Data prediction method, electronic device, storage medium and computer program product

Abstract

The embodiment of the invention provides a data prediction method, electronic equipment, a storage medium and a computer program product. The method comprises the following steps: acquiring portrait data and life cycle data of a target user account, and generating portrait features corresponding to the portrait data; the portrait feature comprises a payment feature and a non-payment feature, and the life cycle data is used for representing the payment state of the target user account; inputting the portrait features and the life cycle data into an LTV prediction model to obtain an LTV prediction result corresponding to the target user account; wherein the LTV prediction model comprises a gating network used for determining a first weight corresponding to a payment feature and a second weight corresponding to a non-payment feature according to a payment state represented by life cycle data; and the payment amount prediction network is used for carrying out weighting processing on the payment features according to the first weight, carrying out weighting processing on the non-payment features according to the second weight, and generating the LTV prediction result based on the payment features and the non-payment features obtained through weighting processing.

Description

Technical Field

[0001] The embodiments of the present disclosure relate to the field of artificial intelligence technology. More specifically, the embodiments of the present disclosure relate to a data prediction method, an electronic device, a storage medium, and a computer program product. Background Art

[0002] This section is intended to provide a background or context to the embodiments of the disclosure that are recited in the claims. No description herein is admitted to be prior art by inclusion in this section.

[0003] Life Time Value (LTV) is an indicator used to measure the total value of a user's contribution to a platform or enterprise throughout the entire product cycle (or a period of time). Predicting a user's LTV helps an enterprise or platform more accurately assess the value of a single user, thereby adopting different marketing strategies for users of different values ​​to maximize the long-term value of users and improve user retention.

[0004] In the related art, LTV prediction is usually performed based on user profile data and historical payment records. However, this method fails to fully consider the differences in payment behaviors between different users, resulting in inaccurate LTV prediction results. Summary of the invention

[0005] In a first aspect of the embodiments of the present disclosure, a data prediction method is provided, comprising:

[0006] Acquire the portrait data and life cycle data of the target user account, and generate the portrait features corresponding to the portrait data; the portrait features include payment features and non-payment features, and the life cycle data is used to characterize the payment status of the target user account;

[0007] The portrait features and the life cycle data are input into the LTV prediction model to obtain the LTV prediction result corresponding to the target user account; wherein the LTV prediction model includes:

[0008] A gating network, configured to determine a first weight corresponding to the payment feature and a second weight corresponding to the non-payment feature according to the payment status represented by the lifecycle data;

[0009] The payment amount prediction network is used to weight the payment feature according to the first weight, and to weight the non-payment feature according to the second weight, and to generate the LTV prediction result based on the payment feature and non-payment feature obtained by weighted processing.

[0010] Optionally, the portrait data includes static attribute data, historical active behavior data and payment data of the target user account, and generating portrait features corresponding to the portrait data includes: extracting features from the payment data to obtain the payment features; extracting features from the static attribute data and the historical active behavior data to obtain the non-payment features, which include static attribute features corresponding to the static attribute data and historical active behavior features corresponding to the historical active behavior data; normalizing the payment features and the non-payment features, respectively, and concatenating the processed payment features with the processed non-payment features to obtain the portrait features.

[0011] Optionally, the LTV prediction model also includes at least one payment behavior prediction subnetwork, each payment behavior prediction subnetwork corresponding to a preset time window; wherein any payment behavior prediction subnetwork is used to: generate a payment behavior prediction result of the target user account within the corresponding preset time window based on the portrait features and the first weight and the second weight output by the gating network; the payment amount prediction network is also used to: generate the LTV prediction result based on the payment features and non-payment features obtained by weighted processing, and the payment behavior prediction results generated by each payment behavior prediction subnetwork.

[0012] Optionally, the LTV prediction model also includes at least one payment amount prediction subnetwork, each payment amount prediction subnetwork corresponds to a preset time window, and all preset time windows are different from the time window corresponding to the payment amount prediction network; wherein, any payment amount prediction subnetwork is used to: generate the LTV prediction result of the target user account within the corresponding preset time window based on the payment features and non-payment features obtained by weighted processing.

[0013] Optionally, the payment amount prediction network includes a prediction layer and at least one feature processing layer, each feature processing layer corresponds to a different feature dimension and each feature processing layer corresponds to a gating network; wherein any feature processing layer is used to: perform feature analysis based on the output of the previous feature processing layer and the first weight and the second weight output by the corresponding gating network, and output the features of the corresponding feature dimension; the input of the first feature processing layer is the payment feature, the non-payment feature and the first weight and the second weight output by the corresponding gating network; the prediction layer is used to: perform prediction based on the output of the last feature processing layer to generate the LTV prediction result.

[0014] In a second aspect of the embodiments of the present disclosure, a method for training a data prediction model is provided, comprising:

[0015] Obtain sample portrait data, sample life cycle data and sample LTV of the sample user account, and generate sample portrait features corresponding to the sample portrait data; wherein the sample portrait features include sample payment features and sample non-payment features, and the sample life cycle data is used to characterize the sample payment status of the sample user account;

[0016] Inputting the sample portrait features and the sample life cycle data into the LTV prediction model to be trained to obtain the sample LTV prediction result corresponding to the sample user account;

[0017] Calculating an LTV loss value based on the sample LTV and the sample LTV prediction result, and iteratively training the LTV prediction model to be trained according to the LTV loss value;

[0018] Wherein, the LTV prediction model to be trained includes:

[0019] A gating network, used to determine a first sample weight corresponding to the sample payment feature and a second sample weight corresponding to the sample non-payment feature according to the sample payment status represented by the sample life cycle data;

[0020] The payment amount prediction network is used to weight the sample payment features according to the first sample weight, and to weight the sample non-payment features according to the second sample weight, and to generate the sample LTV prediction result based on the sample payment features and sample non-payment features obtained by weighted processing.

[0021] Optionally, the LTV prediction model to be trained also includes at least one payment behavior prediction subnetwork, each payment behavior prediction subnetwork corresponding to a sample preset time window; wherein any payment behavior prediction subnetwork is used to: generate a sample payment behavior prediction result of the sample user account within the corresponding sample preset time window based on the sample portrait features and the first sample weight and the second sample weight output by the gating network; the payment amount prediction network is also used to: generate the sample LTV prediction result based on the sample payment features and sample non-payment features obtained by weighted processing, and the sample payment behavior prediction results generated by each payment behavior prediction subnetwork.

[0022] Optionally, it also includes: obtaining the sample payment behavior within each sample preset time window respectively, and calculating the payment behavior loss value corresponding to each sample preset time window according to the sample payment behavior corresponding to each sample preset time window and the sample payment behavior prediction result; the iterative training of the LTV prediction model to be trained according to the LTV loss value includes: iterative training of the LTV prediction model to be trained according to the LTV loss value and the payment behavior loss value corresponding to each sample preset time window.

[0023] Optionally, the LTV prediction model to be trained also includes at least one payment amount prediction subnetwork, each payment amount prediction subnetwork corresponds to a sample preset time window, and all sample preset time windows are different from the sample time windows corresponding to the payment amount prediction network; wherein, any payment amount prediction subnetwork is used to: generate a sample LTV prediction result of the sample user account within the corresponding sample preset time window based on the sample payment features and sample non-payment features obtained by weighted processing.

[0024] Optionally, it also includes: obtaining the sample LTV within each sample preset time window respectively, and calculating the LTV loss value corresponding to each sample preset time window according to the sample LTV corresponding to each sample preset time window and the sample LTV prediction result; the iterative training of the LTV prediction model to be trained according to the LTV loss value includes: iterative training of the LTV prediction model to be trained according to the LTV loss value and the LTV loss value corresponding to each sample preset time window.

[0025] In a third aspect of the embodiments of the present disclosure, there is provided an electronic device, including:

[0026] processor;

[0027] a memory for storing instructions executable by the processor;

[0028] The processor implements the method as described in any one of the first aspect or the second aspect by running the executable instructions.

[0029] In a fourth aspect of the embodiments of the present disclosure, a computer-readable storage medium is provided, on which computer instructions are stored. When the instructions are executed by a processor, the method described in any one of the first aspect or the second aspect is implemented.

[0030] In a fifth aspect of the embodiments of the present disclosure, a computer program product is provided, comprising a computer program / instruction, which, when executed by a processor, implements the steps of the method described in any one of the first aspect or the second aspect.

[0031] The above embodiments of the present disclosure have at least the following beneficial effects:

[0032] The target user account's portrait data and life cycle data are input into the LTV prediction model to generate the LTV prediction result corresponding to the target user account using the LTV prediction model. The LTV prediction model includes a gating network and a payment amount prediction network. During the prediction process, the gating network can determine the weights of the payment features and non-payment features contained in the portrait features corresponding to the portrait data based on the payment status of the target user account. Furthermore, the payment amount prediction network can generate the LTV prediction result based on the weighted payment features and weighted non-payment features.

[0033] Since the payment characteristics of different users reflect the differences in payment behaviors between different users, the weights of the user's payment characteristics and non-payment characteristics are controlled through a gated network, and the influence of the payment characteristics and non-payment characteristics on the LTV prediction results is controlled based on the corresponding weights, so that the generated LTV prediction results can be more in line with the user's payment characteristics. This prediction method fully considers the impact of the differences in payment behaviors of different users on the LTV prediction results, and can significantly improve the accuracy of the LTV prediction results. It can be seen that through the technical solution disclosed in the present invention, accurate LTV prediction results can be generated for users with different payment behaviors. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] The above and other objects, features and advantages of the exemplary embodiments of the present disclosure will become readily understood by reading the detailed description below with reference to the accompanying drawings. In the accompanying drawings, several embodiments of the present disclosure are shown in an exemplary and non-limiting manner, in which:

[0035] Figure 1 is a schematic diagram of a system architecture of a data prediction method provided by an exemplary embodiment;

[0036] Figure 2 is a flow chart of a data prediction method provided by an exemplary embodiment;

[0037] Figure 3 is a structural schematic diagram of an LTV prediction model provided by an exemplary embodiment;

[0038] Figure 4 is a structural schematic diagram of another LTV prediction model provided by an exemplary embodiment;

[0039] Figure 5 is a schematic diagram of a structure of a gating network provided by an exemplary embodiment;

[0040] Figure 6a It is a structural schematic diagram of an LTV prediction model including a payment behavior prediction subnetwork provided by an exemplary embodiment;

[0041] Figure 6bIt is a schematic diagram of the structure of another LTV prediction model including a payment behavior prediction subnetwork provided by an exemplary embodiment;

[0042] Figure 6c is a structural diagram of a payment behavior prediction subnetwork provided by an exemplary embodiment;

[0043] Figure 7a It is a structural schematic diagram of an LTV prediction model including a payment amount prediction subnetwork provided by an exemplary embodiment;

[0044] Figure 7b It is a schematic diagram of the structure of another LTV prediction model including a payment amount prediction subnetwork provided by an exemplary embodiment;

[0045] Figure 8a It is a schematic diagram of the structure of an LTV prediction model including both a payment amount prediction subnetwork and a payment behavior prediction subnetwork provided by an exemplary embodiment;

[0046] Figure 8b is a structural schematic diagram of another LTV prediction model provided by an exemplary embodiment;

[0047] Fig. 9 is a flowchart of a method for training a data prediction model provided by an exemplary embodiment;

[0048] Fig.10 is a block diagram of a data prediction device provided by an exemplary embodiment;

[0049] Fig.11 is a block diagram of a data prediction model training device provided by an exemplary embodiment;

[0050] Fig.12 is a schematic diagram of a readable storage medium corresponding to the above method provided by an exemplary embodiment;

[0051] Fig.13 It is a schematic diagram of an electronic device capable of implementing the above method provided by an exemplary embodiment.

[0052] In the drawings, the same or corresponding reference numerals represent the same or corresponding parts. DETAILED DESCRIPTION

[0053] The principles and spirit of the present disclosure will be described below with reference to several exemplary embodiments. It should be understood that these embodiments are provided only to enable those skilled in the art to better understand and implement the present disclosure, and are not intended to limit the scope of the present disclosure in any way. On the contrary, these embodiments are provided to make the present disclosure more thorough and complete, and to fully convey the scope of the present disclosure to those skilled in the art.

[0054] Those skilled in the art know that the embodiments of the present disclosure can be implemented as a system, device, equipment, method or computer-readable storage medium. Therefore, the present disclosure can be specifically implemented in the following forms, namely: complete hardware, complete software (including firmware, resident software, microcode, etc.), or a combination of hardware and software.

[0055] According to an embodiment of the present disclosure, a data prediction method, an electronic device, a storage medium, and a computer program product are proposed.

[0056] In this document, it should be understood that any number of elements in the drawings is for illustration and not limitation, and any naming is only for distinction and does not have any limiting meaning. In addition, the data involved in this disclosure may be data authorized by the user or fully authorized by all parties.

[0057] The principle and spirit of the present disclosure are explained in detail below with reference to several representative embodiments of the present disclosure.

[0058] Application Scenario Overview

[0059] In the Internet field, LTV is usually used to measure the economic benefits that an application's users bring to the operator of the application. For example, for a music application, the user's LTV can be understood as the total amount of money that the user recharges for the music application over a period of time, which can include the amount of money the user spends on membership purchases, the amount of money the user spends on songs, and the amount paid for other recharge activities. Recharge activities refer to recharge activities that are participated in within the music application. Predicting a user's LTV helps improve the user's retention rate or conversion rate in the application.

[0060] In the related art, LTV prediction is usually performed based on user profile data and historical payment records. However, this method fails to fully consider the differences in payment behavior between different users, resulting in inaccurate LTV prediction results. For example, for users who have never paid, they do not have historical payment records, so the lack of historical payment records during the prediction process will result in inaccurate LTV prediction results.

[0061] It can be seen from this that in the scenario of predicting user LTV, there may be a problem of not being able to generate accurate LTV prediction results for some users.

[0062] It should be noted that the above application scenarios are only shown to facilitate understanding of the spirit and principle of the present disclosure, and the embodiments of the present disclosure are not limited in this respect. On the contrary, the embodiments of the present disclosure can be applied to any applicable scenario. SUMMARY OF THE INVENTION

[0064] In view of this, this specification provides a technical solution for data prediction.

[0065] The core concept of this specification is to introduce a gating network into the LTV prediction model, so as to control the weight of the user's payment features and non-payment features on the LTV prediction results through the gating network, and then generate LTV prediction results based on the weighted payment features and weighted non-payment features. In this way, since the payment features of different users reflect the differences in payment behaviors / records between different users, the influence of payment features and non-payment features on the LTV prediction results can be controlled by corresponding weights, so that the generated LTV prediction results can be more in line with the user's payment characteristics, significantly improve the accuracy of the LTV prediction results, and achieve accurate LTV prediction results for users with different payment behaviors.

[0066] Exemplary Methods

[0067] The following will (combined with Figure 1 The system architecture diagram shown in the figure describes the technical concept of this specification in detail through specific embodiments.

[0068]

System Architecture

[0069] In the embodiments of the present disclosure, the terminal may be an electronic device used by a user. The terminal may specifically include but is not limited to a smart phone, a desktop computer, a tablet computer, a laptop computer, an e-book reader, a smart watch, a smart bracelet, and other electronic devices with certain computing capabilities, and the electronic device may run software or websites of instant messaging, social networking, gaming, education, and other categories, and may be applied to scenarios of multi-person audio and video calls.

[0070] The server may include a server, a server cluster formed by multiple servers, or a cloud server. The server may be an independent physical server, a server cluster or a distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN (Content Delivery Network), and big data and artificial intelligence platforms.

[0071] The terminal and the server may be connected directly or indirectly via wired communication or wireless communication, which is not particularly limited in the present disclosure.

[0072] based on Figure 1As shown in the system architecture diagram, the data prediction system includes a server and a terminal, and the LTV prediction model is deployed on the server. The terminal can obtain the portrait data and life cycle data of the target user account, and upload the obtained data to the server. The server can generate corresponding portrait features for the received portrait data, and input the portrait features and the received life cycle data into the LTV prediction model. Then, the server returns the LTV prediction result output by the LTV prediction model to the terminal that submitted the portrait data and life cycle data.

[0073] In one embodiment of the present disclosure, the data prediction system may also include a terminal, and the LTV prediction model is deployed on the terminal. The terminal may obtain the portrait data and life cycle data of the target user account, and generate corresponding portrait features for the acquired portrait data. Then, the terminal inputs the portrait features and life cycle data into the locally deployed LTV prediction model to obtain the LTV prediction result output by the LTV prediction model.

[0074] The following will take a terminal executing the data prediction method provided by the embodiment of the present disclosure as an example for exemplary description. Figure 2 , Figure 2 This is a flow chart of a data prediction method provided by an exemplary embodiment. The method may include the following steps:

[0075] Step 201: Obtain the portrait data and life cycle data of the target user account, and generate portrait features corresponding to the portrait data; the portrait features include payment features and non-payment features, and the life cycle data is used to characterize the payment status of the target user account.

[0076] Typically, users conduct various activities on a system or platform through their corresponding accounts. A user account is an identity created by a user to access an online service, application, or website, which usually includes login credentials such as a username and password. The user's activity records on the platform or application (such as purchase history, browsing behavior, comments, etc.) are associated with the account created on the corresponding platform or application. Where permitted, a user can have multiple accounts on the same platform / application.

[0077] In this embodiment, the target user's account is referred to as a "target user account", and the target user account can be any account for which LTV prediction is required. The portrait data and life cycle data of the target user account refer to the portrait data and life cycle data associated with the target user account. Among them, the portrait data refers to a data set of various attributes and behavioral characteristics of the target user, which is intended to depict a specific user image, and may include but is not limited to age, gender, geographic location, hobbies, consumption habits, active duration, and payment records. The life cycle of a user on the platform / application can be understood as the entire process of subscribing to the service on the platform / application: from the first registration of the user, through the occurrence of important events (such as completing a purchase, upgrading membership, etc.), until the user cancels the subscription or the service is terminated. In this embodiment, the life cycle data of the target user account may include the payment status of the target user account. Exemplarily, the payment status of the user account can be divided into new members (i.e., users who have never paid), valid members (i.e., currently paid users), and member churn (i.e., from paid users to unpaid users). Alternatively, the payment status of the user account can be divided into non-paying users, active members (i.e., non-paying users turned into paying users), high-value members (i.e., users who pay stably for a long time), and lost members (i.e., users who turn from paid users into non-paying users). It should be noted that the above division method is only an example, and those skilled in the art can divide the payment status of the user account according to their own needs, and this disclosure does not limit this.

[0078] If you want to predict the LTV of the target user on any platform / system, you can obtain the portrait data and life cycle data associated with the target user's account on the platform / system. Then, the terminal can perform feature extraction on the portrait data to generate corresponding portrait features. There are many ways to extract features, such as feature extraction through machine learning models, feature extraction using feature extraction algorithms such as Word2Vec and GloVe. The present disclosure does not specifically limit the specific implementation method of feature extraction. The generated portrait features include paid features and non-paid features, among which paid features are generated based on data related to payment behavior in the portrait data, and non-paid features are generated based on data unrelated to payment behavior in the portrait data.

[0079] In some possible implementations, the profile data of the target user account may include static attribute data, historical active behavior data, and payment data. Among them, payment data refers to data related to payment behavior, including but not limited to: historical purchase times, historical purchase amounts, historical purchase unit prices, remaining days of payment status, and number of days of payment status expiration. Feature extraction of payment data can obtain payment features.

[0080] Static attribute data and historical active behavior data are data that are not related to paid behavior. Static attribute data may include data that will not change over a long period of time, such as age, gender, city level, account level, and price range of the device used by the user. Historical active behavior data refers to data generated by the user's other behaviors other than paid behaviors in the corresponding platform / application over a period of time in the past, including but not limited to the number of login days, number of playbacks, and the content and number of favorites / comments / likes. By extracting features from static attribute data and historical active behavior data respectively, we can obtain static attribute features and historical active behavior features. These two parts of features constitute non-paid features.

[0081] Furthermore, the paid features and the non-paid features can be normalized, and the normalized paid features and the non-paid features can be spliced ​​to obtain the portrait features of the target user account. In this way, the scales of the paid features and the non-paid features can be unified.

[0082] In this embodiment, by obtaining the static attribute data and historical activity data of the target user account, the comprehensiveness of the portrait data is improved, which helps to predict more accurate LTV prediction results based on more comprehensive portrait data. In addition, the paid features and non-paid features are normalized and then spliced ​​to obtain the portrait features, and then the portrait features are input into the LTV prediction model, which can avoid the LTV prediction model ignoring the features with smaller scales during the prediction process due to the inconsistent scales of the paid features and non-paid features, thereby further improving the accuracy of the LTV prediction results.

[0083] Step 202: Input the portrait features and the life cycle data into the LTV prediction model to obtain the LTV prediction result corresponding to the target user account; wherein the LTV prediction model includes: a gating network, which is used to determine the first weight corresponding to the payment feature and the second weight corresponding to the non-payment feature according to the payment status represented by the life cycle data; a payment amount prediction network, which is used to weight the payment feature according to the first weight, and weight the non-payment feature according to the second weight, and generate the LTV prediction result based on the payment feature and non-payment feature obtained by weighted processing.

[0084] Figure 3 is a schematic diagram of the structure of an LTV prediction model provided by an exemplary embodiment. Figure 3, the LTV prediction model includes an embedding layer, a gate network, and a payment amount prediction network. The life cycle data and portrait features are input into the LTV prediction model, and the vector layer can perform vector conversion on the life cycle data and convert the life cycle data into life cycle features. Exemplarily, the vector layer can contain a randomly initialized matrix V, the size of which is s*d, where s is the number of payment states. Taking the payment states including new members, valid members, and member churn as an example, s is equal to 3. These three payment states can be numbered as payment state 1, payment state 2, and payment state 3 respectively. The life cycle data of the target user account can include the number of the payment state represented (represented by stage_id). In this way, the number of the payment state is input into the vector layer, and the corresponding (stage_id-1) row content can be found in the matrix V as the life cycle feature output. For example, when stage_id=2, the first row of the matrix V is output as the life cycle feature of the target user account.

[0085] The input of the gating network includes the portrait features and the life cycle features output by the vector layer. It can determine the first weight corresponding to the payment feature in the portrait features and the second weight corresponding to the non-payment feature according to the payment status represented by the life cycle features for output. For example, for users whose payment status is never paid, the first weight will be much smaller than the second weight, thereby increasing the influence of non-payment features on the LTV prediction results. For users whose payment status is currently paid, the gating network will appropriately increase the ratio of the first weight to the second weight, thereby increasing the influence of payment features on the LTV prediction results.

[0086] The input of the payment amount prediction network includes the portrait features and the first weight and the second weight output by the gating network. The payment amount prediction network can weight the payment features according to the first weight, and weight the non-payment features according to the second weight, and then generate and output the LTV prediction result based on the payment features and non-payment features obtained by weighted processing. The LTV prediction result is the total future payment amount predicted for the target user account.

[0087] In the above embodiment, since the payment characteristics of different users reflect the differences in payment behaviors between different users, the weights of the user's payment characteristics and non-payment characteristics are controlled through a gated network, so as to control the influence of the payment characteristics and non-payment characteristics on the LTV prediction results based on the corresponding weights, so that the generated LTV prediction results can be more in line with the user's payment characteristics. This prediction method fully considers the impact of the differences in payment behaviors of different users on the LTV prediction results, and can significantly improve the accuracy of the LTV prediction results, so that accurate LTV prediction results can be generated for users with different payment behaviors. By predicting accurate LTV prediction results for users with different payment behaviors, it is helpful to finely divide users within the platform / system according to the LTV prediction results, and carry out targeted marketing strategy delivery to improve the marketing delivery effect.

[0088] In some possible implementations, the payment amount prediction network may include a prediction layer and at least one feature processing layer, wherein each feature processing layer corresponds to a different feature dimension and each feature processing layer corresponds to a gating network. Figure 4 FIG. 1 is a schematic diagram of another LTV prediction model provided by an exemplary embodiment. Figure 4 In the example, the payment amount prediction network includes a prediction layer (logit layer) and three feature processing layers (neural layer 1 to 3), and each feature processing layer corresponds to a gating network.

[0089] The life cycle data and portrait features of the target user account are input into the LTV prediction model. First, the life cycle data is converted into life cycle features through the embedding layer. At the same time, the portrait features are processed nonlinearly through the feature input layer. For example, the ReLu activation function is used to process the portrait features nonlinearly. The principle of this process can be referred to formula (1):

[0090] X out =relu(W*X in +b) (1)

[0091] In formula (1), X in represents the image features, W represents the weight of the feature input layer, b represents the bias of the feature input layer, X out Characterize the image features after nonlinear processing.

[0092] Then, the image features after nonlinear processing are concatenated with the life cycle features as the input of the gating network. The input and structure of the three gating networks are consistent, but the parameters are different. In this way, different gating networks can learn features of different feature dimensions. For example, the gating network can use a two-layer neural network, the first layer uses the relu activation function, and the second layer uses the sigmiod activation function, such as Figure 5 It should be noted that Figure 5 This is just an example, and the present disclosure does not specifically limit the specific structure of the gated network. The working principle of any gated network can refer to formula (2):

[0093] g=Gate(concat(X u ,v)) (2)

[0094] In formula (2), X u It represents the image features after nonlinear processing, v represents the life cycle features, concat represents the concatenation operation, and g represents the first weight and the second weight of the gating network output.

[0095] For any feature processing layer, its input includes the output of the previous feature processing layer and the output of the corresponding gating network. The input of the first feature processing layer is the image features after nonlinear processing (including paid features and non-paid features) and the first weight and second weight of the corresponding gating network output. In order to facilitate the splicing of the output of the feature processing layer and the output of the corresponding gating network, the feature dimension corresponding to any feature processing layer is consistent with the feature dimension corresponding to the corresponding gating network. For example, the feature dimensions corresponding to the feature processing layer neural layer1 and the gate network gate network1 are both 20 dimensions, the feature dimensions corresponding to the feature processing layer neural layer2 and the gate network gate network2 are both 50 dimensions, and the feature dimensions corresponding to the feature processing layer neural layer3 and the gate network gate network3 are both 100 dimensions. The principle of the above splicing process can be referred to formula (3):

[0096]

[0097] In formula (3), t represents the tth feature processing layer, Represents the output of the (t-1)th feature processing layer or the image features after nonlinear processing, g t Characterize the first and second weights of the output of the t-th layer gating network, Characterize the concatenated features corresponding to the t-th feature processing layer.

[0098] Since each feature processing layer corresponds to a different feature dimension, any feature processing layer is used to perform feature analysis on its corresponding concatenated features and output features on its own corresponding feature dimension. The principle of the feature analysis process can be referred to formula (4):

[0099]

[0100] In formula (4), W t Characterizes the weight of the t-th feature processing layer, b t Characterizes the bias of the t-th feature processing layer, Characterize the concatenated features corresponding to the t-th feature processing layer, Represents the output of the t-th feature processing layer.

[0101] The output of the last feature processing layer is input into the prediction layer to obtain the LTV prediction result output by the prediction layer. The principle of the prediction process can be referred to formula (5):

[0102] V = W / logit buy +b(5)

[0103] In formula (5), W represents the weight of the prediction layer, b represents the bias of the prediction layer, and logit buy represents the input of the prediction layer, and V represents the LTV prediction result output by the prediction layer.

[0104] In this embodiment, by setting feature processing layers and gating networks of different dimensions, the depth and width of the LTV prediction model are increased, which can significantly improve the expressiveness of the LTV prediction model and enable it to learn more complex features. In addition, the image features and life cycle data input into the LTV prediction model can be gradually extracted from low-level features to high-level features, which helps to capture more comprehensive and accurate feature representations, so as to generate more accurate LTV prediction results based on more comprehensive and accurate feature representations.

[0105] In some possible implementations, the LTV prediction model may also include at least one payment behavior prediction subnetwork, and each payment behavior prediction subnetwork corresponds to a preset time window. The preset time window can be understood as a preset duration. As the name implies, the payment behavior prediction subnetwork is used to predict whether the target user account will pay (i.e., whether there is payment behavior) within the corresponding preset time window. Exemplarily, the input of each payment behavior prediction subnetwork may include the portrait feature and the first weight and the second weight output by the gating network. The payment behavior prediction subnetwork can generate a payment behavior prediction result of the target user account within the corresponding preset time window based on the portrait feature and the first weight and the second weight, and the prediction result includes the payment probability of the target user account within the corresponding preset time window. Furthermore, the payment behavior prediction results generated by each payment behavior prediction subnetwork are input into the payment amount prediction network, so that the payment amount prediction network generates the LTV prediction result based on these inputs and the payment features and non-payment features obtained by weighted processing.

[0106] Figure 6a FIG. 1 is a schematic diagram of a structure of an LTV prediction model including a payment behavior prediction subnetwork provided by an exemplary embodiment. Figure 6a In the example, each payment behavior prediction subnetwork is directly connected to the payment amount prediction network. During the prediction process, the input of each payment behavior prediction subnetwork is the first weight and the second weight output by the image feature and the gating network. The payment amount prediction network can predict the payment behavior prediction results (i.e., payment probability) output by each payment behavior prediction subnetwork, the payment features and non-payment features after weighted processing, and generate the LTV prediction results corresponding to the target user account. Assuming that the time window corresponding to the payment amount prediction network is 180 days, the payment amount prediction network can predict the total payment amount of the target user account in the next 180 days. The preset time windows corresponding to the six payment behavior prediction subnetworks (Buy1, Buy7, Buy30, Buy60, Buy90, ​​Buy180) are 1 day, 7 days, 30 days, 60 days, 90 days, and 180 days respectively. In other words, these six payment behavior prediction subnetworks can predict the payment probability of the target user account in the next 1 day, the next 7 days, the next 30 days, the next 60 days, the next 90 days, and the next 180 days respectively.

[0107] Figure 6b FIG. 1 is a schematic diagram of another LTV prediction model including a payment behavior prediction subnetwork provided by an exemplary embodiment. Figure 6bIn the example, six payment behavior prediction subnetworks (Buy1, Buy7, Buy30, Buy60, Buy90, ​​Buy180) are connected step by step according to the length of the corresponding preset time window, and the payment behavior prediction subnetwork with the largest preset time window (i.e., Buy180) is connected to the payment amount prediction network. During the prediction process, the input of the payment behavior prediction subnetwork (i.e., Buy1) farthest from the payment amount prediction network is the portrait feature, the first weight and the second weight output by the gating network. Then, the payment behavior prediction subnetwork Buy1 will input the generated payment behavior prediction result into the payment behavior prediction subnetwork Buy7. That is, the input of the payment behavior prediction subnetwork Buy7 includes the portrait feature, the first weight and the second weight output by the gating network, and the payment behavior prediction result generated by the payment behavior prediction subnetwork Buy1, and then the payment behavior prediction subnetwork Buy7 will input the payment behavior prediction result generated by itself into the payment behavior prediction subnetwork Buy30. By analogy, except for the payment behavior prediction subnetwork Buy1, the inputs of other payment behavior prediction subnetworks include portrait features, the first weight and the second weight output by the gating network, and the payment behavior prediction results generated by the previous payment behavior prediction subnetwork. In this case, the input of the payment amount prediction network includes the payment behavior prediction results output by the payment behavior prediction subnetwork Buy180, and the weighted payment features and non-payment features. This hierarchical connection architecture allows each payment behavior prediction subnetwork to learn the feature representations of the portrait features and life cycle features in the previous payment behavior prediction subnetwork, thereby improving the accuracy of the payment probability output by each payment behavior prediction subnetwork. At the same time, on the basis of ensuring that the payment amount prediction network learns the payment probability features within multiple preset time windows, the amount of data that needs to be processed by the payment amount prediction network can be further reduced, thereby improving the prediction efficiency. It should be noted that Figure 6a and Figure 6b All time windows involved are only examples, and the present disclosure does not specifically limit the time windows.

[0108] In this embodiment, at least one payment behavior prediction subnetwork is set in the LTV prediction model, so as to split the goal of predicting user LTV into the goal of predicting the total payment amount and the goal of predicting multiple payment probabilities. Compared with directly predicting the total payment amount, this embodiment assists in predicting the total payment amount by predicting multiple payment probabilities under the framework of multi-objective / multi-task joint learning, which helps to further improve the accuracy of the predicted total payment amount.

[0109] Figure 6c FIG. 1 is a structural diagram of a payment behavior prediction subnetwork provided by an exemplary embodiment. Figure 6cAs shown, taking the architecture of each payment behavior prediction subnetwork connected step by step as an example, each payment behavior prediction subnetwork may include a probability generation layer (sigmoid layer), a prediction layer (logit layer) and multiple feature processing layers (neurallayer), each feature processing layer corresponds to a different feature dimension and each feature processing layer corresponds to a gating network. The following is an exemplary description of the payment behavior prediction subnetwork Buy1: In the payment behavior prediction subnetwork Buy1, the feature dimension corresponding to each feature processing layer is consistent with the feature dimension corresponding to the corresponding gating network. The input of the first feature processing layer neurallayer11 includes the portrait feature and the first weight and the second weight output by the first gating network. The portrait feature is concatenated with the output of the first gating network, and then the concatenated features are input into the first feature processing layer neural layer11 to obtain the feature representation on the feature dimension corresponding to the first feature processing layer 11. Except for the first feature processing layer neurallayer11, the inputs of other feature processing layers include the feature representation output by the previous feature processing layer and the first weight and the second weight output by the corresponding gating network. The prediction layer logit layer can make predictions based on the feature representation output by the last feature processing layer. The principle of the above process can be referred to formula (6):

[0110]

[0111] In formula (6), represents the output after being processed by the three-layer feature processing layer in the payment behavior prediction subnetwork Buy1, W represents the weight of the prediction layer in the payment behavior prediction subnetwork Buy1, b represents the bias of the prediction layer in the payment behavior prediction subnetwork Buy1, logit buy1 Represents the output of the prediction layer in the payment behavior prediction subnetwork Buy1.

[0112] Then, the output of the prediction layer is input into the probability generation layer sigmoid layer to obtain the payment probability of the target user account in the next day. The principle can be referred to in formula (7):

[0113] p buy1 =sigmoid(logit buy1 )(7)

[0114] In formula (7), logit buy1 Represents the output of the prediction layer in the payment behavior prediction subnetwork Buy1, p buy1Characterizes the payment probability output by the payment behavior prediction subnetwork Buy1. This structural setting of the payment behavior prediction subnetwork can help the payment behavior subnetwork learn the feature representation of the portrait features and life cycle features in different feature dimensions, thereby improving the diversity and accuracy of feature learning, and helping to improve the accuracy of the payment probability output by the payment behavior prediction subnetwork.

[0115] The input and output of each feature processing layer and probability generation layer in other payment behavior prediction subnetworks (Buy7, Buy30, Buy60, Buy90, ​​Buy180) can refer to the payment behavior prediction subnetwork Buy1, and will not be repeated here. The input of the prediction layer in other payment behavior prediction subnetworks includes the feature representation of the output of the previous feature processing layer in its own network, and the output of the prediction layer of the previous payment behavior prediction subnetwork. That is, the input of the prediction layer in the payment behavior prediction subnetwork Buy7 includes the feature representation of the output of the previous feature processing layer in its own network, and the output of the prediction layer in the payment behavior prediction subnetwork Buy1. The input of the prediction layer in the payment behavior prediction subnetwork Buy30 includes the feature representation of the output of the previous feature processing layer in its own network, and the output of the prediction layer in the payment behavior prediction subnetwork Buy7, and so on.

[0116] Combination Figure 6c For the payment behavior prediction subnetwork Buy7, the working principle of its prediction layer can be referred to formula (8-1), and the working principle of its probability generation layer can be referred to formula (8-2):

[0117]

[0118] p buy7 =sigmoid(logit buy7 ) (8-2)

[0119] In formulas (8-1) and (8-2), logit buy1 Represents the output of the prediction layer in the payment behavior prediction subnetwork Buy1, represents the output after being processed by the three-layer feature processing layer in the payment behavior prediction sub-network Buy7, concat represents the concatenation operation, W represents the weight of the prediction layer in the payment behavior prediction sub-network Buy7, b represents the bias of the prediction layer in the payment behavior prediction sub-network Buy7, logit buy7 Represents the output of the prediction layer in the payment behavior prediction subnetwork Buy7, p buy7 Represents the payment probability output by the payment behavior prediction subnetwork Buy7.

[0120] For the payment behavior prediction subnetwork Buy30, the working principle of its prediction layer can be referred to formula (9-1), and the working principle of its probability generation layer can be referred to formula (9-2):

[0121]

[0122] p buy30 =sigmoid(logit buy30 ) (9-2)

[0123] In formulas (9-1) and (9-2), logit buy7 Represents the output of the prediction layer in the payment behavior prediction subnetwork Buy7, represents the output after being processed by the three-layer feature processing layer in the payment behavior prediction subnetwork Buy30, concat represents the concatenation operation, W represents the weight of the prediction layer in the payment behavior prediction subnetwork Buy30, b represents the bias of the prediction layer in the payment behavior prediction subnetwork Buy30, logit buy30 Represents the output of the prediction layer in the payment behavior prediction subnetwork Buy30, p buy30 Represents the payment probability output by the payment behavior prediction subnetwork Buy30.

[0124] For the payment behavior prediction subnetwork Buy60, the working principle of its prediction layer can be referred to formula (10-1), and the working principle of its probability generation layer can be referred to formula (10-2):

[0125]

[0126] p buy60 =sigmoid(logit buy60 ) (10-2)

[0127] In formulas (10-1) and (10-2), logit buy30 Represents the output of the prediction layer in the payment behavior prediction subnetwork Buy30, represents the output after being processed by the three-layer feature processing layer in the payment behavior prediction subnetwork Buy60, concat represents the concatenation operation, W represents the weight of the prediction layer in the payment behavior prediction subnetwork Buy60, b represents the bias of the prediction layer in the payment behavior prediction subnetwork Buy60, logit buy60 Represents the output of the prediction layer in the payment behavior prediction subnetwork Buy60, p buy60 Represents the payment probability output by the payment behavior prediction subnetwork Buy60.

[0128] For the payment behavior prediction subnetwork Buy90, ​​the working principle of its prediction layer can be referred to formula (11-1), and the working principle of its probability generation layer can be referred to formula (11-2):

[0129] logit buy90 =W*concat(Xb 3 uy90out ,logit buy60 )+b(11-1)

[0130] p buy90 =sigmoid(logit buy90 ) (11-2)

[0131] In formulas (11-1) and (11-2), logit buy60 Represents the output of the prediction layer in the payment behavior prediction subnetwork Buy60, represents the output after being processed by the three-layer feature processing layer in the payment behavior prediction sub-network Buy90, ​​concat represents the concatenation operation, W represents the weight of the prediction layer in the payment behavior prediction sub-network Buy90, ​​b represents the bias of the prediction layer in the payment behavior prediction sub-network Buy90, ​​logit buy90 Represents the output of the prediction layer in the payment behavior prediction subnetwork Buy90, ​​p buy90 Represents the payment probability output by the payment behavior prediction subnetwork Buy90.

[0132] For the payment behavior prediction subnetwork Buy180, the working principle of its prediction layer can be referred to formula (12-1), and the working principle of its probability generation layer can be referred to formula (12-2):

[0133]

[0134] p buy180 =sigmoid(logit buy180 ) (12-2)

[0135] In formulas (12-1) and (12-2), logit buy90 Represents the output of the prediction layer in the payment behavior prediction subnetwork Buy90, represents the output after being processed by the three-layer feature processing layer in the payment behavior prediction subnetwork Buy180, concat represents the concatenation operation, W represents the weight of the prediction layer in the payment behavior prediction subnetwork Buy180, b represents the bias of the prediction layer in the payment behavior prediction subnetwork Buy180, logit buy180 Represents the output of the prediction layer in the payment behavior prediction subnetwork Buy180, p buy180 Represents the payment probability output by the payment behavior prediction subnetwork Buy180.

[0136] exist Figure 6c The working principle of the payment amount prediction network can be referred to formula (13):

[0137]

[0138] In formula (13), logit buy180 represents the output of the prediction layer in the payment behavior prediction subnetwork Buy180, W represents the weight of the prediction layer in the payment amount prediction network, b represents the bias of the prediction layer in the payment amount prediction network, Represents the output after processing by the three-layer feature processing layer in the payment amount prediction network. V represents the LTV prediction result output by the prediction layer, that is, the total payment amount of the target user account in the next 180 days.

[0139] In some possible implementations, the LTV prediction model may further include at least one payment amount prediction subnetwork, each payment amount prediction subnetwork corresponds to a preset time window, and all preset time windows are different from the time windows corresponding to the payment amount prediction network. Figure 7a FIG. 1 is a schematic diagram of a structure of an LTV prediction model including a payment amount prediction subnetwork provided by an exemplary embodiment. Figure 7a As shown, the payment amount prediction network is LTV, and the payment amount prediction sub-networks include LTV1, LTV7, LTV30, LTV60, and LTV90. Figure 7a In the schematic diagram of the model structure shown in the figure, each payment amount prediction subnetwork is independent of each other, and their inputs are the same, including the portrait features, the first weight and the second weight output by the gating network. By weighting the payment features with the first weight, the weighted payment features can be obtained; by weighting the non-payment features with the second weight, the weighted non-payment features can be obtained. Any payment amount prediction subnetwork can generate the LTV prediction result (i.e., the total payment amount) of the target user account in the corresponding preset time window according to its own input. Assuming that the time window corresponding to LTV is 180 days, the preset time windows corresponding to these payment amount prediction subnetworks (LTV1, LTV7, LTV30, LTV60, LTV90) are 1 day, 7 days, 30 days, 60 days, and 90 days respectively. Therefore, LTV can predict the total payment amount of the target user account in the next 180 days, and these five payment amount prediction subnetworks can respectively predict the total payment amount of the target user account in the next 1 day, the next 7 days, the next 30 days, the next 60 days, and the next 90 days. In this way, it is possible to use an LTV prediction model to predict the total payment amount of the same target user account in different time windows, which significantly improves the prediction range of the LTV prediction model and expands the versatility of the LTV prediction model. In addition, users can control the number of total payment amounts output by the LTV prediction model according to their own needs, such as controlling the LTV prediction model to only output the total payment amount within the next 30 days and 60 days, which helps meet the user's personalized needs for predicting LTV.

[0140] Figure 7b FIG. 1 is a schematic diagram of another LTV prediction model including a payment amount prediction subnetwork provided by an exemplary embodiment. Figure 7b As shown, the five payment amount prediction subnetworks (LTV1, LTV7, LTV30, LTV60, LTV90) can be connected step by step according to the length of the corresponding preset time window, and the payment behavior prediction subnetwork (i.e., LTV90) with the largest preset time window is connected to the payment amount prediction network (i.e., LTV). During the prediction process, the input of the payment amount prediction subnetwork (i.e., LTV 1) farthest from the payment amount prediction network is the portrait feature, the first weight and the second weight output by the gating network. Then, the payment amount prediction subnetwork LTV 1 will input the generated LTV prediction result into the payment amount prediction subnetwork LTV7. That is, the input of the payment amount prediction subnetwork LTV7 includes the portrait feature, the first weight and the second weight output by the gating network, and the LTV prediction result generated by the payment amount prediction subnetwork LTV1. Then, the payment amount prediction subnetwork LTV7 will input the LTV prediction result generated by itself into the payment amount prediction subnetwork LTV30. By analogy, except for the payment amount prediction subnetwork LTV1, the inputs of other payment amount prediction subnetworks include portrait features, the first weight and the second weight output by the gating network, and the LTV prediction results generated by the previous payment amount prediction subnetwork. In this case, the input of the payment amount prediction network LTV includes the LTV prediction results output by the payment amount prediction subnetwork LTV90, and the weighted payment features and non-payment features. This hierarchical connection architecture allows each payment amount prediction subnetwork to learn the feature representations of the portrait features and life cycle features in the previous payment amount prediction subnetwork, thereby improving the accuracy of the total payment amount output by each payment amount prediction subnetwork. At the same time, it can ensure that the payment amount prediction network learns the payment amount features within multiple preset time windows, further improving the accuracy of the LTV prediction results output by the payment amount prediction network.

[0141] In some other possible implementations, the LTV prediction model may include payment behavior prediction subnetworks and payment amount prediction subnetworks corresponding to multiple preset time windows. Figure 8aThe preset time window corresponding to the payment amount prediction network LTV is 180 days, and the preset time windows corresponding to the two payment amount prediction sub-networks (LTV60, LTV90) are 60 days and 90 days respectively. The preset time windows corresponding to the three payment behavior prediction sub-networks (Buy 60, Buy 90, Buy180) are 60 days, 90 days and 180 days respectively. The payment behavior prediction sub-network corresponding to the same preset time window is interconnected with the payment amount prediction sub-network / payment amount prediction network, and except for the network Buy60, the payment behavior prediction sub-network corresponding to each preset time window is also connected to the payment amount prediction sub-network corresponding to another preset time window.

[0142] In this case, for the payment amount prediction subnetwork LTV60, its input includes: portrait features, the first weight and the second weight output by the gating network, and the payment behavior prediction result output by the payment behavior prediction subnetwork (i.e. Buy60) corresponding to the same preset time window; the output is the total payment amount of the target user account in the next 60 days. For any other payment amount prediction subnetwork / payment amount prediction network, its input includes: portrait features, the first weight and the second weight output by the gating network, the LTV prediction result output by the previous payment amount prediction subnetwork, and the payment behavior prediction result output by the payment behavior prediction subnetwork corresponding to the same preset time window. Among them, the previous payment amount prediction subnetwork refers to: the payment amount prediction subnetwork under the preset time window adjacent to the preset time window corresponding to itself. For example, for the payment amount prediction network LTV, the corresponding preset time window is 180 days, then its previous payment amount prediction subnetwork is the payment amount prediction subnetwork LTV90 under the preset time window of 90 days. The payment behavior prediction sub-network corresponding to the same preset time window refers to the payment behavior prediction sub-network Buy180 corresponding to the preset time window of 180 days.

[0143] For the payment behavior prediction subnetwork Buy 60, its input includes portrait features, the first weight and the second weight output by the gating network, and the output is the payment probability of the target user account in the next 60 days. For the payment behavior prediction subnetworks Buy 90 and Buy180, the input includes: portrait features, the first weight and the second weight output by the gating network, and the payment behavior prediction results output by the previous payment behavior prediction subnetwork; the output is the payment probability of the target user account within the corresponding preset time window. Buy 60 is the previous payment behavior prediction subnetwork of Buy90, ​​and Buy 90 is the previous payment behavior prediction subnetwork of Buy180. It should be noted that in order to facilitate the display of the model structure, Figure 8a The total payment amount output by each payment amount prediction subnetwork and the payment probability output by each payment behavior prediction subnetwork are not shown.

[0144] based on Figure 8a The model structure shown can use an LTV prediction model to output the payment probability and total payment amount in multiple preset time windows. And through this multi-task / objective learning architecture, the prediction difficulty of each task can be further reduced, and the prediction accuracy and efficiency of each task can be improved.

[0145] exist Figure 8a Based on this, we can learn the image features and life cycle features from multiple feature dimensions to further improve the accuracy of the prediction results. Figure 8b , Figure 8b In the figure, the payment amount prediction subnetwork LTV90 and the payment behavior prediction subnetwork Buy90 are taken as examples to illustrate the structures of the two networks. Figure 8b The input and output of each feature processing layer neural network and the input and output of the probability generation layer in the payment behavior prediction subnetwork Buy90 can be referenced. Figure 6c The illustrated embodiments will not be described in detail here. The input of the prediction layer (logit layer) in the payment behavior prediction subnetwork Buy90 includes: the output of the last feature processing layer contained in the network itself, and the payment behavior prediction result output by the previous payment behavior prediction subnetwork (i.e. Buy60). The input of the prediction layer (logit layer) in the payment amount prediction subnetwork LTV90 includes: the output of the last feature processing layer contained in the network itself, the payment behavior prediction result output by the payment behavior prediction subnetwork (i.e. Buy90) in the same preset time window, and the LTV prediction result output by the previous payment amount prediction subnetwork (i.e. LTV60).

[0146] Corresponding to the above-mentioned data prediction method, in another embodiment of the present disclosure, a data prediction model training method is also provided, through which the LTV prediction model is trained to implement the above-mentioned data prediction method based on the trained LTV prediction model.

[0147] See also Fig. 9 , Fig. 9 The following is a flowchart of a method for training a data prediction model provided by an exemplary embodiment. The method may include the following steps:

[0148] Step 901: Obtain sample portrait data, sample life cycle data and sample LTV of the sample user account, and generate sample portrait features corresponding to the sample portrait data; wherein the sample portrait features include sample payment features and sample non-payment features, and the sample life cycle data is used to characterize the sample payment status of the sample user account.

[0149] The sample user accounts may include user accounts that have paid, user accounts that have never paid, and user accounts whose payment status has expired. In this embodiment, a sample user account on date t may be selected, and then the sample portrait data, sample life cycle data, and sample LTV of the sample user account in the corresponding platform / system are obtained. The sample life cycle data is used to characterize the sample payment status of the sample user account on date t. The sample LTV within the preset time period refers to the sample LTV (i.e., the total amount of sample payment) of the sample user account after the preset time period from date t, such as obtaining the sample LTV within [t, t+180 days].

[0150] Different payment states can be numbered, for example, new members are numbered 1, valid members are numbered 2, and lost members are numbered 3. The number corresponding to the sample user account is used as the sample life cycle data, recorded as stage_id. The sample portrait data of each sample user account can be feature extracted to generate the corresponding sample portrait features, recorded as vector X. In this way, the sample composition corresponding to each sample user account is {t, user_id, stage_id, X, LTV}, where user_id is the sample user account.

[0151] Step 902: Input the sample portrait features and the sample life cycle data into the LTV prediction model to be trained to obtain the sample LTV prediction result corresponding to the sample user account. The LTV prediction model to be trained includes: a gating network, which is used to determine the first sample weight corresponding to the sample payment feature and the second sample weight corresponding to the sample non-payment feature according to the sample payment status represented by the sample life cycle data; a payment amount prediction network, which is used to weight the sample payment feature according to the first sample weight, and weight the sample non-payment feature according to the second sample weight, and generate the sample LTV prediction result based on the sample payment feature and sample non-payment feature obtained by weighted processing.

[0152] The specific implementation method of this step can be found in Figure 2 The relevant contents in the illustrated embodiment will not be repeated here.

[0153] Step 903: Calculate an LTV loss value based on the sample LTV and the sample LTV prediction result, and iteratively train the LTV prediction model to be trained according to the LTV loss value.

[0154] In this embodiment, when the sample LTV is the sample total payment amount of the sample user account after 180 days from date t, the sample LTV prediction result is used to represent the predicted total payment amount of the sample user account after 180 days from date t. The MSE loss function can be used to calculate the LTV loss value. The principle can refer to formula (14):

[0155]

[0156] In formula (14), i represents the i-th sample user account, N represents the total number of sample user accounts, and LTV i Represents the sample LTV corresponding to the i-th sample user account, V i It represents the sample LTV prediction result corresponding to the i-th sample user account, and ltv_loss represents the LTV loss value.

[0157] Through multiple iterations of training, when the obtained LTV prediction model meets the predefined training objectives (such as minimizing the LTV loss value) or reaches the predefined number of iterations, the training of the LTV prediction model is completed and can be used to achieve the following Figure 2 It should be noted that, in this embodiment, when gradient backpropagation is performed, in order not to affect the weight training of the sample image feature, the gradient weight from the gated network will not be backpropagated to the sample image feature X.

[0158] In some possible implementations, the LTV prediction model to be trained may also include at least one payment behavior prediction subnetwork, and each payment behavior prediction subnetwork corresponds to a sample preset time window. In this way, the LTV prediction model to be trained can learn the payment probability within multiple different time windows. Therefore, in this case, the sample portrait data of the sample user account may also include sample payment behaviors within multiple sample preset time windows. Still taking date t as an example, the sample portrait data may also include the payment behavior of the sample user account within [t, t+k], and the value of k may be 1, 7, 30, 60, 90, 180. If the sample user account has paid within the corresponding sample preset time window, the sample payment behavior under the sample preset time window is recorded as 1; if the sample user account has not paid within the corresponding sample preset time window, the sample payment behavior under the sample preset time window is recorded as 0. The sample composition corresponding to each sample user account can be {t, user_id, stage_id, X, LTV, Buy1, Buy7, Buy30, Buy60, Buy90, ​​Buy180}, where Buy1~Buy180 are used to characterize the payment behavior of the sample user account within multiple sample preset time windows.

[0159] Combination Figure 6a-6cDuring the training process, any payment behavior prediction subnetwork is used to generate a sample payment behavior prediction result (i.e., sample payment probability) of a sample user account within the corresponding sample preset time window based on the sample portrait features and the first sample weight and the second sample weight output by the gating network. The payment amount prediction network can be used to generate a sample LTV prediction result based on the sample payment features and sample non-payment features obtained through weighted processing, as well as the sample payment behavior prediction results generated by each payment behavior prediction subnetwork. For specific implementation methods, please refer to Figure 2 The relevant contents in the illustrated embodiment will not be repeated here.

[0160] In this embodiment, by setting at least one payment behavior prediction sub-network in the LTV prediction model to be trained, the training goal of predicting user LTV is split into a training goal of predicting the total payment amount and a plurality of training goals of predicting the payment probability, which greatly simplifies the training difficulty of the LTV prediction model and improves the model training efficiency. In addition, in the related art, there is only one training goal of predicting user LTV, and it is possible that the model training effect is poor due to the excessively high sample LTV of some sample user accounts. In this embodiment, even if the sample LTV of some sample user accounts is too high, this part of the sample data will only have some impact on the payment amount prediction network in the LTV prediction model, and will not affect the learning of the payment behavior prediction sub-network, thereby further improving the prediction accuracy of the LTV prediction model after training.

[0161] In some possible implementations, when the LTV prediction model to be trained includes at least one payment behavior prediction subnetwork, it is also necessary to calculate the payment behavior loss value of each payment behavior prediction subnetwork. Exemplarily, the sample payment behavior within each sample preset time window is obtained, and the payment behavior loss value corresponding to each sample preset time window is calculated based on the sample payment behavior and the sample payment behavior prediction result corresponding to each sample preset time window. For example, the ordinary cross entropy loss function can be used to calculate the payment behavior loss value. The principle can refer to formula (15):

[0162]

[0163] In formula (15), t represents the preset time window of the tth sample, buy t Characterizes the sample payment behavior corresponding to the preset time window of the tth sample, Represents the sample payment behavior prediction result corresponding to the preset time window of the tth sample, CrossEntropy represents the ordinary cross entropy function, and buy_loss represents the sum of the payment behavior loss values ​​corresponding to the preset time windows of each sample.

[0164] Then, the total loss value of the model is calculated based on the LTV loss value and the payment behavior loss value corresponding to the preset time window of each sample, so as to iteratively train the LTV prediction model to be trained based on the total loss value. The principle of this calculation process can be referred to formula (16):

[0165]

[0166] In formula (16), ltv_loss represents the LTV loss value, and the meaning of buy_loss is the same as that of formula (15). Represents the preset loss weight, and loss represents the total loss value.

[0167] In this way, the payment amount prediction network and each payment behavior prediction sub-network can be optimized at the same time, enriching the learnable feature representation of the LTV prediction model, enhancing the generalization ability of the model, improving the stability of model training, and reducing the risk of gradient disappearance or explosion.

[0168] In some possible implementations, the LTV prediction model to be trained may also include at least one payment amount prediction subnetwork, each payment amount prediction subnetwork corresponds to a sample preset time window, and all sample preset time windows are different from the sample time windows corresponding to the payment amount prediction network. In this way, the LTV prediction model to be trained can learn the payment amount in multiple different time windows. Therefore, in this case, the sample portrait data of the sample user account may also include the sample LTV (i.e., the sample payment amount) in multiple sample preset time windows. Still taking date t as an example, the sample portrait data may also include the payment amount of the sample user account in [t, t+k], and the value of k may be 1, 7, 30, 60, 90, 180. The sample composition corresponding to each sample user account may be {t, user_id, stage_id, X, LTV, LTV1, LTV 7, LTV 30, LTV 60, LTV 90}, where LTV 1 to LTV90 are respectively used to characterize the sample LTV of the sample user account in multiple sample preset time windows.

[0169] Combination Figure 7a-7b During the training process, any payment amount prediction subnetwork is used to generate the sample LTV prediction result (i.e., sample predicted payment amount) of the sample user account within the corresponding sample preset time window based on the sample portrait features and the first sample weight and the second sample weight output by the gating network. For specific implementation methods, please refer to Figure 2 The relevant contents in the illustrated embodiment will not be repeated here.

[0170] In this embodiment, by setting at least one payment amount prediction subnetwork in the LTV prediction model to be trained, the training goal of predicting user LTV is split into training goals of predicting payment amounts in multiple different time windows, which greatly simplifies the training difficulty of the LTV prediction model and improves the model training efficiency.

[0171] In some possible implementations, when the LTV prediction model to be trained includes at least one payment amount prediction subnetwork, it is also necessary to calculate the LTV loss value of each payment amount prediction subnetwork. Exemplarily, the sample LTV in each sample preset time window is obtained respectively, and the LTV loss value corresponding to each sample preset time window is calculated respectively according to the sample LTV corresponding to each sample preset time window and the sample LTV prediction result. The calculation principle can refer to formula (14).

[0172] Then, the total loss value of the model is calculated based on the LTV loss value of the payment amount prediction network and the LTV loss value of each payment amount prediction sub-network, so as to iteratively train the LTV prediction model to be trained based on the total loss value. In this way, the payment amount prediction network and each payment amount prediction sub-network can be optimized at the same time, enriching the learnable feature representation of the LTV prediction model, enhancing the generalization ability of the model, improving the stability of model training, and reducing the risk of gradient disappearance or explosion.

[0173] Exemplary Devices

[0174] In an exemplary embodiment of the present disclosure, a data prediction device is also provided.

[0175] See also Fig.10 , Fig.10 It is a block diagram of a data prediction device provided by an exemplary embodiment.

[0176] like Fig.10 As shown, the data prediction device 1000 may include: an acquisition module 1001 and an input module 1002. Among them:

[0177] The acquisition module 1001 is used to acquire the portrait data and life cycle data of the target user account, and generate portrait features corresponding to the portrait data; the portrait features include payment features and non-payment features, and the life cycle data is used to characterize the payment status of the target user account.

[0178] The input module 1002 is used to input the portrait features and the life cycle data into the LTV prediction model to obtain the LTV prediction result corresponding to the target user account; wherein the LTV prediction model includes: a gating network, which is used to determine the first weight corresponding to the payment feature and the second weight corresponding to the non-payment feature according to the payment status represented by the life cycle data; a payment amount prediction network, which is used to weight the payment feature according to the first weight, and weight the non-payment feature according to the second weight, and generate the LTV prediction result based on the payment feature and non-payment feature obtained by weighted processing.

[0179] In one embodiment, the portrait data includes static attribute data, historical active behavior data and payment data of the target user account, and the acquisition module 1001 is specifically used to: extract features from the payment data to obtain the payment features; extract features from the static attribute data and the historical active behavior data respectively to obtain the non-payment features, which include static attribute features corresponding to the static attribute data and historical active behavior features corresponding to the historical active behavior data; normalize the payment features and the non-payment features respectively, and concatenate the processed payment features with the processed non-payment features to obtain the portrait features.

[0180] In one embodiment, the LTV prediction model also includes at least one payment behavior prediction subnetwork, each payment behavior prediction subnetwork corresponding to a preset time window; wherein any payment behavior prediction subnetwork is used to: generate a payment behavior prediction result of the target user account within the corresponding preset time window based on the portrait features and the first weight and the second weight output by the gating network; the payment amount prediction network is also used to: generate the LTV prediction result based on the payment features and non-payment features obtained by weighted processing, and the payment behavior prediction results generated by each payment behavior prediction subnetwork.

[0181] In one embodiment, the LTV prediction model also includes at least one payment amount prediction subnetwork, each payment amount prediction subnetwork corresponds to a preset time window, and all preset time windows are different from the time window corresponding to the payment amount prediction network; wherein, any payment amount prediction subnetwork is used to: generate the LTV prediction result of the target user account within the corresponding preset time window based on the payment features and non-payment features obtained by weighted processing.

[0182] In one embodiment, the payment amount prediction network includes a prediction layer and at least one feature processing layer, each feature processing layer corresponds to a different feature dimension and each feature processing layer corresponds to a gating network; wherein any feature processing layer is used to: perform feature analysis based on the output of the previous feature processing layer and the first weight and the second weight output by the corresponding gating network, and output the features of the corresponding feature dimension; the input of the first feature processing layer is the payment feature, the non-payment feature and the first weight and the second weight output by the corresponding gating network; the prediction layer is used to: perform prediction based on the output of the last feature processing layer to generate the LTV prediction result.

[0183] In an exemplary embodiment of the present disclosure, a training device for a data prediction model is also provided.

[0184] See also Fig.11 , Fig.11 It is a block diagram of a data prediction model training device provided by an exemplary embodiment.

[0185] like Fig.11 As shown, the data prediction model training device 1100 may include: a sample acquisition module 1101, an input module 1102 and a training module 1103. Among them:

[0186] The sample acquisition module 1101 is used to obtain sample portrait data, sample life cycle data and sample LTV of the sample user account, and generate sample portrait features corresponding to the sample portrait data; wherein the sample portrait features include sample payment features and sample non-payment features, and the sample life cycle data is used to characterize the sample payment status of the sample user account.

[0187] The input module 1102 is used to input the sample portrait features and the sample life cycle data into the LTV prediction model to be trained to obtain the sample LTV prediction result corresponding to the sample user account; wherein the LTV prediction model to be trained includes: a gating network, which is used to determine the first sample weight corresponding to the sample payment feature and the second sample weight corresponding to the sample non-payment feature according to the sample payment status represented by the sample life cycle data; a payment amount prediction network, which is used to weight the sample payment feature according to the first sample weight, and weight the sample non-payment feature according to the second sample weight, and generate the sample LTV prediction result based on the sample payment feature and sample non-payment feature obtained by weighted processing.

[0188] The training module 1103 is used to calculate the LTV loss value based on the sample LTV and the sample LTV prediction result, and iteratively train the LTV prediction model to be trained according to the LTV loss value.

[0189] In one embodiment, the LTV prediction model to be trained also includes at least one payment behavior prediction subnetwork, and each payment behavior prediction subnetwork corresponds to a sample preset time window; wherein any payment behavior prediction subnetwork is used to: generate a sample payment behavior prediction result of the sample user account within the corresponding sample preset time window based on the sample portrait features and the first sample weight and the second sample weight output by the gating network; the payment amount prediction network is also used to: generate the sample LTV prediction result based on the sample payment features and sample non-payment features obtained by weighted processing, and the sample payment behavior prediction results generated by each payment behavior prediction subnetwork.

[0190] In one embodiment, the apparatus 1100 further includes:

[0191] The payment behavior loss value calculation module 1104 is used to obtain the sample payment behavior within each sample preset time window respectively, and calculate the payment behavior loss value corresponding to each sample preset time window according to the sample payment behavior corresponding to each sample preset time window and the sample payment behavior prediction result.

[0192] The training module 1103 is specifically used to iteratively train the LTV prediction model to be trained according to the LTV loss value and the payment behavior loss value corresponding to each sample preset time window.

[0193] In one embodiment, the LTV prediction model to be trained also includes at least one payment amount prediction subnetwork, each payment amount prediction subnetwork corresponds to a sample preset time window, and all sample preset time windows are different from the sample time windows corresponding to the payment amount prediction network; wherein, any payment amount prediction subnetwork is used to: generate a sample LTV prediction result of the sample user account within the corresponding sample preset time window based on the sample payment features and sample non-payment features obtained by weighted processing.

[0194] In one embodiment, the apparatus 1100 further includes:

[0195] The LTV loss value calculation module 1105 is used to obtain the sample LTV in each sample preset time window respectively, and calculate the LTV loss value corresponding to each sample preset time window according to the sample LTV corresponding to each sample preset time window and the sample LTV prediction result.

[0196] The training module 1103 is specifically used to iteratively train the LTV prediction model to be trained according to the LTV loss value and the LTV loss value corresponding to each sample preset time window.

[0197] The specific details of the above-mentioned data prediction device 1000 and each module of the data prediction model training device 1100 have been described in detail in the previous description of the data prediction method process, so they will not be repeated here.

[0198] It should be noted that, although the data prediction device 1000 and several modules or units of the training device 1100 of the data prediction model are mentioned in the above detailed description, this division is not mandatory. In fact, according to the embodiments of the present disclosure, the features and functions of two or more modules or units described above can be concretized in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided into multiple modules or units to be concretized.

[0199] Exemplary Media

[0200] In an exemplary embodiment of the present disclosure, a computer-readable storage medium is also provided, on which a program product capable of implementing the above method of the present specification is stored. In some possible embodiments, various aspects of the present disclosure may also be implemented in the form of a program product, which includes a program code, and when the program product is run on a terminal device, the program code is used to enable the terminal device to perform the steps according to various exemplary embodiments of the present disclosure described in the above "Exemplary Method" section of the present specification.

[0201] See also Fig.12 , Fig.12 It is a schematic diagram of a readable storage medium corresponding to a data prediction method or a data prediction model training method provided by an exemplary embodiment.

[0202] refer to Fig.12 As shown, a readable storage medium 1200 for implementing the above method according to an embodiment of the present disclosure is described, which can adopt a portable compact disk read-only memory (CD-ROM) and include program code, and can be run on a terminal device, such as a personal computer. However, the readable storage medium of the present disclosure is not limited thereto. In the present disclosure, the readable storage medium can be any tangible medium containing or storing a program, which can be used by or in combination with an instruction execution system, an apparatus or a device.

[0203] The readable storage medium can adopt any combination of one or more readable media. The readable medium can be a readable signal medium or a readable storage medium. The readable storage medium can be, for example, but not limited to, a system, device or device of electricity, magnetism, light, electromagnetic, infrared, or semiconductor, or any combination of the above. More specific examples (non-exhaustive list) of readable storage media include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0204] Computer readable signal media may include data signals propagated in baseband or as part of a carrier wave, in which readable program code is carried. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Readable signal media may also be any readable medium other than a readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0205] The program code embodied on the readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wired, optical cable, RF, etc., or any suitable combination of the foregoing.

[0206] Program code for performing the operations of the present disclosure may be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java, C++, etc., and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user computing device, partially on the user device, as a separate software package, partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server. In cases involving a remote computing device, the remote computing device may be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computing device (e.g., through the Internet using an Internet service provider).

[0207] Exemplary Electronic Devices

[0208] In an exemplary embodiment of the present disclosure, an electronic device capable of implementing the above-mentioned data prediction method or data prediction model training method is also provided.

[0209] See also Fig.13 , Fig.13It is a schematic diagram of an electronic device capable of implementing the above method provided by an exemplary embodiment.

[0210] Refer to the following Fig.13 1300 according to this embodiment of the present disclosure is described. Fig.13 The electronic device 1300 shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0211] like Fig.13 As shown, the electronic device 1300 is in the form of a general computing device. The components of the electronic device 1300 may include but are not limited to: at least one processing unit 1301, at least one storage unit 1302, and a bus 1303 connecting different system components (including the storage unit 1302 and the processing unit 1301).

[0212] The storage unit stores program codes, and the program codes can be executed by the processing unit 1301, so that the processing unit 1301 executes the steps of the above-mentioned various embodiments of this specification.

[0213] The storage unit 1302 may include a readable medium in the form of a volatile storage unit, such as a random access storage unit (RAM) 13021 and / or a cache storage unit 13022 , and may further include a read-only storage unit (ROM) 13023 .

[0214] The storage unit 1302 may also include a program / utility 13024 having a set (at least one) of program modules 13025, such program modules 13025 including but not limited to: an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include the reality of a network environment.

[0215] Bus 1303 may represent one or more of several types of bus structures, including a memory unit bus or memory unit controller, a peripheral bus, an accelerated graphics port, a processing unit, or a local bus using any of a variety of bus architectures.

[0216] The electronic device 1300 may also communicate with one or more external devices 1304 (e.g., keyboards, pointing devices, Bluetooth devices, etc.), may communicate with one or more devices that enable a user to interact with the electronic device 1300, and / or may communicate with any device that enables the electronic device 1300 to communicate with one or more other computing devices (e.g., routers, modems, etc.). Such communication may be performed via an input / output (I / O) interface 1305. Furthermore, the electronic device 1300 may also communicate with one or more networks (e.g., a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) via a network adapter 1306. As shown, the network adapter 1306 communicates with other modules of the electronic device 1300 via a bus 1303. It should be understood that although Fig.13 Not shown, other hardware and / or software modules may be used in conjunction with electronic device 1300, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0217] Exemplary computer program products

[0218] In an exemplary embodiment of the present disclosure, a computer program product is further provided, including a computer program / instruction, which implements the steps of the method described in any of the above embodiments when executed by a processor.

[0219] Through the description of the above embodiments, it is easy for those skilled in the art to understand that the example embodiments described here can be implemented by software, or by software combined with necessary hardware. Therefore, the technical solution according to the embodiment of the present disclosure can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, and includes several instructions to enable a computing device (which can be a personal computer, a server, a terminal device, or a network device, etc.) to execute the method according to the embodiment of the present disclosure.

[0220] Those skilled in the art will readily appreciate other embodiments of the present disclosure after considering the specification and practicing the disclosure disclosed herein. This application is intended to cover any variations, uses, or adaptations of the present disclosure that follow the general principles of the present disclosure and include common knowledge or customary technical means in the art that are not disclosed in the present disclosure. The specification and examples are to be considered exemplary only, and the true scope and spirit of the present disclosure are indicated by the claims.

[0221] It should be noted that, although several units / modules or subunits / modules of the device are mentioned in the above detailed description, this division is merely exemplary and not mandatory. In fact, according to the embodiments of the present disclosure, the features and functions of two or more units / modules described above can be embodied in one unit / module. Conversely, the features and functions of one unit / module described above can be further divided into multiple units / modules to be embodied.

[0222] In addition, although the operations of the disclosed method are described in a specific order in the drawings, this does not require or imply that the operations must be performed in this specific order, or that all the operations shown must be performed to achieve the desired results. Additionally or alternatively, some steps may be omitted, multiple steps may be combined into one step, and / or one step may be decomposed into multiple steps.

[0223] Although the spirit and principle of the present disclosure have been described with reference to several specific embodiments, it should be understood that the present disclosure is not limited to the disclosed specific embodiments, and the division of various aspects does not mean that the features in these aspects cannot be combined to benefit, and such division is only for the convenience of expression. The present disclosure is intended to cover various modifications and equivalent arrangements included in the spirit and scope of the attached claims.

Claims

1. A data prediction method, characterized in that: The method comprises: Acquire the portrait data and life cycle data of the target user account, and generate the portrait features corresponding to the portrait data; the portrait features include payment features and non-payment features, and the life cycle data is used to characterize the payment status of the target user account; The portrait features and the life cycle data are input into the LTV prediction model to obtain the LTV prediction result corresponding to the target user account; wherein the LTV prediction model includes: A gating network, configured to determine a first weight corresponding to the payment feature and a second weight corresponding to the non-payment feature according to the payment status represented by the lifecycle data; The payment amount prediction network is used to weight the payment feature according to the first weight, and to weight the non-payment feature according to the second weight, and to generate the LTV prediction result based on the payment feature and non-payment feature obtained by weighted processing.

2. The method according to claim 1, characterized in that The portrait data includes static attribute data, historical active behavior data and payment data of the target user account, and the generation of the portrait features corresponding to the portrait data includes: Extracting features from the payment data to obtain the payment features; Extracting features from the static attribute data and the historical active behavior data respectively to obtain the non-payment features, where the non-payment features include static attribute features corresponding to the static attribute data and historical active behavior features corresponding to the historical active behavior data; The paid features and the non-paid features are normalized respectively, and the processed paid features and the processed non-paid features are concatenated to obtain the portrait features.

3. The method according to claim 1, characterized in that The LTV prediction model also includes at least one payment behavior prediction sub-network, each payment behavior prediction sub-network corresponds to a preset time window; wherein, Any payment behavior prediction sub-network is used to: generate a payment behavior prediction result of the target user account within a corresponding preset time window according to the portrait feature and the first weight and the second weight output by the gating network; The payment amount prediction network is also used to generate the LTV prediction result based on the payment features and non-payment features obtained through weighted processing and the payment behavior prediction results generated by each payment behavior prediction sub-network.

4. The method according to claim 1, characterized in that: The LTV prediction model further includes at least one payment amount prediction subnetwork, each payment amount prediction subnetwork corresponds to a preset time window, and all preset time windows are different from the time windows corresponding to the payment amount prediction network; wherein, Any payment amount prediction subnetwork is used to generate an LTV prediction result of the target user account within a corresponding preset time window according to the payment features and non-payment features obtained through weighted processing.

5. The method according to claim 1, characterized in that The payment amount prediction network includes a prediction layer and at least one feature processing layer, each feature processing layer corresponds to a different feature dimension and each feature processing layer corresponds to a gating network; Among them, any feature processing layer is used to: perform feature analysis based on the output of the previous feature processing layer and the first weight and the second weight output by the corresponding gating network, and output the features of the corresponding feature dimension; the input of the first feature processing layer is the paid feature, the non-paid feature and the first weight and the second weight output by the corresponding gating network; The prediction layer is used to make predictions based on the output of the last feature processing layer to generate the LTV prediction result.

6. A method for training a data prediction model, characterized in that: The method includes: Obtain sample portrait data, sample life cycle data and sample LTV of the sample user account, and generate sample portrait features corresponding to the sample portrait data; wherein the sample portrait features include sample payment features and sample non-payment features, and the sample life cycle data is used to characterize the sample payment status of the sample user account; Inputting the sample portrait features and the sample life cycle data into the LTV prediction model to be trained to obtain the sample LTV prediction result corresponding to the sample user account; Calculating an LTV loss value based on the sample LTV and the sample LTV prediction result, and iteratively training the LTV prediction model to be trained according to the LTV loss value; Wherein, the LTV prediction model to be trained includes: A gating network, used to determine a first sample weight corresponding to the sample payment feature and a second sample weight corresponding to the sample non-payment feature according to the sample payment status represented by the sample life cycle data; The payment amount prediction network is used to weight the sample payment features according to the first sample weight, and to weight the sample non-payment features according to the second sample weight, and to generate the sample LTV prediction result based on the sample payment features and sample non-payment features obtained by weighted processing.

7. The method according to claim 6, characterized in that The LTV prediction model to be trained also includes at least one payment behavior prediction sub-network, each payment behavior prediction sub-network corresponds to a sample preset time window; wherein, Any payment behavior prediction sub-network is used to: generate a sample payment behavior prediction result of the sample user account within a preset time window of the corresponding sample according to the sample portrait feature and the first sample weight and the second sample weight output by the gating network; The payment amount prediction network is also used to generate the sample LTV prediction result based on the sample payment features and sample non-payment features obtained through weighted processing, and the sample payment behavior prediction results generated by each payment behavior prediction subnetwork.

8. An electronic device, characterized in that: include: processor; a memory for storing instructions executable by the processor; The processor implements the method according to any one of claims 1 to 7 by running the executable instructions.

9. A computer-readable storage medium having computer instructions stored thereon, characterized in that: When the instruction is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

10. A computer program product, characterized in that The method comprises a computer program / instruction, which, when executed by a processor, implements the steps of the method according to any one of claims 1 to 7.

Images