send object profile data of a set of target objects to be classified to the server 103. The object profile data may include, for example, attribute profile data and behavior profile data of the target objects.

[0029] The server 103 may be a server that provides various services, for example, a background processing server that performs object classification processing according to the object classification request sent by the request terminal 101 (only as an example).

[0030] For example, in response to an object classification request received from the requesting terminal 101, the server 103 divides the target object set to be classified into a first object subset and a second object subset, wherein the target objects in the target object set have a consistent distribution of image features; determines a first conversion rate associated with the target objects in the first object subset, and determines a second conversion rate associated with the target objects in the second object subset; and determines an attribute classification result for the target object set based on the first conversion rate and the second conversion rate. The first conversion rate indicates a positive feedback probability of the corresponding target object based on a content push condition, and the second conversion rate indicates a positive feedback probability of the corresponding target object based on a non-content push condition.

[0031] It should be noted that the object classification method provided in the embodiments of the present disclosure can be executed by the server 103. Accordingly, the object classification device provided in the embodiments of the present disclosure can be set in the server 103. The object classification method provided in the embodiments of the present disclosure can also be executed by a server or server cluster that is different from the server 103 and can communicate with the request terminal 101 and / or the server 103. Accordingly, the object classification device provided in the embodiments of the present disclosure can also be set in a server or server cluster that is different from the server 103 and can communicate with the request terminal 101 and / or the server 103.

[0032] It should be understood that Figure 1 The numbers of request terminals, networks, and servers in the embodiment are merely illustrative. Any number of request terminals, networks, and servers may be provided as required.

[0033] The present disclosure provides an object classification method. Figure 1 The system architecture of Figures 2 and 3 The object classification method according to the exemplary embodiment of the present disclosure is described. The object classification method of the embodiment of the present disclosure can be, for example, Figure 1 The server 103 shown is used for execution.

[0034] Figure 2The flowchart of the object classification method according to an embodiment of the present disclosure is schematically shown.

[0035] like Figure 2 As shown, the object classification method 200 of the embodiment of the present disclosure may include, for example, operations S210 to S230.

[0036] In operation S210 , a target object set to be classified is divided into a first object subset and a second object subset, and the image feature distribution of the target objects in the target object set is consistent.

[0037] In operation S220 , a first conversion rate associated with a target object in the first object subset is determined, and a second conversion rate associated with a target object in the second object subset is determined.

[0038] In operation S230 , an attribute classification result for the target object set is determined according to the first conversion rate and the second conversion rate.

[0039] The first conversion rate indicates a positive feedback probability of the corresponding target object based on the content push condition, and the second conversion rate indicates a positive feedback probability of the corresponding target object based on the non-content push condition.

[0040] The following examples illustrate the operation flow of the object classification method of this embodiment.

[0041] Exemplarily, the target object set may be uploaded by the object classification requester, or may be reported by a third-party client. The third-party client may be, for example, a client capable of processing the target service corresponding to the content to be pushed.

[0042] For a set of target objects to be classified, the set of target objects can be randomly divided into a first object subset and a second object subset. The set of target objects includes multiple target objects, and the image feature distribution of the multiple target objects is consistent. The number of target objects in the first object subset and the second object subset can be the same or different, and this embodiment does not limit this.

[0043] Based on the object profile data of the target objects in the first object subset, the probability of positive feedback from the target objects based on the content push conditions can be determined as a first conversion rate. Based on the object profile data of the target objects in the second object subset, the probability of positive feedback from the target objects based on non-content push conditions can be determined as a second conversion rate. The object profile data may include, for example, attribute profile data and behavior profile data of the target objects.

[0044] Object profile data can be obtained through various public, legal, and compliant means, such as from public datasets or by data collectors after obtaining authorization from the user associated with the object profile data. Object profile data is not contextual data specific to a specific user and does not reflect a specific user's personal information. The scope of application of object profile data is limited to areas where users have the right to know and have authorized its use.

[0045] For example, the pushed content may include pushed marketing content. A positive feedback probability of target objects in the first object subset performing a responsive action toward the marketing objective when the marketing content is pushed is determined as a first conversion rate. A positive feedback probability of target objects in the second object subset performing a responsive action toward the marketing objective when the marketing content is not pushed is determined as a second conversion rate.

[0046] Marketing objectives can include products, services, events, and locations. Products can include both tangible and intangible products. Events can include real-time events to be promoted. Locations can include locations to be promoted and iconic locations. Responses to marketing objectives can include transactions, downloads, clicks, installations, and sharing, among other actions.

[0047] According to the first conversion rate and the second conversion rate, an attribute classification result for the target object set is determined. The attribute classification result may include, for example, marketing-sensitive objects, natural conversion objects, indifferent objects, and reactive objects.

[0048] For example, when marketing-sensitive subjects are pushed marketing content, the probability of positive feedback from executing a response behavior increases. Natural conversion subjects have a positive feedback probability above the target upper limit regardless of whether marketing content is pushed. Indifferent subjects have a positive feedback probability below the target lower limit regardless of whether marketing content is pushed. Reactive subjects have a positive feedback probability decrease when marketing content is pushed.

[0049] Through the embodiment of the present disclosure, the target object set to be classified is divided into a first object subset and a second object subset, the first conversion rate of the target objects in the first object subset based on the content push condition is determined, and the second conversion rate of the target objects in the second object subset based on the non-content push condition is determined, and the attribute classification result for the target object set is determined based on the first conversion rate and the second conversion rate. Attribute classification of the target object set based on the first conversion rate and the second conversion rate can effectively improve the accuracy of the object classification result, help improve the object classification efficiency, and help reduce the object classification cost. By determining the attribute classification result of the target object set, it is helpful to provide reliable decision support for improving the content push effect and reducing the content push cost.

[0050] Figure 3 The flowchart of the object classification method according to another embodiment of the present disclosure is schematically shown.

[0051] like Figure 3 As shown, the object classification method 300 of the embodiment of the present disclosure may include, for example, operation S210, operation S310, and operation S230.

[0052] In operation S210 , a target object set to be classified is divided into a first object subset and a second object subset, and the image feature distribution of the target objects in the target object set is consistent.

[0053] In operation S310, a first conversion rate is output based on object portrait data of target objects in a first object subset using a first prediction model, and a second conversion rate is output based on object portrait data of target objects in a second object subset using a second prediction model.

[0054] In operation S230 , an attribute classification result for the target object set is determined according to the first conversion rate and the second conversion rate.

[0055] The first conversion rate indicates a positive feedback probability of the corresponding target object based on the content push condition, and the second conversion rate indicates a positive feedback probability of the corresponding target object based on the non-content push condition.

[0056] The following illustrates an example flow of each operation of the object classification method of this embodiment.

[0057] For example, the object portrait data of the target objects in the first object subset can be used as input data of the trained first prediction model to obtain the probability of positive feedback of the target objects based on the content push condition as the first conversion rate. The object portrait data of the target objects in the second object subset can be used as input data of the trained second prediction model to obtain the probability of positive feedback of the target objects based on the non-content push condition as the second conversion rate.

[0058] The first prediction model and the second prediction model can be, for example, a logistic regression (LR) model, a deep neural network (DNN) model, a gradient boosting decision tree (GBDT) model or a binary classification model, etc., which is not limited in this embodiment.

[0059] The first prediction model can be trained based on the object portrait data of the content push object, and the second prediction model can be trained based on the object portrait data of the non-content push object. The object portrait data can be, for example, label attribute data under important data dimensions based on preset portrait label matching. The object portrait data can include attribute portrait data and behavior portrait data. The attribute portrait data can include, for example, basic attributes, social attributes, interest preferences, APP preferences and other different data. The behavior portrait data can include, for example, transaction behavior, download behavior, installation behavior, sharing behavior and other different data.

[0060] Based on the first conversion rate and the second conversion rate, an attribute classification result for the target object set is determined. In one exemplary embodiment, a conversion rate difference can be calculated based on a first conversion rate statistic associated with the first object subset and a second conversion rate statistic associated with the second object subset. Based on the first conversion rate, the second conversion rate, and the conversion rate difference, an attribute classification result for the target object set is determined. This detailed classification of target objects provides reliable decision support for improving content push effectiveness and increasing content push revenue.

[0061] The first conversion rate statistic may include, for example, the mean conversion rate and median conversion rate of the target objects in the first object subset, and the second conversion rate statistic may include, for example, the mean conversion rate and median conversion rate of the target objects in the second object subset. The mean conversion rate may include, for example, the arithmetic mean conversion rate, the geometric mean conversion rate, the square mean conversion rate, and the like.

[0062] The difference between the first conversion rate statistic and the second conversion rate statistic is calculated to obtain a conversion rate difference. For example, an attribute classification result for the target object set can be determined based on the first conversion rate statistic, the second conversion rate statistic, and the conversion rate difference.

[0063] When the first conversion rate statistic is greater than a first preset threshold, the second conversion rate statistic is less than the first preset threshold, and the conversion rate difference is greater than a second preset threshold, the target object in the target object set may be determined to be a first type object. The first type object may be, for example, a marketing-sensitive object. The second preset threshold may be, for example, zero.

[0064] When the first conversion rate statistic is greater than the first preset threshold and the second conversion rate statistic is greater than the first preset threshold, the target object in the target object set may be determined to be a second type object. For example, the second type object may be a natural conversion type object.

[0065] When the first conversion rate statistic is less than the third preset threshold and the second conversion rate statistic is less than the third preset threshold, the target object in the target object set may be determined to be a third type object. For example, the third type object may be an indifferent object.

[0066] When the first conversion rate statistic is less than the first preset threshold, the second conversion rate statistic is greater than the first preset threshold, and the conversion rate difference is less than the second preset threshold, the target object in the target object set can be determined to be a fourth type object. The fourth type object can be, for example, a reaction type object.

[0067] In one example, for a first type of object, a first estimated conversion rate for the first type of object based on price content push conditions and a second estimated conversion rate for the first type of object based on non-price content push conditions can be determined. Based on the first estimated conversion rate and the second estimated conversion rate, it is determined whether the first type of object is price-sensitive. The pushed price content may include, for example, subsidy information, discount information, preferential information, etc. By carefully classifying the target objects, the content push effect can be effectively guaranteed, which is conducive to achieving a content traffic distribution with significant response effect.

[0068] In one example, content push parameters for a set of target objects can be determined based on the attribute targeting results. The content push parameters can indicate at least one of the following: whether to push content, the content to be pushed, and the content push method. The content push method can, for example, indicate information such as the content push medium and the content push time.

[0069] For example, content can be pushed to a first type of object to increase the probability of positive feedback from the first type of object based on the content push conditions. For price-sensitive objects within the first type of object, price content can be pushed to price-sensitive objects to increase the probability of positive feedback from price-sensitive objects based on the price content push conditions. For price-insensitive objects within the first type of object, other content can be pushed to price-insensitive objects to increase the probability of cross-product recommendations for price-insensitive objects based on the other content push conditions. Other content can include, for example, marketing reminders and fun marketing content.

[0070] In one example, when content push parameters indicate content push, the statistical values ​​of the image features of target objects in a set of target objects can be input into a trained random forest model, and multiple decision trees can be generated using the statistical values ​​of the image features as classification criteria. Based on the target leaf nodes in each decision tree that match the statistical values ​​of the image features, a decision result for the set of target objects is obtained. If the maximum depth of each decision tree is less than a preset depth threshold, the decision result indicates the target content to be pushed.

[0071] The random forest model is a cluster classification model that can include multiple independent decision trees. Each decision tree can contain multiple judgment nodes and multiple classification nodes. Judgment nodes are used to make attribute judgments based on input data, while classification nodes are used to make decisions and classifications based on the attribute judgments of adjacent judgment nodes.

[0072] The object profile features of the sample objects can be used as training samples for model training to obtain a random forest model. For example, model parameters of the random forest model to be trained can be obtained, and the model parameters can include, for example, a sample threshold and a depth threshold. Splitting is performed using the object profile features of the sample objects as classification conditions to generate a random forest model. The random forest model includes multiple decision trees, each with a maximum depth less than the depth threshold. The leaf nodes of each decision tree can be candidate push content, and the number of training samples corresponding to each candidate push content is greater than the sample threshold.

[0073] When the content push parameter indicates content push, a traffic allocation value for content push can be determined based on the profile feature distribution type of the target objects in the target object set. Based on the traffic allocation value, content is pushed to the target objects in the target object set. For example, when the profile feature is an age feature, the profile feature distribution type can be, for example, the age range.

[0074] For example, the traffic allocation value for content push is determined based on the target content type to be pushed to the target object set. The traffic allocation value indicates the proportion of content push objects in the target object set and also indicates the exposure ratio allocated to the target content to be pushed.

[0075] In one example, data on the response behavior of content push recipients based on the pushed content can be obtained. Based on this response behavior data, the percentage of conversion recipients among the content push recipients that provided positive feedback regarding the pushed content can be determined. Based on the percentage of conversion recipients and the content push cost, a content push gain can be determined, and traffic allocation can be adjusted based on the content push gain. Positive feedback behaviors regarding the pushed content can include, for example, installation, downloading, and transactional activities, though this embodiment does not limit these behaviors.

[0076] When the content push parameters indicate content push, response behavior data of target objects in the target object set can be obtained. Based on the response behavior data, the percentage of conversion objects in the target object set that have provided positive feedback regarding the pushed content is determined. Based on the percentage of conversion objects, the model parameters of the first prediction model are adjusted to obtain an adjusted first prediction model.

[0077] If the content push parameters indicate that content push is not to be performed, response behavior data of target objects in the target object set can be obtained. Based on the response behavior data, the proportion of conversion objects in the target object set that have provided positive feedback regarding the target content is determined. Based on the proportion of conversion objects, the model parameters of the second prediction model are adjusted to obtain an adjusted second prediction model.

[0078] Through the embodiment of the present disclosure, a first prediction model is used to output a first conversion rate of the target object based on content push conditions based on the object portrait data of the target object in the first object subset, and a second prediction model is used to output a second conversion rate of the target object based on non-content push conditions based on the object portrait data of the target object in the second object subset. Based on the first conversion rate and the second conversion rate, the attribute classification result for the target object set is determined. The accuracy of the object classification result can be effectively guaranteed, which is conducive to improving the efficiency of object classification and reducing the cost consumption of object classification. The sensitivity of the target object to the content to be pushed can be effectively identified. By accurately classifying the target object, it is conducive to making a more accurate estimate of the content push gain, which is conducive to achieving a more accurate and reasonable content push strategy.

[0079] Figure 4 The flowchart of the network model training method according to an embodiment of the present disclosure is schematically shown.

[0080] like Figure 4 As shown, the training method 400 may include, for example, operations S410 to S430.

[0081] In operation S410 , a model to be trained that matches each sample object is determined based on whether each sample object in the sample object set is a content push object.

[0082] In operation S420 , the object portrait data of each sample object is used as input data of the corresponding model to be trained to obtain a predicted conversion rate for each sample object.

[0083] In operation S430 , the model parameters of the corresponding model to be trained are adjusted according to the predicted conversion rate and the preset conversion label of each sample object to obtain a trained target network model.

[0084] The following examples illustrate the example process of each operation of the model training method of this embodiment.

[0085] Exemplarily, it is determined whether each sample object in the sample object set is a content push object. Based on whether each sample object in the sample object set is a content push object, the object portrait data of each sample object is used as input data for the corresponding to-be-trained model to obtain a predicted conversion rate for each sample object.

[0086] The model to be trained can be used to extract features from the object portrait data of the sample object, generating an initial feature vector based on at least one feature dimension. Based on the dimension weights of each feature dimension, a preset number of eigenvalues ​​with the largest dimension weights are filtered from the initial feature vector to generate a target feature vector. Based on the target feature vector, the predicted conversion rate of the sample object, either based on content push conditions or non-content push conditions, is output. This filtering of model training features can effectively address the computational overhead of machine learning models, accelerate model training, and improve the prediction accuracy of marketing conversion groups.

[0087] For example, during model training, a multi-way recall strategy and a chi-squared test algorithm can be used to screen the eigenvalues ​​with the largest weights for a preset number of dimensions to obtain a target eigenvector. The idea behind the chi-squared test is to determine whether the assumption of independence between variables holds true based on the deviation between the actual value and the hypothesized value (theoretical value obtained by assuming independence between variables).

[0088] For any target feature dimension in the at least one feature dimension, a precision gain coefficient of the to-be-trained model based on the target feature dimension may be determined, and a dimension weight matching the target feature dimension may be determined based on the precision gain coefficient.

[0089] For example, the object portrait data corresponding to the target feature dimension can be input into the corresponding to-be-trained model to obtain the first MSE (mean squared error) score of the to-be-trained model. Combined with the second MSE score when the object portrait data corresponding to the target feature dimension is not input into the to-be-trained model, the accuracy gain coefficient of the to-be-trained model based on the target feature dimension is determined.

[0090] The MSE score is used to measure the deviation between the model prediction value and the true value. The specific calculation formula can be: P i represents the predicted conversion rate for the i-th sample object, P i ′ represents the actual conversion rate of the i-th sample object, and the actual conversion rate can be 0 or 1, for example.

[0091] For example, the precision gain coefficient associated with the target feature dimension can be expressed by formula (1),

[0092]

[0093] MSE1 represents the first MSE score when the object profile data corresponding to the target feature dimension is not input into the training model, and MSE2 represents the second MSE score when the object profile data corresponding to the target feature dimension is input into the training model. The precision gain coefficient g can be used as the dimension weight of the target feature dimension. The dimension weight indicates the degree of influence of the target feature dimension on the predicted conversion rate. The target feature dimension can be any feature dimension from the at least one feature dimension.

[0094] For example, when the sample objects are content push targets, the object profile data of the sample objects can be used as input data for the first to-be-trained model to obtain a first predicted conversion rate for each sample object. Based on the first predicted conversion rate and first conversion label of the sample objects, the model parameters of the first to-be-trained model are adjusted to obtain a trained target network model as the first prediction model. The first predicted conversion rate indicates the probability of positive feedback for the corresponding sample object based on the content push conditions, and the first conversion label indicates the actual conversion status of the corresponding sample object based on the content push conditions.

[0095] If the sample object is not a content push target, the object profile data of the sample object is used as input data for the second to-be-trained model to obtain a second predicted conversion rate for each sample object. Based on the second predicted conversion rate and second conversion label of the sample object, the model parameters of the second to-be-trained model are adjusted to obtain a trained target network model as the second prediction model. The second predicted conversion rate indicates the probability of positive feedback for the sample object based on the non-content push condition, and the second conversion label indicates the actual conversion status of the corresponding sample object based on the non-content push condition.

[0096] Through the disclosed embodiments, a first prediction model is trained based on the object portrait data of the content push object to predict the first conversion rate of the target object based on the content push condition. A second prediction model is trained based on the object portrait data of the non-content push object to predict the second conversion rate of the target object based on the non-content push condition. This is helpful in assisting in analyzing the attribute classification results of the target object set, and can effectively improve the object classification accuracy and effectively improve the object classification efficiency. It is helpful in improving the prediction accuracy of the marketing conversion group and providing reliable decision support for improving the content push effect.

[0097] Figure 5 The figure schematically shows a training process of a network model according to an embodiment of the present disclosure.

[0098] like Figure 5As shown, content push objects 502A and non-content push objects 502B are determined in sample object set 501. Content push objects 502A may be sample objects to which marketing content has been pushed, and non-content push objects 502B may be sample objects to which marketing content has not been pushed.

[0099] For content push object 502A, a first conversion tag 503A and first object profile data 504A associated with content push object 502A are determined. First conversion tag 503A indicates the actual conversion status of the corresponding sample object based on the content push conditions. First object profile data 504A is used as input data for a first to-be-trained model 505A to obtain a first predicted conversion rate for content push object 502A. Based on the first predicted conversion rate and first conversion tag 503A, the model parameters of first to-be-trained model 505A are adjusted to obtain a first prediction model 506A.

[0100] For non-content push object 502B, a second conversion tag 503B and second object profile data 504B associated with non-content push object 502B are determined. Second conversion tag 503B indicates the actual conversion status of the corresponding sample object based on the non-content push condition. Second object profile data 504B is used as input data for a second to-be-trained model 505B to obtain a second predicted conversion rate for non-content push object 502B. Based on the second predicted conversion rate and second conversion tag 503B, the model parameters of the second to-be-trained model 505B are adjusted to obtain a second prediction model 506B.

[0101] For a set of target objects to be classified, the target object set can be divided into a first object subset and a second object subset, wherein the target objects in the target object set have a consistent distribution of profile features. A first prediction model 506A can be used to determine a first conversion rate 507A of the corresponding target objects based on content push conditions based on the object profile data of the target objects in the first object subset. A second prediction model 506B can be used to determine a second conversion rate 507B of the corresponding target objects based on non-content push conditions based on the object profile data of the target objects in the second object subset.

[0102] According to the first conversion rate 507A based on the content push condition and the second conversion rate 507B based on the non-content push condition, the attribute classification result 508 for the target object set can be determined. According to the attribute classification result 508, the content push parameter 509 for the target object set can be determined.

[0103] By introducing a prediction model based on object profile data for both content push recipients and non-content push recipients, we can achieve detailed classification of target object sets and effectively ensure the accuracy of object classification results. This helps provide reliable decision-making support for improving content push effectiveness and efficiency, and facilitates more accurate and targeted content push.

[0104] Figure 6 The block diagram schematically shows an object classification device according to an embodiment of the present disclosure.

[0105] like Figure 6 As shown, the object classification device 600 of the embodiment of the present disclosure includes, for example, a first processing module 610 , a second processing module 620 and a third processing module 630 .

[0106] The first processing module 610 is used to divide the target object set to be classified into a first object subset and a second object subset, and the portrait feature distribution of the target objects in the target object set is consistent; the second processing module 620 is used to determine the first conversion rate associated with the target objects in the first object subset, and determine the second conversion rate associated with the target objects in the second object subset; and the third processing module 630 is used to determine the attribute classification result for the target object set based on the first conversion rate and the second conversion rate, the first conversion rate indicates the positive feedback probability of the corresponding target object based on the content push condition, and the second conversion rate indicates the positive feedback probability of the corresponding target object based on the non-content push condition.

[0107] Through the embodiment of the present disclosure, the target object set to be classified is divided into a first object subset and a second object subset, the first conversion rate of the target objects in the first object subset based on the content push condition is determined, and the second conversion rate of the target objects in the second object subset based on the non-content push condition is determined, and the attribute classification result for the target object set is determined based on the first conversion rate and the second conversion rate. Attribute classification of the target object set based on the first conversion rate and the second conversion rate can effectively improve the accuracy of the object classification result, help improve the object classification efficiency, and help reduce the object classification cost. By determining the attribute classification result of the target object set, it is helpful to provide reliable decision support for improving the content push effect and reducing the content push cost.

[0108] According to an embodiment of the present disclosure, the second processing module includes: a first processing sub-module, for using a first prediction model to output a first conversion rate based on the object portrait data of the target object in the first object subset, and the first prediction model is trained based on the object portrait data of the content push object; and a second processing sub-module, for using a second prediction model to output a second conversion rate based on the object portrait data of the target object in the second object subset, and the second prediction model is trained based on the object portrait data of the non-content push object.

[0109] According to an embodiment of the present disclosure, the third processing module includes: a third processing sub-module, used to calculate the conversion probability difference based on the first conversion rate statistic associated with the first object subset and the second conversion rate statistic associated with the second object subset; and a fourth processing sub-module, used to determine the attribute classification result for the target object set based on the first conversion rate, the second conversion rate and the conversion probability difference.

[0110] According to an embodiment of the present disclosure, the device also includes a fourth processing module, which is used to: determine content push parameters for the target object set based on the attribute classification results, and the content push parameters indicate at least one of the following information: whether to push content, the content to be pushed, and the content push method.

[0111] According to an embodiment of the present disclosure, the device also includes a fifth processing module, which includes: a fifth processing sub-module, which is used to determine the traffic allocation value for content push based on the portrait feature distribution type of the target object in the target object set when the content push parameter indicates content push; and a sixth processing sub-module, which is used to push content to the target object in the target object set based on the traffic allocation value, and the traffic allocation value indicates the proportion of content push objects in the target object set.

[0112] According to an embodiment of the present disclosure, the device also includes a sixth processing module, which includes: a seventh processing sub-module, which is used to obtain response behavior data of the content push object based on the pushed content; an eighth processing sub-module, which is used to determine the proportion of conversion objects in the content push object that provide positive feedback on the pushed content based on the response behavior data; a ninth processing sub-module, which is used to determine the content push gain based on the proportion of conversion objects and the content push cost; and a tenth processing sub-module, which is used to adjust the traffic allocation value based on the content push gain.

[0113] According to an embodiment of the present disclosure, the device also includes a seventh processing module, which includes: an eleventh processing sub-module, which is used to input the portrait feature statistics of the target objects in the target object set into the trained random forest model when the content push parameters indicate content push, to obtain multiple decision trees generated based on the portrait feature statistics, wherein the maximum depth of each decision tree is less than a preset depth threshold; and a twelfth processing sub-module, which is used to obtain a decision result for the target object set based on the target leaf nodes in each decision tree that match the portrait feature statistics, and the decision result indicates the target content to be pushed.

[0114] Figure 7 A block diagram schematically shows a network model training device according to an embodiment of the present disclosure.

[0115] like Figure 7 As shown, the network model training device 700 of the embodiment of the present disclosure includes, for example, an eighth processing module 710 , a ninth processing module 720 and a tenth processing module 730 .

[0116] The eighth processing module 710 is used to determine the model to be trained that matches each sample object based on whether each sample object in the sample object set is a content push object; the ninth processing module 720 is used to use the object portrait data of each sample object as the input data of the corresponding model to be trained to obtain the predicted conversion rate for each sample object; and the tenth processing module 730 is used to adjust the model parameters of the corresponding model to be trained based on the predicted conversion rate of each sample object and the preset conversion label to obtain a trained target network model.

[0117] Through the disclosed embodiments, a first prediction model is trained based on the object portrait data of the content push object to predict the first conversion rate of the target object based on the content push condition. A second prediction model is trained based on the object portrait data of the non-content push object to predict the second conversion rate of the target object based on the non-content push condition. This is helpful in assisting in analyzing the attribute classification results of the target object set, and can effectively improve the object classification accuracy and effectively improve the object classification efficiency. It is helpful in improving the prediction accuracy of the marketing conversion group and providing reliable decision support for improving the content push effect.

[0118] According to an embodiment of the present disclosure, the ninth processing module includes: a thirteenth processing sub-module, which is used to use the model to be trained to perform feature extraction on the object portrait data of the sample object, and obtain an initial feature vector based on at least one feature dimension; a fourteenth processing sub-module, which is used to screen a preset number of eigenvalues ​​with the largest dimension weights from the initial feature vector according to the dimension weights of each feature dimension, and obtain a target feature vector; and a fifteenth processing sub-module, which is used to output the predicted conversion rate for the sample object based on the target feature vector, and the dimension weight indicates the degree of influence of the corresponding feature dimension on the predicted conversion rate.

[0119] According to an embodiment of the present disclosure, the device also includes an eleventh processing module, which includes: a sixteenth processing sub-module, for determining the precision gain coefficient of the model to be trained based on the target feature dimension for any target feature dimension in at least one feature dimension; and a seventeenth processing sub-module, for determining the dimension weight matching the target feature dimension based on the precision gain coefficient.

[0120] According to an embodiment of the present disclosure, the tenth processing module includes: an eighteenth processing sub-module, which is used to adjust the model parameters of the first model to be trained according to the first predicted conversion rate and the first conversion label of the sample object when the sample object is a content push object, so as to obtain a trained target network model as the first prediction model; and a nineteenth processing sub-module, which is used to adjust the model parameters of the second model to be trained according to the second predicted conversion rate and the second conversion label of the corresponding sample object when the sample object is a non-content push object, so as to obtain a trained target network model as the second prediction model, the first predicted conversion rate indicating the positive feedback probability of the corresponding sample object based on the content push condition, and the second predicted conversion rate indicating the positive feedback probability of the corresponding sample object based on the non-content push condition.

[0121] It should be noted that the information collection, storage, use, processing, transmission, provision and disclosure involved in the technical solutions disclosed herein are in compliance with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0122] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium, and a computer program product.

[0123] Figure 8 A block diagram of an electronic device for object classification according to an embodiment of the present disclosure is schematically shown.

[0124] Figure 8A schematic block diagram of an example electronic device 800 that can be used to implement embodiments of the present disclosure is shown. The electronic device 800 is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0125] like Figure 8 As shown, the device 800 includes a computing unit 801, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 802 or a computer program loaded from a storage unit 808 into a random access memory (RAM) 803. Various programs and data required for the operation of the device 800 can also be stored in the RAM 803. The computing unit 801, the ROM 802, and the RAM 803 are connected to each other via a bus 804. An input / output (I / O) interface 805 is also connected to the bus 804.

[0126] Various components in device 800 are connected to I / O interface 805, including an input unit 806, such as a keyboard, mouse, etc.; an output unit 807, such as various types of displays, speakers, etc.; a storage unit 808, such as a magnetic disk, optical disk, etc.; and a communication unit 809, such as a network card, modem, wireless communication transceiver, etc. The communication unit 809 allows device 800 to exchange information / data with other devices via a computer network such as the Internet and / or various telecommunication networks.

[0127] The computing unit 801 can be a variety of general-purpose and / or specialized processing components with processing and computing capabilities. Some examples of the computing unit 801 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various specialized artificial intelligence (AI) computing chips, various computing units that run deep learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 801 performs the various methods and processes described above, such as the object classification method. For example, in some embodiments, the object classification method can be implemented as a computer software program that is tangibly contained in a machine-readable medium, such as the storage unit 808. In some embodiments, part or all of the computer program can be loaded and / or installed on the device 800 via the ROM 802 and / or the communication unit 809. When the computer program is loaded into the RAM 803 and executed by the computing unit 801, one or more steps of the object classification method described above can be performed. Alternatively, in other embodiments, the computing unit 801 can be configured to perform the object classification method by any other suitable means (e.g., by means of firmware).

[0128] Various embodiments of the systems and techniques described above can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), system-on-chip systems (SOCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include being implemented in one or more computer programs that are executable and / or interpreted on a programmable system that includes at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0129] The program code for implementing the method of the present disclosure can be written in any combination of one or more programming languages. These program codes can be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable model training device so that when the program code is executed by the processor or controller, the functions / operations specified in the flow chart and / or block diagram are implemented. The program code can be executed entirely on the machine, partially on the machine, as a stand-alone software package, partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0130] In the context of the present disclosure, a machine-readable medium can be a tangible medium that can contain or store a program for use by or in conjunction with an instruction execution system, device or equipment. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium can include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0131] To provide interaction with an object, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the object; and a keyboard and pointing device (e.g., a mouse or trackball) through which the object can provide input to the computer. Other types of devices can also be used to provide interaction with the object; for example, the feedback provided to the object can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the object can be received in any form (including acoustic input, voice input, or tactile input).

[0132] The systems and techniques described herein can be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., an object computer having a graphical object interface or a web browser through which the object can interact with an embodiment of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), and the Internet.

[0133] A computer system may include a client and a server. The client and server are generally remote from each other and typically interact through a communication network. The client-server relationship arises through computer programs running on the respective computers and having a client-server relationship with each other. The server may be a cloud server, a server in a distributed system, or a server integrated with a blockchain.

[0134] It should be understood that the various forms of the processes shown above can be used to reorder, add, or delete steps. For example, the steps described in this disclosure can be performed in parallel, sequentially, or in a different order, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved. This is not limited herein.

[0135] The above specific embodiments do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art will appreciate that various modifications, combinations, sub-combinations, and substitutions may be made based on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure shall be included within the scope of protection of this disclosure.

Claims

1. A method for object classification, comprising: Dividing a target object set to be classified into a first object subset and a second object subset, wherein the image feature distributions of the target objects in the target object set are consistent; determining a first conversion rate associated with target objects in the first subset of objects, and determining a second conversion rate associated with target objects in the second subset of objects; and determining an attribute classification result for the target object set according to the first conversion rate and the second conversion rate; The first conversion rate indicates the probability of positive feedback of the corresponding target object based on the content push condition, and the second conversion rate indicates the probability of positive feedback of the corresponding target object based on the non-content push condition; Determining a first conversion rate associated with a target object in the first subset of objects, and determining a second conversion rate associated with a target object in the second subset of objects, includes: Outputting the first conversion rate using a first prediction model based on the object portrait data of the target object in the first object subset, wherein the first prediction model is trained based on the object portrait data and conversion labels of the content push object; and The second conversion rate is outputted based on the object portrait data of the target object in the second object subset using a second prediction model, wherein the second prediction model is trained based on the object portrait data and conversion labels of the non-content push object.

2. The method according to claim 1, wherein Determining the attribute classification result for the target object set according to the first conversion rate and the second conversion rate includes: Calculating a conversion probability difference based on a first conversion rate statistic associated with the first subset of objects and a second conversion rate statistic associated with the second subset of objects; and An attribute classification result for the target object set is determined according to the first conversion rate, the second conversion rate, and the conversion probability difference.

3. The method according to claim 1 or 2, further comprising: Determine content push parameters for the target object set based on the attribute classification result, wherein the content push parameters indicate at least one of the following information: Whether to push content, the content to be pushed, and the content push method.

4. The method according to claim 3, further comprising: In the case where the content push parameter indicates content push, determining a traffic allocation value for content push according to a distribution type of a portrait feature of a target object in the target object set; as well as Based on the traffic allocation value, content is pushed to the target objects in the target object set, wherein the traffic allocation value indicates a proportion of the content push objects in the target object set.

5. The method according to claim 4, further comprising: Obtaining response behavior data of the content push object based on the pushed content; Determining, based on the response behavior data, a proportion of conversion objects among the content push objects that provide positive feedback on the pushed content; Determining a content push gain based on the conversion target ratio and the content push cost; and Based on the content push gain, the traffic allocation value is adjusted.

6. The method according to claim 3, further comprising: When the content push parameter indicates content push, inputting the portrait feature statistics of the target objects in the target object set into a trained random forest model to obtain a plurality of decision trees generated based on the portrait feature statistics, wherein the maximum depth of each decision tree is less than a preset depth threshold; and According to the target leaf nodes in each decision tree that match the statistical value of the portrait feature, a decision result for the target object set is obtained, and the decision result indicates the target content to be pushed.

7. A method for training a network model, comprising: Determining a model to be trained that matches each sample object in the sample object set according to whether each sample object in the sample object set is a content push object; Using the object portrait data of each of the sample objects as input data of the corresponding model to be trained, and obtaining a predicted conversion rate for each of the sample objects; as well as Adjusting the model parameters of the corresponding model to be trained according to the predicted conversion rate and the preset conversion label of each of the sample objects to obtain a trained target network model; The step of adjusting the model parameters of the corresponding model to be trained according to the predicted conversion rate and the preset conversion label of each sample object to obtain a trained target network model includes: When the sample object is a content push object, adjusting the model parameters of the first to-be-trained model according to the first predicted conversion rate and the first conversion label of the sample object to obtain a trained target network model as the first prediction model; and In the case where the sample object is not a content push object, the model parameters of the second to-be-trained model are adjusted according to the second predicted conversion rate and the second conversion label of the corresponding sample object to obtain a trained target network model as the second prediction model. The first predicted conversion rate indicates the positive feedback probability of the corresponding sample object based on the content push condition, and the second predicted conversion rate indicates the positive feedback probability of the corresponding sample object based on the non-content push condition.

8. The method according to claim 7, wherein: The object portrait data of each of the sample objects is used as input data of the corresponding model to be trained to obtain a predicted conversion rate for each of the sample objects, including: Using the model to be trained, extracting features from the object portrait data of the sample object to obtain an initial feature vector based on at least one feature dimension; According to the dimension weights of the feature dimensions, a preset number of eigenvalues ​​with the largest dimension weights are selected from the initial feature vector to obtain a target feature vector; and Based on the target feature vector, output the predicted conversion rate for the sample object, The dimension weight indicates the degree of influence of the corresponding feature dimension on the predicted conversion rate.

9. The method according to claim 8, further comprising: For any target feature dimension among the at least one feature dimension, determining a precision gain coefficient of the to-be-trained model based on the target feature dimension; as well as According to the precision gain coefficient, a dimension weight matching the target feature dimension is determined.

10. An object classification device, comprising: A first processing module is configured to divide a target object set to be classified into a first object subset and a second object subset, wherein the image feature distribution of the target objects in the target object set is consistent; a second processing module configured to determine a first conversion rate associated with a target object in the first subset of objects, and to determine a second conversion rate associated with a target object in the second subset of objects; and A third processing module is configured to determine an attribute classification result for the target object set based on the first conversion rate and the second conversion rate. The first conversion rate indicates the probability of positive feedback of the corresponding target object based on the content push condition, and the second conversion rate indicates the probability of positive feedback of the corresponding target object based on the non-content push condition; The second processing module includes: a first processing submodule, configured to output the first conversion rate using a first prediction model based on the object portrait data of the target object in the first object subset, wherein the first prediction model is trained based on the object portrait data and conversion labels of the content push object; and The second processing submodule is used to use a second prediction model to output the second conversion rate based on the object portrait data of the target object in the second object subset, wherein the second prediction model is trained based on the object portrait data and conversion labels of the non-content push object.

11. The device according to claim 10, wherein The third processing module includes: a third processing submodule, configured to calculate a conversion probability difference based on a first conversion rate statistic associated with the first object subset and a second conversion rate statistic associated with the second object subset; and The fourth processing submodule is configured to determine an attribute classification result for the target object set according to the first conversion rate, the second conversion rate, and the conversion probability difference.

12. The apparatus according to claim 10 or 11, further comprising a fourth processing module, configured to: Determine content push parameters for the target object set based on the attribute classification result, wherein: The content push parameter indicates at least one of the following information: Whether to push content, the content to be pushed, and the content push method.

13. The apparatus according to claim 12, further comprising a fifth processing module, The fifth processing module includes: A fifth processing submodule is configured to determine a traffic allocation value for content push according to a distribution type of a portrait feature of a target object in the target object set when the content push parameter indicates content push; as well as A sixth processing submodule is configured to push content to target objects in the target object set based on the traffic allocation value, wherein the traffic allocation value indicates a proportion of content push objects in the target object set.

14. The apparatus according to claim 13, further comprising a sixth processing module, The sixth processing module includes: a seventh processing submodule, configured to obtain response behavior data of the content push object based on the pushed content; an eighth processing submodule, configured to determine, based on the response behavior data, a proportion of conversion objects among the content push objects that provide positive feedback on the pushed content; a ninth processing submodule, configured to determine a content push gain based on the conversion target ratio and the content push cost; as well as A tenth processing submodule is configured to adjust the traffic allocation value based on the content push gain.

15. The apparatus according to claim 12, further comprising a seventh processing module, The seventh processing module includes: an eleventh processing submodule, configured to, when the content push parameter indicates content push, input the portrait feature statistics of the target objects in the target object set into a trained random forest model to obtain a plurality of decision trees generated based on the portrait feature statistics, wherein a maximum depth of each of the decision trees is less than a preset depth threshold; and The twelfth processing submodule is used to obtain a decision result for the target object set based on the target leaf node in each decision tree that matches the portrait feature statistical value, and the decision result indicates the target content to be pushed.

16. A network model training device, comprising: an eighth processing module, configured to determine, based on whether each sample object in the sample object set is a content push object, a model to be trained that matches each sample object; a ninth processing module, configured to use the object portrait data of each sample object as input data of a corresponding model to be trained, and obtain a predicted conversion rate for each sample object; as well as a tenth processing module, configured to adjust the model parameters of the corresponding model to be trained according to the predicted conversion rate and the preset conversion label of each sample object, to obtain a trained target network model; The tenth processing module includes: An eighteenth processing submodule is configured to, when the sample object is a content push object, adjust the model parameters of the first to-be-trained model according to the first predicted conversion rate and the first conversion label of the sample object to obtain a trained target network model as the first prediction model; and The nineteenth processing submodule is configured to adjust the model parameters of the second to-be-trained model according to the second predicted conversion rate and the second conversion label of the corresponding sample object when the sample object is not a content push object, to obtain a trained target network model as the second prediction model. The first predicted conversion rate indicates the positive feedback probability of the corresponding sample object based on the content push condition, and the second predicted conversion rate indicates the positive feedback probability of the corresponding sample object based on the non-content push condition.

17. The device according to claim 16, wherein The ninth processing module includes: A thirteenth processing submodule is configured to perform feature extraction on the object portrait data of the sample object using the to-be-trained model to obtain an initial feature vector based on at least one feature dimension; A fourteenth processing submodule is configured to filter a preset number of eigenvalues ​​with the largest dimensional weights from the initial eigenvector according to the dimensional weights of the respective feature dimensions, to obtain a target eigenvector; and A fifteenth processing submodule is configured to output the predicted conversion rate for the sample object based on the target feature vector. The dimension weight indicates the degree of influence of the corresponding feature dimension on the predicted conversion rate.

18. The apparatus according to claim 17, further comprising an eleventh processing module, The eleventh processing module includes: A sixteenth processing submodule is configured to determine, for any target feature dimension among the at least one feature dimension, a precision gain coefficient of the to-be-trained model based on the target feature dimension; as well as The seventeenth processing submodule is used to determine the dimension weight that matches the target feature dimension according to the precision gain coefficient.

19. An electronic device comprising: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the object classification method described in any one of claims 1 to 6, or execute the network model training method described in any one of claims 7 to 9.

20. A non-transitory computer-readable storage medium storing computer instructions, wherein: The computer instructions are used to enable the computer to execute the object classification method according to any one of claims 1 to 6, or to execute the network model training method according to any one of claims 7 to 9.

21. A computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the computer program implements the object classification method according to any one of claims 1 to 6, or implements the network model training method according to any one of claims 7 to 9.

Images