Model training method and device, business execution method and device and storage medium

By combining the user's historical LTV and behavioral data in the prediction model, predicting future behavioral characteristics and LTV, the problem of low prediction accuracy in the existing technology is solved, and more accurate business strategy adjustment is achieved.

CN119066418AInactive Publication Date: 2024-12-03GUANGZHOU SANQI DREAM NETWORK TECH CO LTD
View PDF 5 Cites 0 Cited by

Patent Information

Application Number
CN202411054656.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-01
Publication Date
2024-12-03
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

The prior art is low in predicting the total life cycle value (LTV) of users in the future, mainly because it only refers to the user's LTV for the past period and lacks behavioral data analysis.

Method used

By obtaining the total value and behavioral data of the user's historical life cycle, input it into the prediction model, using a multi-layer structure model (including feature extraction layer, prediction layer and regression layer) to predict the behavioral characteristics and LTV of the user's specified time in the future, and according to The deviation between the predicted value and the label value is trained in model to improve the prediction accuracy.

Benefits of technology

By combining historical LTV and behavioral data for comprehensive analysis, the prediction accuracy of users' LTVs at specified time in the future has been significantly improved, helping enterprises to adjust their business strategies more effectively.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119066418A_ABST
    Figure CN119066418A_ABST
Patent Text Reader

Abstract

According to the model training method, the business execution method, the business execution device and the storage medium provided by the invention, the historical LTV of the user for the target business and the behavior data of the user in the corresponding historical time period can be acquired, and the historical LTV and the behavior data are input into the to-be-trained prediction model, so that the target business can be predicted through the first prediction layer in the prediction model. The method comprises the following steps: predicting a first behavior feature of a user in a specified time in the future through a first prediction layer, predicting a second behavior feature of the user in the specified time in the future through a second prediction layer, inputting the first behavior feature and the second behavior feature into a regression layer, determining an LTV of the user in the specified time in the future through the regression layer, and taking the LTV as a predicted LTV, according to the deviation between the predicted LTV and the label LTV corresponding to the user in the specified time in the future, a loss value is determined, the prediction model is trained according to the loss value, and the deviation and the loss value are in a positive correlation relationship.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a model training method, a business execution method, a device, and a storage medium. Background Art

[0002] At present, the business strategy can be adjusted according to the total life cycle value (LTV) of users within a period of time and the customer acquisition cost (CAC), so that the adjusted business strategy is more popular with users, thereby improving the user experience. Among them, LTV can be used to measure the benefits brought by users to the enterprise, and CAC can be used to measure the cost of the enterprise to encourage users to use the services it provides.

[0003] In actual application, it is often necessary to adjust the business strategy based on the user's LTV over a longer period of time. For example, the user's LTV within 90 days can be obtained, and when the user's LTV within 90 days is significantly less than the user's corresponding CAC, it can be determined that the current business strategy is ineffective and can be adjusted.

[0004] In the above process of obtaining the user's LTV, it usually takes a long time. For example, if you need to obtain the user's LTV within 90 days, it takes 90 days to obtain it. In order to solve this problem, a neural network model can be used to predict the user's LTV for a period of time in the future.

[0005] However, in the prior art, when predicting the LTV of a user in the future, a simple prediction is made by referring to the LTV of the user in the past, and the prediction accuracy is low. Summary of the invention

[0006] This specification provides a model training method, a business execution method, a device and a storage medium to partially solve the above-mentioned problems existing in the prior art.

[0007] This manual adopts the following technical solutions:

[0008] This specification provides a model training method, including:

[0009] Obtain the user's historical life cycle total value LTV for the target business and the user's behavior data for the target business in the historical period corresponding to the historical LTV;

[0010] Input the historical LTV and the behavior data into the prediction model to be trained, so as to predict the first behavior feature of the user at a specified future time according to the historical LTV and the behavior data through a first prediction layer included in the prediction model, and predict the second behavior feature of the user at the specified future time according to the historical LTV and the behavior data through a second prediction layer included in the prediction model, and input the first behavior feature and the second behavior feature into the regression layer included in the prediction model, so as to determine the LTV of the user at the specified future time through the regression layer as the predicted LTV, the first behavior feature is used to characterize whether the user performs a specified business operation for the target business at the specified future time, the second behavior feature is used to characterize whether the user continues to perform the target business at the specified future time, and the specified business operation is a business operation that prompts the user to continue to perform the target business;

[0011] A loss value is determined based on a deviation between the predicted LTV and the tag LTV corresponding to the user at the specified future time, and the prediction model is trained based on the loss value, wherein there is a positive correlation between the deviation and the loss value.

[0012] Optionally, the step of obtaining the user's historical lifetime value LTV for the target business and the user's behavior data for the target business in a historical period corresponding to the historical LTV specifically includes:

[0013] Obtaining the user's initial historical LTV for the target business and the user's initial behavior data for the target business during the historical period;

[0014] The initial historical LTV is preprocessed to use the processed data as the acquired historical LTV of the user for the target business, and the initial behavior data is preprocessed to use the processed data as the behavior data of the user for the target business during the historical period, wherein the preprocessing includes at least one of deduplication processing, normalization processing, and outlier processing.

[0015] Optionally, the step of inputting the historical LTV and the behavior data into a prediction model to be trained, so as to predict the first behavior feature of the user at a specified future time according to the historical LTV and the behavior data through a first prediction layer included in the prediction model, specifically includes:

[0016] Inputting the historical LTV and the behavior data into a prediction model to be trained, so as to extract user features from the historical LTV and the behavior data through a feature extraction layer included in the prediction model;

[0017] The user features are input into a first prediction layer included in the prediction model, so as to predict a first behavior feature of the user at a specified future time according to the user features through the first prediction layer.

[0018] Optionally, the step of inputting the historical LTV and the behavior data into a prediction model to be trained, and predicting the second behavior feature of the user at the specified future time according to the historical LTV and the behavior data through a second prediction layer included in the prediction model, specifically includes:

[0019] Inputting the historical LTV and the behavior data into a prediction model to be trained, so as to extract user features from the historical LTV and the behavior data through a feature extraction layer included in the prediction model;

[0020] The user features are input into a second prediction layer included in the prediction model, so as to predict the second behavior features of the user at the specified future time according to the user features through the second prediction layer.

[0021] Optionally, the step of inputting the first behavior feature and the second behavior feature into a regression layer included in the prediction model to determine the LTV of the user at the specified future time through the regression layer as a step of predicting the LTV specifically includes:

[0022] The user feature, the first behavior feature, and the second behavior feature are input into a regression layer included in the prediction model, so as to determine the LTV of the user at the future specified time through the regression layer as the predicted LTV.

[0023] Optionally, before the step of determining a loss value according to a deviation between the predicted LTV and the tag LTV corresponding to the user at the specified future time, and training the prediction model according to the loss value, the method further includes:

[0024] Inputting the behavior data into the behavior feature layer in the prediction model to obtain the behavior features corresponding to the user;

[0025] Inputting the behavior feature into the prediction layer in the prediction model to predict the behavior data of the user at the specified future time as predicted behavior data;

[0026] The step of determining a loss value according to a deviation between the predicted LTV and the tag LTV corresponding to the user at the specified future time, and training the prediction model according to the loss value specifically includes:

[0027] A loss value is determined based on the deviation between the predicted LTV and the label LTV corresponding to the user at the specified future time, and the deviation between the predicted behavior data and the actual behavior data of the user at the specified future time, and the prediction model is trained based on the loss value, wherein the deviation between the predicted behavior data and the actual behavior data of the user at the specified future time is positively correlated with the loss value.

[0028] This specification provides a service execution method, including:

[0029] Obtain the user's historical LTV for the target business and the user's behavior data for the target business in the historical period corresponding to the historical LTV;

[0030] Inputting the historical LTV and the behavior data into a pre-trained prediction model so that the prediction model determines the predicted LTV of the user at a specified time in the future, wherein the prediction model is trained by the above-mentioned model training method;

[0031] A service strategy for the user is determined based on the predicted LTV, and the target service is executed according to the service strategy.

[0032] This specification provides a model training device, including:

[0033] Acquisition module: used to acquire the historical life cycle total value LTV of the user for the target business and the behavior data of the user for the target business in the historical period corresponding to the historical LTV;

[0034] Prediction module: used for inputting the historical LTV and the behavior data into the prediction model to be trained, so as to predict the first behavior feature of the user at a specified future time according to the historical LTV and the behavior data through a first prediction layer included in the prediction model, and predict the second behavior feature of the user at the specified future time according to the historical LTV and the behavior data through a second prediction layer included in the prediction model, and inputting the first behavior feature and the second behavior feature into the regression layer included in the prediction model, so as to determine the LTV of the user at the specified future time through the regression layer as the predicted LTV, the first behavior feature is used to characterize whether the user performs a specified business operation for the target business at the specified future time, the second behavior feature is used to characterize whether the user continues to perform the target business at the specified future time, and the specified business operation is a business operation that prompts the user to continue to perform the target business;

[0035] Training module: used to determine a loss value according to the deviation between the predicted LTV and the second historical LTV, and train the target model according to the loss value, wherein there is a positive correlation between the deviation and the loss value.

[0036] This specification provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the above-mentioned model training method or business execution method.

[0037] This specification provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned model training method or business execution method when executing the program.

[0038] At least one of the above technical solutions adopted in this specification can achieve the following beneficial effects:

[0039] The model training method provided in this specification can obtain the user's historical LTV for the target business and the user's behavior data in the corresponding historical period, and input the historical LTV and behavior data into the prediction model to be trained, so as to predict the first behavior feature of the user at a specified future time through the first prediction layer in the prediction model, and predict the second behavior feature of the user at a specified future time through the second prediction layer, and input the first behavior feature and the second behavior feature into the regression layer, so as to determine the user's LTV at a specified future time through the regression layer as the predicted LTV, determine the loss value according to the deviation between the predicted LTV and the label LTV corresponding to the user at the specified future time, and train the prediction model according to the loss value, wherein the deviation and the loss value are positively correlated.

[0040] It can be seen from this that in the process of model training using the above method, instead of simply making a prediction based on the user's LTV in the past period of time as in the prior art, in addition to inputting the user's historical LTV, the user's behavior data in the historical period is also used as the input of the model, so that a more comprehensive analysis of the user's behavior in the historical period can be performed, and the results of the analysis can be used to predict the user's LTV at a specified time in the future, thereby improving the prediction accuracy of the trained model for the user's LTV at a specified time in the future. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] The drawings described herein are used to provide a further understanding of this specification and constitute a part of this specification. The illustrative embodiments and descriptions of this specification are used to explain this specification and do not constitute an improper limitation on this specification. In the drawings:

[0042] Figure 1 A flowchart of a model training method provided in this specification;

[0043] Figure 2 A schematic diagram of a model training process provided in this manual;

[0044] Figure 3 A flowchart of a business execution method provided in this specification;

[0045] Figure 4 A schematic diagram of a model training device provided in this specification;

[0046] Figure 5 A schematic diagram of a service execution device provided in this specification;

[0047] Figure 6 A method corresponding to the Figure 1 or Figure 3 Schematic diagram of the structure of an electronic device. DETAILED DESCRIPTION

[0048] In order to make the purpose, technical solutions and advantages of this specification more clear, the technical solutions of this specification will be clearly and completely described below in combination with the specific embodiments of this specification and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of this specification, not all of them. Based on the embodiments in this specification, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this specification.

[0049] At present, in actual application, it is often necessary to adjust the business strategy based on the user's LTV over a long period of time. For example, the user's LTV within 90 days can be obtained, and when the user's LTV within 90 days is significantly less than the user's corresponding CAC, it can be determined that the current business strategy is ineffective and can be adjusted.

[0050] In the above process of obtaining the user's LTV, it usually takes a long time. For example, if you need to obtain the user's LTV within 90 days, it takes 90 days to obtain it. In order to solve this problem, a neural network model can be used to predict the user's LTV for a period of time in the future.

[0051] However, in the prior art, when predicting the LTV of a user in the future, a simple prediction is made by referring to the LTV of the user in the past, and the prediction accuracy is low.

[0052] Therefore, this specification provides a model training method, which can effectively solve the problem of low accuracy in predicting the LTV of users in the future in the prior art.

[0053] The technical solutions provided by the embodiments of this specification are described in detail below in conjunction with the accompanying drawings.

[0054] Figure 1 A flow chart of a model training method provided in this specification includes the following steps:

[0055] S101: Obtain the historical lifetime total value (LTV) of a user for a target business and the behavior data of the user for the target business in a historical period corresponding to the historical LTV.

[0056] The execution subject of the model training method involved in this specification can be a terminal device such as a desktop computer, a laptop computer, or a client installed in the terminal device, or a server. The following only takes the server as the execution subject as an example to illustrate the model training method in the embodiment of this specification.

[0057] In this specification, the server may use the historical LTV of the user for the target business and the user's behavior data for the target business in the historical period corresponding to the historical LTV as sample data to train the prediction model to be trained. For example, the user's LTV for the target business within 3 days and the user's behavior data within these 3 days may be used as sample data for subsequent model training.

[0058] In actual application, in order to improve the training effect of the model, the data to be input into the model can often be preprocessed, and the processed data can be input into the model as sample data for subsequent training.

[0059] Therefore, the server may first pre-process the acquired initial historical LTV of the user for the target business, so as to use the processed data as the historical LTV of the user for the target business. Also, the acquired initial behavior data of the user for the target business during the historical period may be pre-processed, so as to use the processed data as the behavior data of the user for the target business during the historical period. For example, the pre-processing may be to perform de-duplication processing on it. For another example, the pre-processing may be to perform normalization processing on it. For another example, the pre-processing may be to perform outlier processing on it, etc.

[0060] Furthermore, the determined historical LTV of the user for the target business and the user's behavior data for the target business in the historical period can be used as sample data for subsequent model training.

[0061] In this specification, the target business can be the corresponding business provided by the enterprise (such as: game product experience business), and the behavior data can be the behavior data generated during the interaction between the user and the target business (such as: the user's online time, the record data of the user's operation of related products, etc.).

[0062] S102: Input the historical LTV and the behavior data into the prediction model to be trained, so as to predict the first behavior feature of the user at a specified future time according to the historical LTV and the behavior data through a first prediction layer included in the prediction model, and predict the second behavior feature of the user at the specified future time according to the historical LTV and the behavior data through a second prediction layer included in the prediction model, and input the first behavior feature and the second behavior feature into the regression layer included in the prediction model, so as to determine the LTV of the user at the specified future time through the regression layer as the predicted LTV, the first behavior feature is used to characterize whether the user performs a specified business operation for the target business at the specified future time, the second behavior feature is used to characterize whether the user continues to perform the target business at the specified future time, and the specified business operation is a business operation that prompts the user to continue to perform the target business.

[0063] The server can input the historical LTV and the user's behavior data in the corresponding historical period into the prediction model to be trained, and then, the first prediction layer included in the prediction model can predict the user's first behavior feature at a specified time in the future based on the historical LTV and the user's behavior data in the corresponding historical period. Among them, the first behavior feature can characterize whether the user performs a specified business operation at a specified time in the future. Performing a specified business operation can refer to the user performing the specified business provided by the enterprise to the user. For example, the user can go to punch in every day to receive rewards, the user can participate in daily competitions to improve their own level, the user can obtain the products of interest through redemption, and the user can purchase the desired products through recharge, etc. These can all be called the above-mentioned specified business operations. In other words, the first behavior feature can characterize whether the user will punch in every day at a specified time in the future (such as: the next 5 days), or whether to obtain the product of interest through redemption.

[0064] Furthermore, the server can predict the second behavior feature of the user at a specified future time based on the historical LTV and the user's behavior data in the corresponding historical period through the second prediction layer (such as the Softmax layer) included in the prediction model. The second behavior feature can characterize whether the user continues to perform the target business at a specified future time. For example, the second behavior feature can characterize whether the user continues to participate in the game product experience business within the next 5 days. Then, the first behavior feature and the second behavior feature can be input into the regression layer included in the prediction model, and then the user's LTV at a specified future time can be determined through the regression layer, and the determined user's LTV at a specified future time is used as the predicted LTV.

[0065] In this specification, the prediction model may also include a feature extraction layer. After the historical LTV and the user's behavior data in the corresponding historical period are input into the prediction model to be trained, the user features may be first extracted from the historical LTV and the user's behavior data in the corresponding historical period through the feature extraction layer. The user features may characterize the features of the user's LTV and behavior data when the user is in the corresponding historical period. For example, the user features may characterize the speed of change of the user's LTV in the corresponding historical period, the fluctuation of the user's online time in the corresponding historical period, etc.

[0066] Then, the extracted user features can be input into the first prediction layer included in the prediction model, and then the first prediction layer can be used to predict the first behavior features of the user at a specified time in the future based on the user features. Also, the extracted user features can be input into the second prediction layer included in the prediction model, and then the second prediction layer can be used to predict the second behavior features of the user at a specified time in the future based on the user features, and the first behavior features and the second behavior features can be input into the regression layer included in the prediction model to determine the predicted LTV.

[0067] In addition, in order to improve the training effect of the model, in this specification, in addition to determining the predicted LTV by inputting the first behavior feature and the second behavior feature into the regression layer included in the prediction model, the user features extracted by the feature extraction layer can also be input into the regression layer, so that the regression layer can obtain more comprehensive features of the users in the corresponding historical period, so as to determine a more accurate LTV of the user in the future specified time through the regression layer as the predicted LTV, thereby improving the training effect of the model.

[0068] S103: Determine a loss value based on a deviation between the predicted LTV and the tag LTV corresponding to the user at the specified future time, and train the prediction model based on the loss value, wherein there is a positive correlation between the deviation and the loss value.

[0069] The server can determine the loss value during the training process based on the deviation between the predicted LTV and the tag LTV corresponding to the user at a specified time in the future. The deviation is positively correlated with the loss value. The server can train the prediction model by gradually reducing the loss value. Accordingly, the trained prediction model can be used to predict the user's LTV at a specified time in the future.

[0070] It should be noted that the training data used in the training process and the tag LTV corresponding to the specified time in the future can come from historical data or from a pre-set training set, and this specification does not limit this.

[0071] In addition, in order to make the LTV of the user predicted by the prediction model at a specified future time more accurate, in the process of training the prediction model, in addition to referring to the deviation between the predicted LTV and the label LTV corresponding to the user at the specified future time, the prediction model can also be used to predict the user's behavior data at a specified future time, and the loss value can be determined by referring to the deviation between the predicted user's behavior data at a specified future time and the actual behavior data corresponding to the specified future time, so as to train the prediction model based on the loss value.

[0072] Specifically, the prediction model may also include a behavior feature layer and a prediction layer. During the model training process, the server may input the user's behavior data for the target business in the corresponding historical period into the behavior feature layer of the prediction model, and then determine the user's corresponding behavior features, which may characterize the user's various behaviors in the corresponding historical period (such as: the frequency of the user's execution of the target business in the historical period, the user's operation of redeeming the product of interest in the historical period, etc.).

[0073] Then, the behavior features can be input into the prediction layer of the prediction model to predict the user's behavior data at a specified time in the future as the predicted behavior data. For example, the predicted behavior data can be the frequency of the user performing the target business in the next 50 days. Another example is whether the user will perform daily check-ins during the execution of the target business in the next 50 days, whether he will redeem the products he is interested in, etc.

[0074] Accordingly, when determining the loss value in the process of training the prediction model, in addition to determining it based on the deviation between the predicted LTV and the tag LTV corresponding to the user's specified time in the future, it is also necessary to determine it based on the deviation between the predicted behavior data and the actual behavior data corresponding to the user's specified time in the future. That is, the loss value can be determined based on these two deviations, and the prediction model can be trained by gradually reducing the loss value. Among them, the deviation between the predicted behavior data and the actual behavior data corresponding to the user's specified time in the future is positively correlated with the loss value.

[0075] It can be seen from the above method that in the process of training the prediction model, it is not like the prior art that only refers to the user's LTV in the past period for simple prediction, but in addition to inputting the user's historical LTV, the user's behavior data in the historical period is also used as the input of the model, so that a more comprehensive analysis of the user's behavior in the historical period can be performed, and the results of the analysis can be used to predict the user's LTV at a specified time in the future, thereby improving the prediction accuracy of the trained model for the user's LTV at a specified time in the future.

[0076] In addition, in the above model training process, in addition to building the first prediction layer, the second prediction layer, and the regression layer in the same prediction model, they can also be built in different models. The corresponding model training process can be specifically referred to as follows Figure 2 .

[0077] Figure 2 A schematic diagram of a model training process provided in this manual.

[0078] Depend on Figure 2It can be seen that the first prediction model can be constructed by the first prediction layer, the second prediction model can be constructed by the second prediction layer, and the regression model can be constructed by the regression layer. First, the user's historical LTV and the user's behavior data in the historical period corresponding to the historical LTV can be used as training data, and the training data can be input into the first prediction model, the second prediction model and the regression model respectively. The first prediction model can input the predicted first behavior feature into the regression model, and the second prediction model can input the predicted second behavior feature into the regression model. The regression model can extract user features from the user's historical LTV and the user's behavior data in the historical period corresponding to the historical LTV, and determine the user's predicted LTV at a specified time in the future based on the user features, the first behavior features and the second behavior features, and jointly train the first prediction model, the regression model and the second prediction model by gradually reducing the predicted LTV and the label LTV corresponding to the specified time in the future as the training goal. Furthermore, the user's LTV at a specified time in the future can be predicted by the trained regression model.

[0079] The above mainly introduces the model training method. After the model training is completed, it can be deployed for business execution. Next, a business execution method provided in this specification will be described.

[0080] Figure 3 A flowchart of a business execution method provided in this specification includes the following steps:

[0081] S301: Acquire the historical LTV of a user for a target service and the behavior data of the user for the target service in a historical period corresponding to the historical LTV.

[0082] The execution subject of the service execution method involved in this specification can be a terminal device such as a desktop computer, a laptop computer, etc., or a client installed in the terminal device, or a server. The following only takes the server as an execution subject as an example to illustrate the service execution method in the embodiment of this specification.

[0083] In actual application, the server may be deployed with a prediction model pre-trained by the above model training method, which can predict the user's predicted LTV at a specified time in the future based on the user's historical LTV and the user's behavior data in the corresponding historical period, and then determine the corresponding business strategy based on the predicted LTV, and execute the target business according to the business strategy. For example, the user's LTV in the next 90 days can be predicted based on the user's LTV within 3 days and the behavior data within 3 days, and then the business strategy can be determined based on the predicted LTV.

[0084] S302: Inputting the historical LTV and the behavior data into a pre-trained prediction model so that the prediction model determines the predicted LTV of the user at a specified time in the future, wherein the prediction model is trained by the above-mentioned model training method.

[0085] In this specification, the prediction model deployed in the server can be used to determine the user's predicted LTV at a specified time in the future based on the input user's historical LTV for the target business and the user's behavior data for the target business in the historical time period corresponding to the historical LTV.

[0086] Of course, in order to determine a more reasonable business strategy for users, it is often necessary to collect the historical LTV of a large number of users and the behavioral data in the corresponding historical time period. In order to improve efficiency, the historical LTV of multiple users can be averaged, and the behavioral data of each user in the corresponding historical time period can be integrated. The average value and the integrated behavioral data can be input into the prediction model to determine the corresponding business strategy according to the prediction results of the prediction model.

[0087] S303: Determine a service strategy for the user according to the predicted LTV, and execute the target service according to the service strategy.

[0088] After determining the predicted LTV, a more reasonable business strategy for the user can be determined based on the predicted LTV, and the corresponding target business (such as the game business provided by the enterprise to the user) can be executed according to the business strategy. In actual application, the predicted LTV determined by the prediction model can often be compared with the CAC corresponding to the user, and the business strategy can be determined based on the comparison results. For example, the predicted LTV for the next 90 days can be compared with the CAC corresponding to the user. If the predicted LTV for the next 90 days is significantly lower than the CAC corresponding to the user, it can reflect that the user's current satisfaction may be low, or the user's interest in the services provided by the enterprise is low. To this end, the corresponding business strategy can be determined (such as simplifying the operating procedures of the current target business, beautifying the current service interface, etc.), and then the target business can be executed according to the business strategy.

[0089] The above are one or more implementation methods of this specification. Based on the same idea, this specification also provides a corresponding model training device and a business execution device, such as Figure 4 , Figure 5 shown.

[0090] Figure 4 A schematic diagram of a model training device provided in this specification includes:

[0091] Acquisition module 401: used to acquire the historical life cycle total value LTV of the user for the target business and the behavior data of the user for the target business in the historical period corresponding to the historical LTV;

[0092] Prediction module 402: used for inputting the historical LTV and the behavior data into the prediction model to be trained, so as to predict the first behavior feature of the user at a specified future time according to the historical LTV and the behavior data through the first prediction layer included in the prediction model, and predict the second behavior feature of the user at the specified future time according to the historical LTV and the behavior data through the second prediction layer included in the prediction model, and inputting the first behavior feature and the second behavior feature into the regression layer included in the prediction model, so as to determine the LTV of the user at the specified future time through the regression layer as the predicted LTV, the first behavior feature is used to characterize whether the user performs a specified business operation for the target business at the specified future time, the second behavior feature is used to characterize whether the user continues to perform the target business at the specified future time, and the specified business operation is a business operation that prompts the user to continue to perform the target business;

[0093] Training module 403: used to determine a loss value according to the deviation between the predicted LTV and the second historical LTV, and train the target model according to the loss value, wherein there is a positive correlation between the deviation and the loss value.

[0094] Optionally, the acquisition module 401 is specifically used for the step of obtaining the behavioral data for the target business within the historical period, specifically including: obtaining the user's initial historical LTV for the target business and the user's initial behavioral data for the target business within the historical period; preprocessing the initial historical LTV to use the processed data as the obtained historical LTV of the user for the target business, and preprocessing the initial behavioral data to use the processed data as the user's behavioral data for the target business within the historical period, wherein the preprocessing includes at least one of deduplication processing, normalization processing, and outlier processing.

[0095] Optionally, the prediction module 402 is specifically used to input the historical LTV and the behavior data into the prediction model to be trained, so as to extract user features from the historical LTV and the behavior data through the feature extraction layer included in the prediction model; and input the user features into the first prediction layer included in the prediction model, so as to predict the first behavior features of the user at a specified future time based on the user features through the first prediction layer.

[0096] Optionally, the prediction module 402 is specifically used to input the historical LTV and the behavior data into the prediction model to be trained, so as to extract user features from the historical LTV and the behavior data through the feature extraction layer included in the prediction model; and input the user features into the second prediction layer included in the prediction model, so as to predict the second behavior features of the user at the specified future time based on the user features through the second prediction layer.

[0097] Optionally, the prediction module 402 is specifically used to input the user characteristics, the first behavior characteristics and the second behavior characteristics into a regression layer included in the prediction model, so as to determine the LTV of the user at the specified future time through the regression layer as the predicted LTV.

[0098] Optionally, the prediction module 402 is further configured to input the behavior data into a behavior feature layer in the prediction model to obtain a behavior feature corresponding to the user; input the behavior feature into a prediction layer in the prediction model to predict the behavior data of the user at the specified future time as predicted behavior data;

[0099] The training module 403 is specifically used to determine the loss value based on the deviation between the predicted LTV and the label LTV corresponding to the user at the specified future time, and the deviation between the predicted behavior data and the actual behavior data of the user at the specified future time, and train the prediction model based on the loss value, wherein the deviation between the predicted behavior data and the actual behavior data of the user at the specified future time is positively correlated with the loss value.

[0100] Figure 5 A schematic diagram of a service execution device provided in this specification includes:

[0101] Acquisition module 501: used to acquire the historical LTV of a user for a target service and the behavior data of the user for the target service in a historical period corresponding to the historical LTV;

[0102] Prediction module 502: used to input the historical LTV and the behavior data into a pre-trained prediction model, so that the prediction model determines the predicted LTV of the user at a specified time in the future, and the prediction model is trained by the above-mentioned model training method;

[0103] Execution module 503: used to determine the service strategy for the user according to the predicted LTV, and execute the target service according to the service strategy.

[0104] This specification also provides a computer-readable storage medium, which stores a computer program, which can be used to execute the above Figure 1 The model training method shown or Figure 3 The business execution method shown.

[0105] This manual also provides Figure 6 The one shown corresponds to Figure 1 or Figure 3 A schematic diagram of the electronic device. Figure 6 As shown, at the hardware level, the electronic device includes a processor, an internal bus, a network interface, a memory, and a non-volatile memory, and may also include other hardware required for the business. The processor reads the corresponding computer program from the non-volatile memory into the memory and then runs it to achieve the above Figure 1 The model training method shown or Figure 3 The business execution method shown.

[0106] Of course, in addition to software implementation, this specification does not exclude other implementation methods, such as logic devices or a combination of software and hardware, etc., that is to say, the executor of the following processing flow is not limited to each logic unit, but can also be hardware or logic devices.

[0107] In the 1990s, improvements to a technology could be clearly distinguished as hardware improvements (for example, improvements to the circuit structure of diodes, transistors, switches, etc.) or software improvements (improvements to the method flow). However, with the development of technology, many improvements to the method flow today can be regarded as direct improvements to the hardware circuit structure. Designers almost always obtain the corresponding hardware circuit structure by programming the improved method flow into the hardware circuit. Therefore, it cannot be said that an improvement in a method flow cannot be implemented using a hardware entity module. For example, a programmable logic device (PLD) (such as a field programmable gate array (FPGA)) is such an integrated circuit whose logical function is determined by the user's programming of the device. Designers can "integrate" a digital system on a PLD by programming it themselves, without having to ask a chip manufacturer to design and produce a dedicated integrated circuit chip. Moreover, nowadays, instead of manually making integrated circuit chips, this kind of programming is mostly implemented by "logic compiler" software, which is similar to the software compiler used when developing and writing programs, and the original code before compilation must also be written in a specific programming language, which is called hardware description language (HDL). There is not only one HDL, but many kinds, such as ABEL (Advanced Boolean Expression Language), AHDL (Altera Hardware Description Language), Confluence, CUPL (Cornell University Programming Language), HDCal, JHDL (Java Hardware Description Language), Lava, Lola, MyHDL, PALASM, RHDL (Ruby Hardware Description Language), etc. The most commonly used ones are VHDL (Very-High-Speed ​​Integrated Circuit Hardware Description Language) and Verilog. Those skilled in the art should also know that it is only necessary to program the method flow slightly in the above-mentioned hardware description languages ​​and program it into the integrated circuit, and then it is easy to obtain the hardware circuit that implements the logic method flow.

[0108] The controller can be implemented in any appropriate manner, for example, the controller can take the form of a microprocessor or processor and a computer-readable medium storing a computer-readable program code (such as software or firmware) that can be executed by the (micro)processor, a logic gate, a switch, an application-specific integrated circuit (ASIC), a programmable logic controller, and an embedded microcontroller. Examples of controllers include, but are not limited to, the following microcontrollers: ARC 625D, Atmel AT91SAM, Microchip PIC18F26K20, and Silicone Labs C8051F320. The memory controller can also be implemented as part of the control logic of the memory. Those skilled in the art also know that in addition to implementing the controller in a purely computer-readable program code manner, the controller can be implemented in the form of a logic gate, a switch, an application-specific integrated circuit, a programmable logic controller, and an embedded microcontroller by logically programming the method steps. Therefore, this controller can be considered as a hardware component, and the devices included therein for implementing various functions can also be regarded as structures within the hardware component. Or even, the devices for implementing various functions can be regarded as both software modules for implementing the method and structures within the hardware component.

[0109] The systems, devices, modules or units described in the above embodiments may be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, the computer may be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

[0110] For the convenience of description, the above device is described by dividing it into various units according to its functions. Of course, when implementing this specification, the functions of each unit can be implemented in the same or multiple software and / or hardware.

[0111] Those skilled in the art will appreciate that the embodiments of this specification may be provided as methods, systems, or computer program products. Therefore, this specification may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, this specification may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0112] This specification is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of this specification. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0113] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0114] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0115] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0116] The memory may include non-permanent storage in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0117] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0118] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0119] Those skilled in the art will appreciate that the embodiments of this specification may be provided as methods, systems or computer program products. Therefore, this specification may take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware. Moreover, this specification may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0120] This specification may be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. This specification may also be practiced in distributed computing environments where tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules may be located in local and remote computer storage media, including storage devices.

[0121] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0122] The above description is only an embodiment of the present specification and is not intended to limit the present specification. For those skilled in the art, the present specification may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present specification shall be included in the scope of the claims of the present specification.

Claims

1. A model training method, characterized in that: include: Obtain the user's historical life cycle total value LTV for the target business and the user's behavior data for the target business in the historical period corresponding to the historical LTV; Input the historical LTV and the behavior data into the prediction model to be trained, so as to predict the first behavior feature of the user at a specified future time according to the historical LTV and the behavior data through a first prediction layer included in the prediction model, and predict the second behavior feature of the user at the specified future time according to the historical LTV and the behavior data through a second prediction layer included in the prediction model, and input the first behavior feature and the second behavior feature into the regression layer included in the prediction model, so as to determine the LTV of the user at the specified future time through the regression layer as the predicted LTV, the first behavior feature is used to characterize whether the user performs a specified business operation for the target business at the specified future time, the second behavior feature is used to characterize whether the user continues to perform the target business at the specified future time, and the specified business operation is a business operation that prompts the user to continue to perform the target business; A loss value is determined based on a deviation between the predicted LTV and the tag LTV corresponding to the user at the specified future time, and the prediction model is trained based on the loss value, wherein there is a positive correlation between the deviation and the loss value.

2. The method according to claim 1, characterized in that The step of obtaining the historical life cycle total value LTV of the user for the target business and the behavior data of the user for the target business in the historical period corresponding to the historical LTV specifically includes: Obtaining the user's initial historical LTV for the target business and the user's initial behavior data for the target business during the historical period; The initial historical LTV is preprocessed to use the processed data as the acquired historical LTV of the user for the target business, and the initial behavior data is preprocessed to use the processed data as the behavior data of the user for the target business during the historical period, wherein the preprocessing includes at least one of deduplication processing, normalization processing, and outlier processing.

3. The method according to claim 1, characterized in that The step of inputting the historical LTV and the behavior data into the prediction model to be trained, so as to predict the first behavior feature of the user at a specified time in the future according to the historical LTV and the behavior data through the first prediction layer included in the prediction model, specifically includes: Inputting the historical LTV and the behavior data into a prediction model to be trained, so as to extract user features from the historical LTV and the behavior data through a feature extraction layer included in the prediction model; The user features are input into a first prediction layer included in the prediction model, so as to predict a first behavior feature of the user at a specified future time according to the user features through the first prediction layer.

4. The method according to claim 1, characterized in that The step of inputting the historical LTV and the behavior data into the prediction model to be trained, and predicting the second behavior feature of the user at the specified future time according to the historical LTV and the behavior data through the second prediction layer included in the prediction model, specifically includes: Inputting the historical LTV and the behavior data into a prediction model to be trained, so as to extract user features from the historical LTV and the behavior data through a feature extraction layer included in the prediction model; The user features are input into a second prediction layer included in the prediction model, so as to predict the second behavior features of the user at the specified future time according to the user features through the second prediction layer.

5. The method according to claim 3 or 4, characterized in that The step of inputting the first behavior feature and the second behavior feature into the regression layer included in the prediction model to determine the LTV of the user at the specified future time through the regression layer as a step of predicting the LTV specifically includes: The user feature, the first behavior feature, and the second behavior feature are input into a regression layer included in the prediction model, so as to determine the LTV of the user at the future specified time through the regression layer as the predicted LTV.

6. The method according to claim 1, characterized in that Before the step of determining a loss value according to a deviation between the predicted LTV and the tag LTV corresponding to the user at the specified future time, and training the prediction model according to the loss value, the method further includes: Inputting the behavior data into the behavior feature layer in the prediction model to obtain the behavior features corresponding to the user; Inputting the behavior feature into the prediction layer in the prediction model to predict the behavior data of the user at the specified future time as predicted behavior data; The step of determining a loss value according to a deviation between the predicted LTV and the tag LTV corresponding to the user at the specified future time, and training the prediction model according to the loss value specifically includes: A loss value is determined based on the deviation between the predicted LTV and the label LTV corresponding to the user at the specified future time, and the deviation between the predicted behavior data and the actual behavior data of the user at the specified future time, and the prediction model is trained based on the loss value, wherein the deviation between the predicted behavior data and the actual behavior data of the user at the specified future time is positively correlated with the loss value.

7. A service execution method, characterized in that: include: Obtain the user's historical LTV for the target business and the user's behavior data for the target business in the historical period corresponding to the historical LTV; Inputting the historical LTV and the behavior data into a pre-trained prediction model so that the prediction model determines the predicted LTV of the user at a specified time in the future, wherein the prediction model is trained by the method according to any one of claims 1 to 6; A service strategy for the user is determined based on the predicted LTV, and the target service is executed according to the service strategy.

8. A model training device, characterized in that: include: Acquisition module: used to acquire the historical life cycle total value LTV of the user for the target business and the behavior data of the user for the target business in the historical period corresponding to the historical LTV; Prediction module: used for inputting the historical LTV and the behavior data into the prediction model to be trained, so as to predict the first behavior feature of the user at a specified future time according to the historical LTV and the behavior data through a first prediction layer included in the prediction model, and predict the second behavior feature of the user at the specified future time according to the historical LTV and the behavior data through a second prediction layer included in the prediction model, and inputting the first behavior feature and the second behavior feature into the regression layer included in the prediction model, so as to determine the LTV of the user at the specified future time through the regression layer as the predicted LTV, the first behavior feature is used to characterize whether the user performs a specified business operation for the target business at the specified future time, the second behavior feature is used to characterize whether the user continues to perform the target business at the specified future time, and the specified business operation is a business operation that prompts the user to continue to perform the target business; Training module: used to determine a loss value according to the deviation between the predicted LTV and the second historical LTV, and train the target model according to the loss value, wherein there is a positive correlation between the deviation and the loss value.

9. A computer-readable storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the method according to any one of claims 1 to 7 is implemented.

Citation Information

Patent Citations

  • User life value cycle detection method and device and computer equipment

    CN112258238A

  • Model training and information recommendation method and device

    CN112966186A

  • Prediction method and device based on data mining, equipment and storage medium

    CN114742568A

  • Object type prediction method and device, computer equipment and storage medium

    CN115249081A

  • Marketing scene prediction model training method and device, electronic equipment and storage medium

    CN118070916A