Transaction data value estimation method and device for information promotion bit, medium and equipment

Abstract

The invention provides a transaction data value estimation method and device for an information promotion bit, a storage medium, a product and electronic equipment, and the method is applied to the technical field of computers, and comprises the steps: obtaining target user feature information corresponding to a target object, and employing a conversion rate prediction model and an uncertainty quantification model, and obtaining a target predicted conversion rate based on the target user feature information, and obtaining a target transaction data output value of the target object for the information promotion bit based on the target predicted conversion rate, an estimated click rate corresponding to the target object and a transaction data issuing strategy.

Description

Technical Field

[0001] This application relates to the field of computer technology, and more particularly, to a method, apparatus, storage medium, and electronic device for predicting transaction data values of information promotion positions in the field of computer technology. Background Art

[0002] Placing promotional information is an important way for merchants to enhance the popularity and sales volume of their products. However, the number of information promotion positions available to merchants is limited. In the prior art, auctions are often used to determine the ownership of information promotion positions. Different merchants have different costs and benefits for different information promotion positions, and merchants need to consider multiple factors when participating in the auction. Due to the rapid development of the Internet, users can browse a large number of web pages and view a large number of applications every day, generating numerous information promotion positions. Therefore, merchants need to participate in auctions for numerous information promotion positions in a short period of time, resulting in low auction efficiency for information promotion positions. There is a need to provide a method that can improve the efficiency of solving the problem of information promotion position ownership. Summary of the Invention

[0003] Embodiments of this application provide a method, apparatus, storage medium, and electronic device for predicting transaction data values of information promotion positions. This method can predict the target transaction data output value for information promotion positions through a conversion rate prediction model and an uncertainty quantification model, improve the efficiency of solving the problem of information promotion position ownership, and make the target transaction data output value conform to the actual situation of the target object.

[0004] In a first aspect, embodiments of this application provide a method for predicting transaction data values of information promotion positions, the method comprising:

[0005] Obtaining target user feature information corresponding to a target object;

[0006] Using a conversion rate prediction model and an uncertainty quantification model, obtaining a target predicted conversion rate based on the target user feature information;

[0007] Based on the target predicted conversion rate, the estimated click-through rate corresponding to the target object, and a transaction data distribution strategy, obtaining a target transaction data output value of the target object for an information promotion position.

[0008] In a second aspect, embodiments of this application provide an apparatus for predicting transaction data values of information promotion positions, the apparatus comprising:

[0009] A feature information acquisition unit, configured to obtain target user feature information corresponding to a target object;

[0010] A conversion rate prediction unit, configured to use a conversion rate prediction model and an uncertainty quantification model to obtain a target predicted conversion rate based on the target user characteristic information;

[0011] An output value prediction unit, configured to obtain a target transaction data output value of the target object for the information promotion position based on the target predicted conversion rate, the estimated click-through rate corresponding to the target object, and a transaction data distribution policy.

[0012] In a third aspect, an embodiment of the present application provides a computer storage medium, which stores multiple instructions, and the instructions are adapted to be loaded and executed by a processor to perform the above method steps.

[0013] In a fourth aspect, an embodiment of the present application provides a computer program product, which stores multiple instructions, and the instructions are adapted to be loaded and executed by a processor to perform the above method steps.

[0014] In a fifth aspect, an embodiment of the present application provides an electronic device, which may include: a processor and a memory; wherein, the memory stores a computer program, and the computer program is adapted to be loaded and executed by the processor to perform the above method steps.

[0015] In one or more embodiments of the present application, target user characteristic information corresponding to a target object is obtained, a conversion rate prediction model and an uncertainty quantification model are used to obtain a target predicted conversion rate based on the target user characteristic information, and a target transaction data output value of the target object for the information promotion position is obtained based on the target predicted conversion rate, the estimated click-through rate corresponding to the target object, and a transaction data distribution policy. By using a conversion rate prediction model and an uncertainty quantification model to predict the target transaction data output value for the information promotion position, the solution efficiency of the information promotion position attribution problem is improved, and the target transaction data output value conforms to the actual situation of the target object. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present application, and those of ordinary skill in the art can obtain other drawings without creative efforts based on these drawings.

[0017] Figure 1 is an example schematic diagram of the attribution of an information promotion position provided by an embodiment of the present application;

[0018] Figure 2 is a flowchart of a method for estimating the transaction data value of an information promotion position provided by an embodiment of the present application;

[0019] Figure 3 It is a schematic flow chart for training a conversion rate prediction model and an uncertainty quantification model provided by an embodiment of the present application;

[0020] Figure 4 It is a schematic flow chart for sample preprocessing provided by an embodiment of the present application;

[0021] Figure 5 It is a schematic flow chart for obtaining a target predicted conversion rate provided by an embodiment of the present application;

[0022] Figure 6 It is a schematic flow chart for obtaining an uncertainty quantification value provided by an embodiment of the present application;

[0023] Figure 7 It is a schematic flow chart for conversion rate adjustment provided by an embodiment of the present application;

[0024] Figure 8 It is a schematic flow chart for calculating a transaction data output value provided by an embodiment of the present application;

[0025] Figure 9 It is a schematic structural diagram of a transaction data value estimation device provided by an embodiment of the present application;

[0026] Figure 10 It is a schematic structural diagram of a transaction data value estimation device provided by an embodiment of the present application;

[0027] Figure 11 It is a schematic structural diagram of an electronic device provided by an embodiment of the present application. Detailed implementation manners

[0028] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present application without creative efforts shall fall within the protection scope of the present application.

[0029] Real-Time Bidding (RTB) is an information promotion technology that can be used to solve the attribution problem of information promotion positions online. It enables merchants, platforms, and other entities to obtain information promotion positions through real-time bidding at the moment when Internet users access websites or use applications, and then place promotional information on these positions. The website or application that generates the information promotion position is the provider of the information promotion position. Among them, the promotional information can be information used to enable more users to understand or pay attention to a certain thing and guide users to be interested in that thing. The promotional information can be in various forms, such as text, pictures, or videos. For example, the promotional information can be an advertisement. For instance, merchants and other entities can place advertisements on the information promotion positions provided by media platforms on the Internet to convey commercial information and promote the promotion of products or services. Please also refer to Figure 1 , which provides an example schematic diagram of the attribution of information promotion positions for the embodiments of this application. When a user accesses a website or uses an application through the terminal device they hold, an information promotion position can be generated. For example, when a user opens an application, a splash screen advertisement position can be generated during the startup of the application, and the splash screen advertisement position can be an information promotion position. After the information promotion position is generated, various entities can conduct real-time bidding for this position. The entities can be those that need information promotion positions, merchants, platforms, etc. that need to place promotional information on the information promotion positions. Each entity can provide a transaction data output value. Among them, the transaction data can be assets, such as tradable goods like goods, or rights relationships such as bonds and stocks that can be transferred. The transaction data output value is the amount of transaction data that the entity can provide. For example, entity A can provide a transaction data output value A for the information promotion position, entity B can provide a transaction data output value B for the information promotion position, and entity C can provide a transaction data output value C. The provider of the information promotion position can determine the attribution of the information promotion position by comparing the transaction data output values. For example, if the transaction data output value B is the largest, the provider of the information promotion position can make the information promotion position belong to entity B, and entity B can then place promotional information on this position.

[0030] It can be understood that when determining the transaction data output value, each object can consider maximizing the goal and some high-level goals, such as budget constraints, key performance indicators (KPIs), and return on advertising spend (ROAS) of the object for promotional information. If the transaction data output value provided by the object is too low, it will not be able to obtain an information promotion position and miss the opportunity to place promotional information. If the transaction data output value provided by the object is too high, even if it obtains an information promotion position, the return brought by the information promotion position may be much lower than the transaction data output value paid. For the large number of real-time bidding events for information promotion positions, it is difficult for each object to quickly determine a suitable transaction data output value for each information promotion position, resulting in low efficiency in solving the problem of information promotion position ownership and may also cause the object to be unable to obtain an information promotion position with a suitable transaction data output value. The transaction data value estimation device provided by the embodiments of the present application can calculate a suitable transaction data output value for the object based on the object's previous promotional information when generating an information promotion position. The transaction data value estimation method for the information promotion position provided by the embodiments of the present application can be implemented depending on a computer program and can run on a transaction data value estimation device for an information promotion position based on the von Neumann architecture. This computer program can be integrated into an application or run as an independent tool-type application.

[0031] The following specifically describes the transaction data value estimation method for the information promotion position provided by the present application with reference to specific embodiments.

[0032] Please refer to Figure 2 , which is a schematic flowchart of a transaction data value estimation method for an information promotion position provided by an embodiment of the present application. As Figure 2 shown, the method of the embodiment of the present application may include the following steps S102-S106.

[0033] S102, obtaining target user characteristic information corresponding to a target object.

[0034] Specifically, when a target object wants to obtain a generated information promotion position, a transaction data value estimation device can be used to calculate a target transaction data output value, which is the transaction data output value calculated by the transaction data value estimation device for the target object and suitable for the information promotion position. Among them, the target object can be a manufacturer, a platform, a commodity or service provided by the manufacturer or platform, etc. The transaction data value estimation device can obtain target user characteristic information corresponding to the target object. The target user characteristic information is the user characteristic information used to calculate the transaction data output value for the target object this time, which can reflect the attributes and preferences of the user and the effect of the target object's previous delivery of promotional information to the user.

[0035] Optionally, the user feature information may include user features and promotion information features corresponding to the user features. The user features in the target user feature information may be the feature information of the user corresponding to the target object and / or the user corresponding to the information promotion position provider, and may include the user's basic profile and the user's historical interaction sequence. For example, the user's basic profile may include the user's gender, age, occupation, region, or user account, etc. The user's historical interaction sequence may be the interaction behavior of the user with the target object and / or the information promotion position provider within a certain period of time in the past, arranged in chronological order to form a sequence. For example, it may include a click sequence and a purchase sequence. The click sequence may be a sequence of promotion information, products, or services clicked by the user, and the purchase sequence may be a sequence of products or services purchased by the user. The user's historical interaction sequence reflects the user's behavior habits, interest preferences, and dynamic changes. The promotion information feature may be the feature information of the promotion information that has been delivered to the user corresponding to the user feature information. The promotion information feature may include the name of the object to which the promotion information belongs, the category of the promotion information, and interaction statistical features, etc.

[0036] S104. Using a conversion rate prediction model and an uncertainty quantification model, obtain a target predicted conversion rate based on the target user feature information.

[0037] Specifically, the transaction data value estimation device may input the target user feature information into the conversion rate prediction model and the uncertainty quantification model, so as to obtain the target predicted conversion rate. The target predicted conversion rate is predicted by the transaction data value estimation device, which is the conversion rate of the target object after the promotion information is delivered at this information promotion position. The conversion rate is the proportion of users who have seen the promotion information and actually completed the conversion behavior. The conversion behavior may be set by the object to which the promotion information belongs and is related to the types, attributes, etc. of the products or services corresponding to the promotion information. For example, it may include clicking on the promotion information, registering an account, purchasing products or services, downloading an application program, etc.

[0038] Since the actual scenario usually changes continuously, the conversion rate predicted by the conversion rate prediction model based on historical data such as target feature information may deviate from the optimal solution. Therefore, the transaction data estimation device can adjust the conversion rate predicted by the conversion rate prediction model in combination with the uncertainty quantification model to obtain the final target predicted conversion rate. Uncertainty Quantification (UQ) technology is a tool for evaluating the impact of uncertainty in complex systems. It is particularly important when dealing with models containing multiple uncertainty factors. These uncertainties may stem from dataset noise, the incompleteness of the model structure, and changes in the external environment. The uncertainty quantification technology aims to enable the deep learning model to give the confidence level of the predicted value while giving the predicted value, that is, the uncertainty quantification value. The uncertainty quantification model can give a lower uncertainty quantification value for in-distribution (ID) samples and a higher uncertainty quantification value for out-of-distribution (OOD) samples. Based on the uncertainty quantification technology, the predicted conversion rate of OOD samples can be calibrated to improve the accuracy of the target predicted conversion rate.

[0039] S106. Based on the target predicted conversion rate, the estimated click-through rate corresponding to the target object, and the transaction data distribution strategy, obtain the target transaction data output value of the target object for the information promotion position.

[0040] Specifically, the transaction data value estimation device can obtain the estimated click-through rate corresponding to the target object and the transaction data distribution strategy. The estimated click-through rate is the estimated click-through rate calculated based on the click-through rate of the promotion information previously placed by the target object, and is used to represent the probability that the promotion information placed by the target object is clicked and viewed by the user. The estimated click-through rate can be calculated by the transaction data value estimation device or provided by the target object. The transaction data distribution strategy can be pre-set by the target object or the transaction data value estimation device, indicating how to adjust the transaction data output value according to the target predicted conversion rate. For example, the higher the target predicted conversion rate, the higher the success rate of the user completing the conversion behavior after placing the promotion information, so the transaction data output value can also be increased. Conversely, the lower the target predicted conversion rate, the lower the success rate of the user completing the conversion behavior after placing the promotion information, so the transaction data output value can be decreased.

[0041] The transaction data value estimation device can calculate the target transaction data output value of the target object for the information promotion position based on the target predicted conversion rate, the estimated click-through rate corresponding to the target object, and the transaction data distribution strategy. The target object can use the target transaction data output value to conduct real-time bidding for the information promotion position.

[0042] In an embodiment of the present application, the target user feature information corresponding to the target object is obtained, and the conversion rate prediction model and the uncertainty quantification model are used to obtain the target predicted conversion rate based on the target user feature information. Based on the target predicted conversion rate, the estimated click-through rate corresponding to the target object, and the transaction data distribution strategy, the target transaction data output value of the target object for the information promotion position is obtained. By using the conversion rate prediction model and the uncertainty quantification model to predict the target transaction data output value for the information promotion position, the solution efficiency of the information promotion position attribution problem is improved, and the target transaction data output value conforms to the actual situation of the target object.

[0043] Please refer to Figure 3 , which is a schematic flowchart of the training of a conversion rate prediction model and an uncertainty quantification model provided by an embodiment of the present application. In one or more embodiments of the present application, the following steps S202-S210 may be included before step S104.

[0044] S202, create an initial conversion rate prediction model and an initial uncertainty quantification model, and obtain the sample user feature information and the conversion label corresponding to each sample user feature information.

[0045] Specifically, the transaction data value estimation device may create an initial conversion rate prediction model and an initial uncertainty quantification model. The structures of the initial conversion rate prediction model and the initial uncertainty quantification model are both deep neural networks. Then the transaction data estimation device may obtain a sample data set for model training of the initial conversion rate prediction model and the initial uncertainty quantification model. The sample data set may include the sample user feature information and the conversion label corresponding to each sample user feature information, where the conversion label is used to indicate whether the user in the sample user feature information has successfully completed the conversion behavior.

[0046] Optionally, the initial conversion rate prediction model may be a conversion rate prediction model (Conversion Rate Prediction Model, CVR). The CVR model may be used to predict the probability that a user completes the conversion behavior after clicking on the promotion information or entering the promotion information page. The model structure consists of a bottom embedding matrix, a sequence modeling module, and a top binary classification sub-network (deep neural network). The initial uncertainty quantification model may adopt an ensemble structure and may be composed of multiple member models. The model structures of each member model are the same and are all simplified conversion rate prediction models, which consist of a bottom Embedding matrix and a top binary classification sub-network (deep neural network). The parameter initializations of each member model are different, and the shuffling operation of the data set during the model training process also brings randomness, further ensuring the uncertainty quantification ability.

[0047] Optionally, the transaction data value estimation device may obtain a sample data set for model training from online logs. The form of the sample data set may be {(x 1 ,y 1 ),(x 2 ,y 2 ),…,(x N ,y N )}, where N is the total number of samples in the sample data set. x i represents the sample user feature information in the i-th sample data, which may include user features, promotion information features, and scenario features. User features may include user basic portraits and user historical interaction sequences. For example, user basic portraits may include user gender, age, occupation, region, or user accounts, etc. Promotion feature information may include the feature information of the promotion information sent to the user. Promotion information features may include the name of the object to which the promotion information belongs, the category of the promotion information, and interaction statistical features, etc. Scenario features are the feature information of the scenario where the promotion information is sent to the user. For example, it may include scenario statistical features, scenario ID, etc. y i represents the label corresponding to whether the conversion behavior is successfully completed. y i ∈ {0, 1}. y i = 0 indicates that the conversion behavior is not completed, and y i = 1 indicates that the conversion behavior is completed.

[0048] S204. Input the sample user feature information into the initial conversion rate prediction model to obtain the training prediction conversion rate output by the initial conversion rate prediction model.

[0049] Specifically, the transaction data value estimation device may input the sample user feature information into the initial conversion rate prediction model. The initial conversion rate prediction model may perform conversion rate prediction based on the sample user feature information and obtain the training prediction conversion rate output by the initial conversion rate prediction model.

[0050] Optionally, the transaction data value estimation device may use the batch processing method to train the initial conversion rate prediction model. It may first randomly sample a batch of samples from the sample data where the batch size N bs may be preset by the transaction data value estimation device. Then, input the N bs sample user feature information in the batch sample B into the initial conversion rate prediction model.

[0051] S206. Based on the training prediction conversion rate and the conversion label, perform parameter adjustment processing on the initial conversion rate prediction model until the model training is completed to obtain the conversion rate prediction model.

[0052] Specifically, the transaction data value prediction device can calculate a loss function based on the training prediction conversion rate and the conversion label, and adjust the parameters of the initial conversion rate prediction model during the backpropagation training process based on the loss function until the initial conversion rate prediction model completes model training to obtain a conversion rate prediction model.

[0053] Optionally, the loss function can include a conversion rate prediction loss function and a regularization constraint. Among them, the conversion rate prediction loss function L cvr is calculated by the following formula:

[0054]

[0055] where p i ∈[0,1] is the training prediction conversion rate, y i ∈{0,1} is the conversion label, and the conversion rate prediction loss function L cvr combined with the regularization constraint can obtain the loss function, and the formula is as follows:

[0056] L = L cvr + η * L reg

[0057] where L reg is the regularization term, η is the regularization term coefficient preset by the transaction data value prediction device, which is used to adjust the constraint strength of the regularization term. By combining the conversion rate prediction loss function and the regularization constraint, while accurately fitting the sample data during the model training process, it can also avoid overfitting through the regularization term, better handle outliers and noise in the sample data, and balance the accuracy and stability of model prediction.

[0058] S208. Input the sample user feature information into the initial uncertainty quantification model to obtain the training uncertainty quantification value output by the initial uncertainty quantification model.

[0059] Specifically, the transaction data value prediction device can input the sample user feature information into the initial uncertainty quantification model. The initial uncertainty quantification model can score the confidence of the prediction conversion rate based on the sample user feature information, and can obtain the training uncertainty quantification value output by the initial uncertainty quantification model.

[0060] Optionally, the transaction data value prediction device can also train the initial uncertainty quantification model in a batch processing manner. Similar to step S204, it can first randomly sample a batch of samples from the sample data and then input the N bs sample user feature information in the batch sample B into the initial uncertainty quantification model.

[0061] Optionally, for the Embedding matrix of each member model in the initial uncertainty quantification model, if there are N model member models in the initial uncertainty quantification model, then expand its embedding dimension to D emb times the original dimension, that is, N model *D emb . For a multi-layer perceptron (MLP), expand its parameters from a two-dimensional matrix to a three-dimensional tensor, that is, from N in *N out to N model *N in *N out . Construct a computational graph through a fusion method for unified training and inference, thereby reducing the computational and memory resource overhead while maintaining the uncertainty quantification ability of the model, and improving the stability during the actual use of the subsequent uncertainty quantification model. During the use of the trained uncertainty quantification model, that is, during the forward calculation process, the extended embedding matrix can be sliced first and divided into N model input matrices of N bs *D emb , where N bs is the batch size of the batch of samples, and then perform batch matrix multiplication with the three-dimensional parameter tensor of N model *N in *N out to obtain a final prediction vector of dimension N model , that is, different prediction scores of N model member models for the same target user feature information.

[0062] S210. Based on the training uncertainty quantification value and the conversion label, adjust the parameters of the initial uncertainty quantification model until the model training is completed to obtain the uncertainty quantification model.

[0063] Specifically, the transaction data value estimation device can calculate the loss function based on the training uncertainty quantification value and the conversion label, and adjust the parameters of the initial uncertainty quantification model during the backpropagation training process based on the loss function until the initial uncertainty quantification model completes the model training to obtain the uncertainty quantification model. Among them, the loss function during the model training process of the uncertainty quantification model can be the same as the loss function during the model training process of the prediction conversion rate model.

[0064] In the embodiments of the present application, an initial conversion rate prediction model and an initial uncertainty quantification model are created. Sample user feature information and conversion labels corresponding to each sample user feature information are obtained. The sample user feature information is input into the initial conversion rate prediction model, and the training prediction conversion rate output by the initial conversion rate prediction model is obtained. Based on the training prediction conversion rate and the conversion labels, parameter adjustment processing is performed on the initial conversion rate prediction model until the model training is completed, and a conversion rate prediction model is obtained. The sample user feature information is input into the initial uncertainty quantification model, and the training uncertainty quantification value output by the initial uncertainty quantification model is obtained. Based on the training uncertainty quantification value and the conversion labels, parameter adjustment processing is performed on the initial uncertainty quantification model until the model training is completed, and an uncertainty quantification model is obtained. By training the conversion rate prediction model and the uncertainty quantification model simultaneously, the conversion rate prediction model is used to focus on improving the accuracy of predicting the conversion rate, and the uncertainty quantification model helps to identify the reliability of the predicted conversion rate. Moreover, by combining the conversion rate estimation loss function and the regularization term constraint, overfitting can be prevented while ensuring the model prediction accuracy, and the robustness of the model to outliers can be enhanced.

[0065] The model training of the conversion rate prediction model and the uncertainty quantification model uses the same sample data set, and the same batch processing method and the same batch of samples can also be used for model training. However, the functions of the conversion rate prediction model and the uncertainty quantification model are different. Therefore, all features of the user feature information can be retained during the training process of the conversion rate prediction model. The comprehensiveness of the sample data enables the conversion rate prediction model to make full use of all information to improve the accuracy of conversion rate prediction, and the user behavior sequence features can be used to understand the dynamic behavior and preferences of users, which can help the conversion rate prediction model further improve the understanding of user behavior and the accuracy of conversion rate prediction. However, the user behavior sequence features may bring complexity and noise to the model training of the uncertainty quantification model. Therefore, in order to prevent the uncertainty quantification model from being unable to focus on uncertainty assessment during the model training process, the transaction data value estimation device can preprocess the sample data during the model training process of the uncertainty quantification model.

[0066] Please refer to Figure 4 , which provides a schematic flowchart of sample preprocessing for the embodiments of the present application. In one or more embodiments of the present application, step S208 may include the following steps S302 - S304.

[0067] S302, perform a deletion process on the user behavior sequence features in the sample user feature information.

[0068] Specifically, the transaction data value prediction device can perform elimination processing on the user behavior sequence features in the sample user feature information, so that only the user basic portrait, promotion information features, and scenario features are retained in the sample user feature information after the elimination processing. After eliminating the behavior sequence features, the uncertainty quantification model is more robust to data sparsity, and the learning of the user basic portrait, promotion information features, and scenario features is more stable, thereby providing a more accurate uncertainty assessment and reducing the computational complexity during the model training process.

[0069] S304. Input the sample user feature information after the elimination processing into the initial uncertainty quantification model to obtain the training uncertainty quantification value output by the initial uncertainty quantification model.

[0070] Specifically, the transaction data value prediction device can input the sample user feature information after the elimination processing into the initial uncertainty quantification model. The initial uncertainty quantification model can perform confidence scoring on the predicted conversion rate based on the sample user feature information after the elimination processing, and the training uncertainty quantification value output by the initial uncertainty quantification model can be obtained.

[0071] In the embodiment of the present application, the user behavior sequence features in the sample user feature information are eliminated, and the sample user feature information after the elimination processing is input into the initial uncertainty quantification model to obtain the training uncertainty quantification value output by the initial uncertainty quantification model. By eliminating the user behavior features during the training process of the uncertainty quantification model, the uncertainty quantification model is more focused on learning the features related to the uncertainty quantification value, improving the accuracy of the uncertainty assessment, and reducing the computational complexity and improving the training efficiency during the model training process.

[0072] Please refer to Figure 5 , which provides a schematic flowchart of the process for obtaining the target predicted conversion rate in the embodiment of the present application. In one or more embodiments of the present application, step S104 may include the following steps S402 - S404.

[0073] S402. Input the target user feature information into the conversion rate prediction model to obtain the target initial conversion rate, and input the target user feature information into the uncertainty quantification model to obtain the target uncertainty quantification value corresponding to the target predicted conversion rate.

[0074] Specifically, the transaction data value estimation device may input the target user feature information into the conversion rate prediction model. The conversion rate prediction model may predict the conversion rate based on the Forward Computation process, and may obtain the target initial conversion rate output by the conversion rate prediction model. The target initial conversion rate is the conversion rate predicted by the conversion rate prediction model for the target user feature information. The transaction data value estimation device may input the target user feature information into the uncertainty quantification model, and may obtain the target uncertainty quantification value output by the uncertainty quantification model. The target uncertainty quantification value represents the confidence level of the uncertainty quantification model for the target predicted conversion rate.

[0075] S404. Obtain the target predicted conversion rate corresponding to the target object based on the target initial conversion rate and the target uncertainty quantification value.

[0076] Specifically, the transaction data value estimation device may adjust the target initial conversion rate according to the target uncertainty quantification value, so as to obtain the target predicted conversion rate corresponding to the target object.

[0077] In the embodiments of the present application, the target user feature information is input into the conversion rate prediction model to obtain the target initial conversion rate, the target user feature information is input into the uncertainty quantification model to obtain the target uncertainty quantification value corresponding to the target predicted conversion rate, and the target predicted conversion rate corresponding to the target object is obtained based on the target initial conversion rate and the target uncertainty quantification value. Using the uncertainty quantification value to adjust the initial conversion rate to obtain the predicted conversion rate can identify the error situation of the model, thereby solving the problems of data noise and model deviation, and improving the robustness and accuracy of the model.

[0078] Please refer to Figure 6 , which provides a schematic flow diagram for obtaining the uncertainty quantification value in the embodiments of the present application. In one or more embodiments of the present application, the step S402 may include the following steps S502-S506.

[0079] S502. Input the target user feature information into the uncertainty quantification model, and obtain the prediction scores output by each member model in the uncertainty quantification model.

[0080] Specifically, the transaction data value estimation device may input the target user feature information into the uncertainty quantification model. Since the uncertainty quantification model is an integrated structure and there are multiple member models therein, the prediction scores output by each member model for the target user feature information can be obtained. The prediction score represents the confidence level of each member model for the target initial conversion rate.

[0081] S504. Calculate the average score and the standard deviation of each prediction score.

[0082] Specifically, the standard deviation can reflect the degree of fluctuation of the predicted scores, and the average score can reflect the central tendency of the predicted scores. The average score and the standard deviation of each predicted score can be calculated, and the formulas are as follows:

[0083]

[0084] Among them, is the predicted score of the conversion rate of the j-th member model for the i-th target user feature information, and N model is the number of member models, and μ i is the average score of the predicted scores of N model member models for the i-th target user feature information, and σ i is the standard deviation of the predicted scores of N model member models for the i-th target user feature information. It can be understood that since a target user feature information is a pair of a user feature and a promotion information feature, a target user feature information can also be understood as an exposure opportunity of a promotion information in front of a user.

[0085] S506. Calculate the quotient of the standard deviation and the average score to obtain the target uncertainty quantification value corresponding to the target predicted conversion rate.

[0086] Specifically, the quotient of the standard deviation and the average score can be calculated, and then this quotient is determined as the target uncertainty quantification value output by the uncertainty quantification model corresponding to the target predicted conversion rate. The formula is as follows:

[0087]

[0088] Among them, is the target uncertainty quantification value.

[0089] In the embodiments of the present application, the target user feature information is input into the uncertainty quantification model to obtain the predicted scores output by each member model in the uncertainty quantification model, calculate the quotient of the standard deviation and the average score, and obtain the target uncertainty quantification value corresponding to the target predicted conversion rate. The standard deviation can reflect the degree of fluctuation of the predicted scores, and the average score can reflect the central tendency of the predicted scores. The uncertainty quantification value is determined by the quotient of the two, and the deviation is reduced through model integration, enhancing the accuracy and stability of the uncertainty quantification value.

[0090] Please refer to Figure 7 , which provides a schematic flowchart of a conversion rate adjustment for the embodiments of the present application. In one or more embodiments of the present application, the step S404 may include the following steps S602-S604.

[0091] S602, if the target initial conversion rate is less than the preset conversion rate threshold, then confirm the target initial conversion rate as the target prediction conversion rate of the target object.

[0092] Specifically, if the target initial conversion rate is less than the preset conversion rate threshold, there is no need to adjust the target initial conversion rate, and the target initial conversion rate can be directly confirmed as the target prediction conversion rate of the target object. Among them, the preset conversion rate threshold can be the initial setting estimated by the transaction data value. For a lower target initial conversion rate, it can be directly confirmed as the target prediction conversion rate, thus avoiding unnecessary calculations and resource waste.

[0093] S604, if the target initial conversion rate is greater than the preset conversion rate threshold, then perform a suppression process on the target initial conversion rate based on the preset coefficient and the target uncertainty quantification value to obtain the target prediction conversion rate corresponding to the target object.

[0094] Specifically, if the target initial conversion rate is greater than the preset conversion rate threshold, a suppression process can be performed on the target initial conversion rate based on the preset coefficient and the target uncertainty quantification value, so as to obtain the target prediction conversion rate corresponding to the target object. The formula is as follows:

[0095]

[0096] Among them, pCVR i is the target initial conversion rate, is the target uncertainty quantification value, τ is the preset conversion rate threshold, post_pCVR i is the target prediction conversion rate, κ is the coefficient for controlling the conservative scoring intensity, and κ and τ can be set manually or can be found through the Grid Search method to find the optimal value.

[0097] In the embodiments of the present application, if the target initial conversion rate is less than the preset conversion rate threshold, then confirm the target initial conversion rate as the target prediction conversion rate of the target object. If the target initial conversion rate is greater than the preset conversion rate threshold, then perform a suppression process on the target initial conversion rate based on the preset coefficient and the target uncertainty quantification value to obtain the target prediction conversion rate corresponding to the target object. By suppressing the overly high conversion rate, the overly optimistic prediction conversion rate is avoided, and the robustness of the model is improved.

[0098] Please refer to Figure 8 , which provides a flow schematic diagram for calculating the output value of transaction data in the embodiments of the present application. In one or more embodiments of the present application, the step S106 may include the following steps S702 - S704.

[0099] S702, obtain the single - time transaction data output value for the information promotion position based on the target prediction conversion rate and the transaction data distribution strategy.

[0100] Specifically, the transaction data value prediction device can obtain a transaction data distribution policy, which can be pre-set by the target object or the transaction data value prediction device, and represents how to adjust the single transaction data output value for the information promotion position according to the target prediction conversion rate. For example, the higher the target prediction conversion rate, the higher the success rate of the user completing the conversion behavior after the promotion information is put on the market. Therefore, the single transaction data output value can also be increased. On the contrary, the lower the target prediction conversion rate, the lower the success rate of the user completing the conversion behavior after the promotion information is put on the market. Therefore, the single transaction data output value can be decreased. Therefore, the transaction data value prediction device can obtain the single transaction data output value for the information promotion position based on the target prediction conversion rate and the transaction data distribution policy.

[0101] S704, calculate the product of the target prediction conversion rate, the single transaction data output value, and the predicted click-through rate corresponding to the target object to obtain the target transaction data output value of the target object for the information promotion position.

[0102] Specifically, the product of the target prediction conversion rate, the single transaction data output value, and the predicted click-through rate corresponding to the target object can be calculated to obtain the target transaction data output value of the target object for the information promotion position. The formula is as follows:

[0103] b_bid i =pCTR i *post_pCVR i *c_bid i

[0104] Among them, b_bid i is the target transaction data output value, pCTR i is the target prediction conversion rate, post_pCVR i is the predicted click-through rate corresponding to the target object, and c_bid i is the single transaction data output value.

[0105] In the embodiment of the present application, based on the target prediction conversion rate and the transaction data distribution policy, the single transaction data output value for the information promotion position is obtained, and the product of the target prediction conversion rate, the single transaction data output value, and the predicted click-through rate corresponding to the target object is calculated to obtain the target transaction data output value of the target object for the information promotion position. By combining the prediction conversion rate, the predicted click-through rate, and the transaction data distribution policy to determine the final target transaction data output value, the conversion effect of the promotion activity is predicted from multiple dimensions, avoiding the limitations of a single index and further improving the accuracy of the target transaction data output value.

[0106] Next, it will be combined with the attached Figure 9 - attached Figure 10, a detailed introduction to the transaction data value prediction device for the information promotion position provided by the embodiments of the present application is given. It should be noted that the Figure 9 - appendix Figure 10 The transaction data value prediction device in is used to execute the method of the embodiments of the present application Figures 1-8 shown. For the sake of convenience of description, only the parts related to the embodiments of the present application are shown. For the specific technical details not disclosed, please refer to the embodiments Figures 1-8 shown in the present application.

[0107] Please refer to Figure 9 , which shows a schematic structural diagram of the transaction data value prediction device provided by an exemplary embodiment of the present application. The transaction data value prediction device can be implemented as all or part of the device through software, hardware or a combination of both. The device 1 includes a feature information acquisition unit 11, a conversion rate prediction unit 12 and an output value prediction unit 13.

[0108] The feature information acquisition unit 11 is used to acquire the target user feature information corresponding to the target object;

[0109] The conversion rate prediction unit 12 is used to adopt a conversion rate prediction model and an uncertainty quantification model to obtain a target prediction conversion rate based on the target user feature information;

[0110] The output value prediction unit 13 is used to obtain the target transaction data output value of the target object for the information promotion position based on the target prediction conversion rate, the estimated click-through rate corresponding to the target object and the transaction data distribution strategy.

[0111] In this embodiment, the target user feature information corresponding to the target object is acquired, a conversion rate prediction model and an uncertainty quantification model are adopted, a target prediction conversion rate is obtained based on the target user feature information, and the target transaction data output value of the target object for the information promotion position is obtained based on the target prediction conversion rate, the estimated click-through rate corresponding to the target object and the transaction data distribution strategy. By using the conversion rate prediction model and the uncertainty quantification model to predict the target transaction data output value for the information promotion position, the solution efficiency of the information promotion position attribution problem is improved, and the target transaction data output value conforms to the actual situation of the target object.

[0112] Please refer to Figure 10 , which shows a schematic structural diagram of the transaction data value prediction device provided by an exemplary embodiment of the present application. The transaction data value prediction device can be implemented as all or part of the device through software, hardware or a combination of both. The device 1 includes a feature information acquisition unit 11, a model training unit 14, a conversion rate prediction unit 12 and an output value prediction unit 13.

[0113] A feature information acquisition unit 11 for acquiring target user feature information corresponding to a target object;

[0114] A model training unit 14 for creating an initial conversion rate prediction model and an initial uncertainty quantification model, acquiring sample user feature information and conversion labels corresponding to each sample user feature information;

[0115] Inputting the sample user feature information into the initial conversion rate prediction model to obtain a training prediction conversion rate output by the initial conversion rate prediction model;

[0116] Based on the training prediction conversion rate and the conversion label, performing parameter adjustment processing on the initial conversion rate prediction model until model training is completed to obtain a conversion rate prediction model.

[0117] Inputting the sample user feature information into the initial uncertainty quantification model to obtain a training uncertainty quantification value output by the initial uncertainty quantification model;

[0118] Based on the training uncertainty quantification value and the conversion label, performing parameter adjustment processing on the initial uncertainty quantification model until model training is completed to obtain an uncertainty quantification model.

[0119] Optionally, the model training unit 14 is specifically configured to perform a removal process on the user behavior sequence features in the sample user feature information;

[0120] Inputting the sample user feature information after the removal process into the initial uncertainty quantification model to obtain a training uncertainty quantification value output by the initial uncertainty quantification model.

[0121] A conversion rate prediction unit 12 for using the conversion rate prediction model and the uncertainty quantification model to obtain a target prediction conversion rate based on the target user feature information;

[0122] Optionally, the conversion rate prediction unit 12 is specifically configured to input the target user feature information into the conversion rate prediction model to obtain a target initial conversion rate, input the target user feature information into the uncertainty quantification model to obtain a target uncertainty quantification value corresponding to the target prediction conversion rate;

[0123] Based on the target initial conversion rate and the target uncertainty quantification value, obtaining a target prediction conversion rate corresponding to the target object.

[0124] Optionally, the conversion rate prediction unit 12 is specifically configured to input the target user feature information into the uncertainty quantification model to obtain prediction scores output by each member model in the uncertainty quantification model;

[0125] Calculate the average value and standard deviation of each of the predicted scores;

[0126] Calculate the quotient of the standard deviation and the average value to obtain the target uncertainty quantification value corresponding to the target predicted conversion rate.

[0127] Optionally, the conversion rate prediction unit 12 is specifically configured to, if the target initial conversion rate is less than a preset conversion rate threshold, confirm the target initial conversion rate as the target predicted conversion rate of the target object;

[0128] If the target initial conversion rate is greater than the preset conversion rate threshold, perform a suppression process on the target initial conversion rate based on a preset coefficient and the target uncertainty quantification value to obtain the target predicted conversion rate corresponding to the target object.

[0129] The output value prediction unit 13 is configured to obtain the target transaction data output value of the target object for the information promotion position based on the target predicted conversion rate, the estimated click-through rate corresponding to the target object, and the transaction data distribution strategy.

[0130] Optionally, the output value prediction unit 13 is specifically configured to obtain the single-time transaction data output value for the information promotion position based on the target predicted conversion rate and the transaction data distribution strategy;

[0131] Calculate the product of the target predicted conversion rate, the single-time transaction data output value, and the estimated click-through rate corresponding to the target object to obtain the target transaction data output value of the target object for the information promotion position.

[0132] In the embodiments of the present application, the target user feature information corresponding to the target object is obtained, an initial conversion rate prediction model and an initial uncertainty quantification model are created, the sample user feature information and the conversion labels corresponding to each sample user feature information are obtained, the sample user feature information is input into the initial conversion rate prediction model, and the training prediction conversion rate output by the initial conversion rate prediction model is obtained. Based on the training prediction conversion rate and the conversion labels, the parameter adjustment process is performed on the initial conversion rate prediction model until the model training is completed, and the conversion rate prediction model is obtained. The user behavior sequence features in the sample user feature information are removed, and the sample user feature information after the removal process is input into the initial uncertainty quantification model, and the training uncertainty quantification value output by the initial uncertainty quantification model is obtained. By removing the user behavior features during the training process of the uncertainty quantification model, the uncertainty quantification model is more focused on learning the features related to the uncertainty quantification value, improving the accuracy of uncertainty evaluation, and reducing the computational complexity and improving the training efficiency during the model training process. Based on the training uncertainty quantification value and the conversion labels, the parameter adjustment process is performed on the initial uncertainty quantification model until the model training is completed, and the uncertainty quantification model is obtained. By training the conversion rate prediction model and the uncertainty quantification model simultaneously, the conversion rate prediction model is used to focus on improving the accuracy of the predicted conversion rate, and the uncertainty quantification model helps to identify the reliability of the predicted conversion rate. Moreover, by combining the conversion rate estimation loss function and the regularization term constraint, overfitting can be prevented while ensuring the prediction accuracy of the model, and the robustness of the model to outliers can be enhanced.

[0133] Input the target user feature information into the conversion rate prediction model to obtain the target initial conversion rate. Input the target user feature information into the uncertainty quantification model to obtain the prediction scores output by each member model in the uncertainty quantification model. Calculate the quotient of the standard deviation value and the average score to obtain the target uncertainty quantification value corresponding to the target predicted conversion rate. The standard deviation value can reflect the fluctuation degree of the prediction scores, and the average score can reflect the central tendency of the prediction scores. Determine the uncertainty quantification value through the quotient of the two, and reduce the bias through model integration, enhancing the accuracy and stability of the uncertainty quantification value. If the target initial conversion rate is less than the preset conversion rate threshold, then confirm the target initial conversion rate as the target predicted conversion rate of the target object. If the target initial conversion rate is greater than the preset conversion rate threshold, then perform a suppression process on the target initial conversion rate based on the preset coefficient and the target uncertainty quantification value to obtain the target predicted conversion rate corresponding to the target object. By performing a suppression process on the overly high conversion rate, over-optimistic predicted conversion rates are avoided, enhancing the robustness of the model. Use the uncertainty quantification value to adjust the initial conversion rate to obtain the predicted conversion rate, which can identify the error situations of the model, thereby solving the problems of data noise and model bias, and improving the robustness and accuracy of the model. Obtain the target user feature information corresponding to the target object, adopt the conversion rate prediction model and the uncertainty quantification model, obtain the target predicted conversion rate based on the target user feature information, and obtain the target transaction data output value of the target object for the information promotion position based on the target predicted conversion rate, the estimated click-through rate corresponding to the target object, and the transaction data distribution strategy. Predict the target transaction data output value for the information promotion position through the conversion rate prediction model and the uncertainty quantification model, improving the solution efficiency of the information promotion position attribution problem and making the target transaction data output value conform to the actual situation of the target object. Predict the target transaction data output value for the information promotion position through the conversion rate prediction model and the uncertainty quantification model, improving the solution efficiency of the information promotion position attribution problem and making the target transaction data output value conform to the actual situation of the target object.

[0134] It should be noted that when the transaction data value estimation device provided in the above embodiment executes the transaction data value estimation method, only the above division of each functional module is used for illustration. In practical applications, the above functions can be allocated to different functional modules according to needs, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. In addition, the transaction data value estimation device provided in the above embodiment and the embodiment of the transaction data value estimation method belong to the same concept, and the implementation process thereof is detailed in the method embodiment, which will not be elaborated here.

[0135] The serial numbers of the above embodiments of the present application are only for description and do not represent the advantages or disadvantages of the embodiments.

[0136] The embodiments of the present application further provide a computer storage medium, which can store multiple instructions, and the instructions are suitable for being loaded and executed by a processor to perform the transaction data value prediction method as described in the above Figures 1-8 shown embodiment. The specific execution process can be referred to the Figures 1-8 specific description of the shown embodiment, which will not be elaborated here.

[0137] The present application also provides a computer program product, which stores at least one instruction, and the at least one instruction is loaded and executed by the processor to perform the transaction data value prediction method as described in the above Figures 1-8 shown embodiment. The specific execution process can be referred to the Figures 1-8 specific description of the shown embodiment, which will not be elaborated here.

[0138] Please refer to Figure 11 , which shows a block diagram of the structure of an electronic device provided by an exemplary embodiment of the present application. The electronic device in the present application may include one or more of the following components: a processor 110, a memory 120, an input device 130, an output device 140, and a bus 150. The processor 110, the memory 120, the input device 130, and the output device 140 may be connected through the bus 150.

[0139] The processor 110 may include one or more processing cores. The processor 110 uses various interfaces and lines to connect various parts within the entire electronic device, and by running or executing instructions, programs, code sets, or instruction sets stored in the memory 120, and by calling data stored in the memory 120, it executes various functions of the terminal 100 and processes data. Optionally, the processor 110 may be implemented in at least one hardware form of digital signal processing (DSP), field-programmable gate array (FPGA), or programmable logic array (PLA). The processor 110 may integrate one or a combination of a central processing unit (CPU), a graphics processing unit (GPU), and a modem, etc. Among them, the CPU mainly processes the operating system, user pages, and application programs, etc.; the GPU is responsible for rendering and drawing display content; the modem is used to process wireless communication. It can be understood that the above modem may not be integrated into the processor 110 and may be implemented separately through a communication chip.

[0140] The memory 120 may include a Random Access Memory (RAM), and may also include a Read-Only Memory (ROM). Optionally, the memory 120 includes a Non-Transitory Computer-Readable Storage Medium. The memory 120 can be used to store instructions, programs, code, code sets, or instruction sets. The memory 120 may include a program storage area and a data storage area. Among them, the program storage area can store instructions for implementing the operating system, instructions for implementing at least one function (such as touch function, sound playback function, image playback function, etc.), instructions for implementing the above-mentioned various method embodiments, etc. The operating system can be the Android system, including a system developed based on the Android system in depth, the IOS system developed by Apple Inc., including a system developed based on the IOS system in depth, or other systems.

[0141] The memory 120 can be divided into an operating system space and a user space. The operating system runs in the operating system space, and native and third-party application programs run in the user space. To ensure that different third-party application programs can achieve better running effects, the operating system allocates corresponding system resources for different third-party application programs. However, there are also differences in the system resource requirements of different application scenarios in the same third-party application program. For example, in the local resource loading scenario, the third-party application program has a higher requirement for the disk read speed; in the animation rendering scenario, the third-party application program has a higher requirement for the GPU performance. The operating system and the third-party application program are independent of each other, and the operating system often cannot timely perceive the current application scenario of the third-party application program, resulting in the operating system being unable to perform targeted system resource adaptation according to the specific application scenario of the third-party application program.

[0142] In order to enable the operating system to distinguish the specific application scenarios of third-party application programs, it is necessary to establish data communication between the third-party application programs and the operating system, so that the operating system can obtain the current scenario information of the third-party application programs at any time, and then perform targeted system resource adaptation based on the current scenario.

[0143] Among them, the input device 130 is used to receive input instructions or data. The input device 130 includes but is not limited to a keyboard, a mouse, a camera, a microphone, or a touch device. The output device 140 is used to output instructions or data. The output device 140 includes but is not limited to a display device and a speaker, etc. In one example, the input device 130 and the output device 140 can be combined, and the input device 130 and the output device 140 are a touch display screen.

[0144] The touch display screen can be designed as a full-screen, a curved screen or a special-shaped screen. The touch display screen can also be designed as a combination of a full-screen and a curved screen, or a combination of a special-shaped screen and a curved screen. The embodiments of the present application do not limit this.

[0145] In addition, those skilled in the art can understand that the structure of the electronic device shown in the above drawings does not limit the electronic device. The electronic device may include more or fewer components than shown in the drawings, or combine some components, or have different component arrangements. For example, the electronic device also includes components such as a radio frequency circuit, an input unit, a sensor, an audio circuit, a Wireless Fidelity (WiFi) module, a power supply, and a Bluetooth module, which will not be elaborated here.

[0146] In Figure 11 In the electronic device shown, the processor 110 can be used to call the transaction data value prediction application for the information promotion position stored in the memory 120, and specifically perform the following operations:

[0147] Obtain the target user feature information corresponding to the target object;

[0148] Adopt a conversion rate prediction model and an uncertainty quantification model to obtain a target prediction conversion rate based on the target user feature information;

[0149] Based on the target prediction conversion rate, the estimated click-through rate corresponding to the target object, and the transaction data distribution strategy, obtain the target transaction data output value of the target object for the information promotion position.

[0150] In one embodiment, before the processor 110 executes the operation of adopting a conversion rate prediction model and an uncertainty quantification model to obtain a target prediction conversion rate based on the target user feature information, the following operations are also performed:

[0151] Create an initial conversion rate prediction model and an initial uncertainty quantification model, obtain sample user feature information and conversion labels corresponding to each sample user feature information;

[0152] Input the sample user feature information into the initial conversion rate prediction model to obtain the training prediction conversion rate output by the initial conversion rate prediction model;

[0153] Based on the training prediction conversion rate and the conversion label, perform parameter adjustment processing on the initial conversion rate prediction model until the model training is completed to obtain a conversion rate prediction model.

[0154] Input the sample user feature information into the initial uncertainty quantification model to obtain the training uncertainty quantification value output by the initial uncertainty quantification model;

[0155] Based on the training uncertainty quantification value and the conversion label, perform parameter adjustment processing on the initial uncertainty quantification model until the model training is completed to obtain an uncertainty quantification model.

[0156] In one embodiment, when the processor 110 executes inputting the sample user feature information into the initial uncertainty quantification model to obtain the training uncertainty quantification value output by the initial uncertainty quantification model, the following operations are specifically performed:

[0157] Perform elimination processing on the user behavior sequence feature in the sample user feature information;

[0158] Input the sample user feature information after the elimination processing into the initial uncertainty quantification model to obtain the training uncertainty quantification value output by the initial uncertainty quantification model.

[0159] In one embodiment, when the processor 110 executes using the conversion rate prediction model and the uncertainty quantification model to obtain the target prediction conversion rate based on the target user feature information, the following operations are specifically performed:

[0160] Input the target user feature information into the conversion rate prediction model to obtain the target initial conversion rate, and input the target user feature information into the uncertainty quantification model to obtain the target uncertainty quantification value corresponding to the target prediction conversion rate;

[0161] Based on the target initial conversion rate and the target uncertainty quantification value, obtain the target prediction conversion rate corresponding to the target object.

[0162] In one embodiment, when the processor 110 executes inputting the target user feature information into the uncertainty quantification model to obtain the target uncertainty quantification value corresponding to the target prediction conversion rate, the following operations are specifically performed:

[0163] Input the target user feature information into the uncertainty quantification model to obtain the prediction scores output by each member model in the uncertainty quantification model;

[0164] Calculate the average score and the standard deviation of each of the prediction scores;

[0165] Calculate the quotient of the standard deviation and the average score to obtain the target uncertainty quantification value corresponding to the target prediction conversion rate.

[0166] In one embodiment, when the processor 110 executes based on the target initial conversion rate and the target uncertainty quantification value to obtain the target prediction conversion rate corresponding to the target object, the following operations are specifically performed:

[0167] If the target initial conversion rate is less than the preset conversion rate threshold, the target initial conversion rate is confirmed as the target predicted conversion rate of the target object;

[0168] If the target initial conversion rate is greater than the preset conversion rate threshold, the target initial conversion rate is suppressed based on a preset coefficient and the target uncertainty quantification value to obtain the target predicted conversion rate corresponding to the target object.

[0169] In one embodiment, when the processor 110 executes to obtain the target transaction data output value of the target object for the information promotion position based on the target predicted conversion rate, the estimated click-through rate corresponding to the target object, and the transaction data distribution strategy, the following operations are specifically executed:

[0170] Based on the target predicted conversion rate and the transaction data distribution strategy, obtain the single-transaction data output value for the information promotion position;

[0171] Calculate the product of the target predicted conversion rate, the single-transaction data output value, and the estimated click-through rate corresponding to the target object to obtain the target transaction data output value of the target object for the information promotion position.

[0172] In the embodiments of the present application, the target user feature information corresponding to the target object is obtained, an initial conversion rate prediction model and an initial uncertainty quantification model are created, the sample user feature information and the conversion labels corresponding to each sample user feature information are obtained, the sample user feature information is input into the initial conversion rate prediction model, the training predicted conversion rate output by the initial conversion rate prediction model is obtained, and based on the training predicted conversion rate and the conversion label, the parameters of the initial conversion rate prediction model are adjusted until the model training is completed to obtain the conversion rate prediction model. The user behavior sequence features in the sample user feature information are removed, and the sample user feature information after the removal process is input into the initial uncertainty quantification model to obtain the training uncertainty quantification value output by the initial uncertainty quantification model. By removing user behavior features during the training process of the uncertainty quantification model, the uncertainty quantification model focuses more on learning features related to the uncertainty quantification value, improving the accuracy of uncertainty assessment, reducing the computational complexity during the model training process, and improving the training efficiency. Based on the training uncertainty quantification value and the conversion label, the parameters of the initial uncertainty quantification model are adjusted until the model training is completed to obtain the uncertainty quantification model. By training the conversion rate prediction model and the uncertainty quantification model simultaneously, the conversion rate prediction model focuses on improving the accuracy of the predicted conversion rate, while the uncertainty quantification model helps to identify the reliability of the predicted conversion rate. Moreover, by combining the conversion rate estimation loss function and the regularization term constraint, overfitting can be prevented while ensuring the model prediction accuracy, and the robustness of the model to outliers can be enhanced.

[0173] Input the target user characteristic information into the conversion rate prediction model to obtain the target initial conversion rate. Input the target user characteristic information into the uncertainty quantification model to obtain the prediction scores output by each member model in the uncertainty quantification model. Calculate the quotient of the standard deviation value and the average score to obtain the target uncertainty quantification value corresponding to the target predicted conversion rate. The standard deviation value can reflect the fluctuation degree of the prediction scores, and the average score can reflect the central tendency of the prediction scores. Determine the uncertainty quantification value through the quotient of the two, and reduce the deviation through model integration, enhancing the accuracy and stability of the uncertainty quantification value. If the target initial conversion rate is less than the preset conversion rate threshold, then confirm the target initial conversion rate as the target predicted conversion rate of the target object. If the target initial conversion rate is greater than the preset conversion rate threshold, then perform a suppression process on the target initial conversion rate based on the preset coefficient and the target uncertainty quantification value to obtain the target predicted conversion rate corresponding to the target object. By performing a suppression process on the overly high conversion rate, over-optimistic predicted conversion rates are avoided, enhancing the robustness of the model. Using the uncertainty quantification value to adjust the initial conversion rate to obtain the predicted conversion rate can identify the error situations of the model, thereby solving the problems of data noise and model deviation, and improving the robustness and accuracy of the model. Obtain the target user characteristic information corresponding to the target object, adopt the conversion rate prediction model and the uncertainty quantification model, obtain the target predicted conversion rate based on the target user characteristic information, and obtain the target transaction data output value of the target object for the information promotion position based on the target predicted conversion rate, the estimated click-through rate corresponding to the target object, and the transaction data distribution strategy. Predicting the target transaction data output value for the information promotion position through the conversion rate prediction model and the uncertainty quantification model improves the solution efficiency of the information promotion position attribution problem and makes the target transaction data output value conform to the actual situation of the target object. Predicting the target transaction data output value for the information promotion position through the conversion rate prediction model and the uncertainty quantification model improves the solution efficiency of the information promotion position attribution problem and makes the target transaction data output value conform to the actual situation of the target object.

[0174] Those of ordinary skill in the art can understand that all or part of the processes in implementing the above method embodiments can be completed by instructing relevant hardware through a computer program. The program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the above method embodiments. Among them, the storage medium can be a magnetic disk, an optical disc, a read-only memory, a random access memory, etc.

[0175] The above-disclosed content is only the preferred embodiment of the present application. Of course, it cannot be used to limit the scope of rights of the present application. Therefore, equivalent changes made according to the claims of the present application still fall within the scope covered by the present application.

[0176] It should be noted that the information (including but not limited to user device information, user personal information, etc.), data (including but not limited to data for analysis, stored data, displayed data, etc.) and signals involved in the embodiments of this specification are all authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data need to comply with the relevant laws, regulations and standards of relevant countries and regions. For example, the target user characteristic information involved in this specification is obtained under full authorization.

Claims

1. A method for estimating transaction data value of information promotion position, characterized in that: The method comprises: Obtain target user feature information corresponding to the target object; Using a conversion rate prediction model and an uncertainty quantification model, obtaining a target predicted conversion rate based on the target user feature information; Based on the target predicted conversion rate, the estimated click rate corresponding to the target object and the transaction data release strategy, the target transaction data output value of the target object for the information promotion position is obtained.

2. The method according to claim 1, characterized in that Before adopting the conversion rate prediction model and the uncertainty quantification model to obtain the target predicted conversion rate based on the target user feature information, the method further includes: Create an initial conversion rate prediction model and an initial uncertainty quantification model, obtain sample user feature information and conversion labels corresponding to each sample user feature information; Inputting the sample user feature information into the initial conversion rate prediction model to obtain the training predicted conversion rate output by the initial conversion rate prediction model; Based on the training predicted conversion rate and the conversion label, the initial conversion rate prediction model is subjected to parameter adjustment processing until the model training is completed to obtain a conversion rate prediction model. Inputting the sample user feature information into the initial uncertainty quantization model to obtain a training uncertainty quantization value output by the initial uncertainty quantization model; Based on the training uncertainty quantization value and the conversion label, the initial uncertainty quantization model is subjected to parameter adjustment processing until the model training is completed to obtain an uncertainty quantification model.

3. The method according to claim 2, characterized in that The step of inputting the sample user feature information into the initial uncertainty quantization model to obtain a training uncertainty quantization value output by the initial uncertainty quantization model includes: Eliminating user behavior sequence features in the sample user feature information; The sample user feature information after elimination is input into the initial uncertainty quantization model to obtain the training uncertainty quantization value output by the initial uncertainty quantization model.

4. The method according to claim 1, characterized in that: The method of adopting the conversion rate prediction model and the uncertainty quantification model to obtain the target predicted conversion rate based on the target user feature information includes: Inputting the target user feature information into a conversion rate prediction model to obtain a target initial conversion rate, and inputting the target user feature information into an uncertainty quantification model to obtain a target uncertainty quantification value corresponding to the target predicted conversion rate; Based on the target initial conversion rate and the target uncertainty quantification value, a target predicted conversion rate corresponding to the target object is obtained.

5. The method according to claim 4, characterized in that The step of inputting the target user feature information into an uncertainty quantification model to obtain a target uncertainty quantification value corresponding to the target predicted conversion rate includes: Inputting the target user feature information into an uncertainty quantification model to obtain a prediction score output by each member model in the uncertainty quantification model; Calculate the average score and standard deviation of each of the predicted scores; The quotient of the standard deviation value and the average score is calculated to obtain a target uncertainty quantification value corresponding to the target predicted conversion rate.

6. The method according to claim 4, characterized in that The step of obtaining the target predicted conversion rate corresponding to the target object based on the target initial conversion rate and the target uncertainty quantization value includes: If the target initial conversion rate is less than a preset conversion rate threshold, the target initial conversion rate is confirmed as the target predicted conversion rate of the target object; If the target initial conversion rate is greater than the preset conversion rate threshold, the target initial conversion rate is suppressed based on a preset coefficient and the target uncertainty quantization value to obtain the target predicted conversion rate corresponding to the target object.

7. The method according to claim 1, characterized in that The step of obtaining the target transaction data output value of the target object for the information promotion position based on the target predicted conversion rate, the estimated click rate corresponding to the target object, and the transaction data release strategy includes: Based on the target predicted conversion rate and the transaction data issuance strategy, a single transaction data output value for the information promotion position is obtained; The product of the target predicted conversion rate, the single transaction data output value and the estimated click rate corresponding to the target object is calculated to obtain the target transaction data output value of the target object for the information promotion position.

8. A transaction data value estimation device for information promotion bits, characterized in that: The device comprises: A feature information acquisition unit, used to acquire target user feature information corresponding to the target object; A conversion rate prediction unit, configured to obtain a target predicted conversion rate based on the target user feature information by using a conversion rate prediction model and an uncertainty quantification model; The output value prediction unit is used to obtain the target transaction data output value of the target object for the information promotion position based on the target predicted conversion rate, the estimated click rate corresponding to the target object and the transaction data issuance strategy.

9. A computer storage medium, characterized in that The computer storage medium stores a plurality of instructions, and the instructions are suitable for being loaded by a processor and executing the method steps according to any one of claims 1 to 7.

10. A computer program product, wherein the computer program product stores a plurality of instructions, wherein the instructions are suitable for being loaded by a processor and executing the method steps according to any one of claims 1 to 7.

11. An electronic device, characterized in that: include: A processor and a memory; wherein the memory stores a computer program, and the computer program is suitable for being loaded by the processor and executing the method steps as claimed in any one of claims 1 to 7.

Images