Life cycle value prediction method, device and non-volatile storage medium

Abstract

The present invention discloses a method, device and non-volatile storage medium for predicting life cycle value. The method comprises: obtaining data sets corresponding to target users in multiple business scenarios, wherein the data sets include user information and business scenario information; encoding data sets corresponding to multiple business scenarios to obtain encoded data sets corresponding to multiple business scenarios; inputting the encoded data sets corresponding to multiple business scenarios into the target scenario shared network layer in the target neural network model, and outputting processed data corresponding to multiple business scenarios; inputting the processed data corresponding to multiple business scenarios into the corresponding scenario-specific network layer in the target neural network model, respectively, to obtain the prediction result of the target user. The present invention solves the technical problem that there are multiple business scenarios in the field of life cycle value prediction, and a single model often finds it difficult to capture the complex relationship across scenarios, resulting in insufficient prediction accuracy.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a life cycle value prediction method, device and non-volatile storage medium. Background Art

[0002] In the field of LTV (lifetime value) prediction, most of the predictions are for a single business scenario, and there is no LTV prediction task for multiple business scenarios. In multiple business scenarios, a single model often cannot capture the complex relationships across scenarios, resulting in insufficient prediction accuracy.

[0003] To address the above-mentioned problems, no effective solution has been proposed yet. Summary of the invention

[0004] The embodiments of the present invention provide a method, device and non-volatile storage medium for life cycle value prediction, so as to at least solve the technical problem that there are multiple business scenarios in the field of life cycle value prediction, and a single model often finds it difficult to capture the complex relationships across scenarios, resulting in insufficient prediction accuracy.

[0005] According to one aspect of an embodiment of the present invention, a method for predicting lifetime value is provided, comprising: obtaining data sets corresponding to target users in multiple business scenarios, wherein the data sets include user information and business scenario information; encoding the data sets corresponding to the multiple business scenarios to obtain encoded data sets corresponding to the multiple business scenarios; inputting the encoded data sets corresponding to the multiple business scenarios into a target scenario shared network layer in a target neural network model, and outputting processed data corresponding to the multiple business scenarios; respectively inputting the processed data corresponding to the multiple business scenarios into corresponding scenario-specific network layers in the target neural network model to obtain prediction results for the target users, wherein the prediction results represent the lifetime value of the target users, and the target neural network model includes scenario-specific network layers corresponding to the multiple business scenarios.

[0006] Optionally, before inputting the encoded data sets corresponding to each of the multiple business scenarios into the target scenario shared network layer in the target neural network model, the method includes: obtaining historical data sets corresponding to each of the multiple business scenarios; dividing the historical data sets corresponding to each of the multiple business scenarios into a training set and a test set; using the training set to train a preset initial neural network model to obtain a trained neural network model, wherein the initial neural network model includes an initial scenario shared network layer and initial scenario-specific network layers corresponding to each of the multiple business scenarios; and using the test set to adjust the trained neural network model to obtain a target neural network model.

[0007] Optionally, a training set is used to train a preset initial neural network model to obtain a trained neural network model, wherein the training set includes training sets corresponding to multiple business scenarios, including: inputting the training sets corresponding to each of the multiple business scenarios into the initial scenario shared network layer, and training the initial scenario shared network layer to obtain the trained scene shared network layer and outputting processed data sets corresponding to each of the multiple business scenarios; inputting the processed data sets corresponding to each of the multiple business scenarios into the corresponding initial scenario-specific network layer, training the initial scenario-specific network layer corresponding to each of the multiple business scenarios, and obtaining the trained scene-specific network layer corresponding to each of the multiple business scenarios; determining the trained neural network model based on the trained scene shared network layer and the trained scene-specific network layer corresponding to each of the multiple business scenarios.

[0008] Optionally, the processed data sets corresponding to each of the multiple business scenarios are respectively input into the corresponding initial scenario-specific network layer, and the initial scenario-specific network layer corresponding to each of the multiple business scenarios is trained to obtain the trained scenario-specific network layer corresponding to each of the multiple business scenarios, including: setting a loss function in the initial scenario-specific network layer corresponding to each of the multiple business scenarios; inputting the processed data sets corresponding to each of the multiple business scenarios into the corresponding initial scenario-specific network layer to obtain the loss value corresponding to the scenario-specific network layer corresponding to each of the multiple business scenarios; determining the sum of the loss values ​​of the initial scenario-specific network layer corresponding to each of the multiple business scenarios based on preset weights; adjusting the parameters corresponding to the initial scenario-specific network layer corresponding to each of the multiple business scenarios based on the sum of the loss values ​​of the initial scenario-specific network layer corresponding to each of the multiple business scenarios to obtain the trained scenario-specific network layer corresponding to each of the multiple business scenarios.

[0009] Optionally, before determining the sum of the loss values ​​of the initial scenario-specific network layers corresponding to multiple business scenarios based on preset weights, it includes: obtaining the priorities corresponding to each of the multiple business scenarios; and setting weights based on the priorities corresponding to each of the multiple business scenarios.

[0010] Optionally, the target neural network model includes multiple layers of neural network layers, wherein any one of the multiple layers of neural network layers includes a target scenario shared network layer and target scenario-specific network layers corresponding to each of the multiple business scenarios.

[0011] According to another aspect of an embodiment of the present invention, a lifetime value prediction device is also provided, including: an acquisition module, used to obtain data sets corresponding to target users in multiple business scenarios, wherein the data sets include user information and business scenario information; an encoding module, used to encode data sets corresponding to multiple business scenarios, and obtain encoded data sets corresponding to multiple business scenarios; a first input module, used to input the encoded data sets corresponding to multiple business scenarios into a target scenario shared network layer in a target neural network model, and output processed data corresponding to multiple business scenarios; a second input module, used to input the processed data corresponding to multiple business scenarios into corresponding scenario-specific network layers in the target neural network model, and obtain prediction results for the target user, wherein the prediction results represent the lifetime value of the target user, and the target neural network model includes scenario-specific network layers corresponding to multiple business scenarios.

[0012] According to another aspect of an embodiment of the present invention, a non-volatile storage medium is provided, the non-volatile storage medium including a stored program, wherein when the program is running, the device where the non-volatile storage medium is located is controlled to execute any one of the above-mentioned life cycle value prediction methods.

[0013] According to another aspect of the embodiments of the present invention, a computer device is provided. The computer device includes a processor, and the processor is used to run a program. When the program is run, any one of the above-mentioned life cycle value prediction methods is executed.

[0014] According to another aspect of the embodiments of the present invention, there is further provided a computer program product, including a computer program, which implements any one of the above-mentioned life cycle value prediction methods when executed by a processor.

[0015] In an embodiment of the present invention, a life cycle value prediction method is adopted, by obtaining data sets corresponding to target users in multiple business scenarios, wherein the data sets include user information and business scenario information; encoding the data sets corresponding to the multiple business scenarios to obtain encoded data sets corresponding to the multiple business scenarios; inputting the encoded data sets corresponding to the multiple business scenarios into the target scenario shared network layer in the target neural network model, and outputting processed data corresponding to the multiple business scenarios; inputting the processed data corresponding to the multiple business scenarios into the corresponding scenario-specific network layers in the target neural network model, respectively, to obtain a prediction result for the target user, wherein the prediction result represents the life cycle value of the target user, and the target neural network model includes scenario-specific network layers corresponding to the multiple business scenarios, thereby achieving the purpose of setting a combined prediction of a scenario shared layer and a scenario-specific layer, thereby realizing the technical effect of improving prediction accuracy, and further solving the technical problem that there are multiple business scenarios in the field of life cycle value prediction, and a single model is often difficult to capture the complex relationship across scenarios, resulting in insufficient prediction accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0017] Figure 1 A hardware structure block diagram of a computer terminal for implementing a method for predicting life cycle value is shown;

[0018] Figure 2 is a flow chart of a method for predicting life cycle value according to an embodiment of the present invention;

[0019] Figure 3 is a training flow chart of a target neural network model in a method for predicting life cycle value according to an optional embodiment of the present invention;

[0020] Figure 4 It is a structural block diagram of a life cycle value prediction device provided according to an embodiment of the present invention. DETAILED DESCRIPTION

[0021] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0022] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0023] According to an embodiment of the present invention, a method embodiment of a method for predicting life cycle value is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0024] The method embodiment provided in the first embodiment of the present application can be executed in a mobile terminal, a computer terminal or a similar computing device. Figure 1 FIG. 1 shows a hardware structure block diagram of a computer terminal for implementing a method for predicting life cycle value. Figure 1 As shown, the computer terminal 10 may include one or more (102a, 102b, ..., 102n are used to illustrate) processors (the processor may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA), a memory 104 for storing data. In addition, it may also include: a display, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of the BUS bus), a network interface, a power supply and / or a camera. It can be understood by those skilled in the art that Figure 1 The structure shown is only for illustration and does not limit the structure of the above electronic device. Figure 1 More or fewer components as shown, or with Figure 1 Different configurations shown.

[0025] It should be noted that the one or more processors and / or other data processing circuits described above may generally be referred to herein as "data processing circuits". The data processing circuits may be embodied in whole or in part as software, hardware, firmware, or any other combination thereof. In addition, the data processing circuit may be a single independent processing module, or may be incorporated in whole or in part into any of the other components in the computer terminal 10. As described in the embodiments of the present application, the data processing circuit acts as a processor control (e.g., selection of a variable resistor terminal path connected to an interface).

[0026] The memory 104 can be used to store software programs and modules of application software, such as the program instructions / data storage device corresponding to the life cycle value prediction method in the embodiment of the present invention. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory 104, that is, realizing the life cycle value prediction method of the above-mentioned application program. The memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory 104 may further include a memory remotely arranged relative to the processor, and these remote memories may be connected to the computer terminal 10 via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0027] The display may be, for example, a touch screen liquid crystal display (LCD) that enables a user to interact with a user interface of the computer terminal 10 .

[0028] Figure 2 FIG. 1 is a flow chart of a method for predicting life cycle value according to an embodiment of the present invention. Figure 2 As shown, the method comprises the following steps:

[0029] Step S202: Obtain data sets corresponding to target users in multiple business scenarios, wherein the data sets include user information and business scenario information.

[0030] In this step, data sets of target users in multiple business scenarios are obtained, where business scenarios can be e-commerce scenarios, lending scenarios, etc. The data of each scenario includes basic user information and data specific to the business scenario. Among them, the user information usually includes the user's age, gender, geographic location, etc. This information is the key to understanding the user's basic attributes and behavioral habits. Business scenario information is for each specific business scenario, and specific data closely related to the scenario needs to be collected. For example, in an e-commerce scenario, it is necessary to collect the user's browsed item sequence, the user's clicked item sequence, the user's transaction item sequence, the user's transaction amount, etc.; in a lending scenario, it is necessary to collect the user's credit expression data, loan record data, and repayment behavior data, etc. These data can provide deep insights specific to the scenario and help the model capture behavioral patterns and trends directly related to the business scenario.

[0031] Step S204: Encode the data sets corresponding to the multiple business scenarios to obtain encoded data sets corresponding to the multiple business scenarios.

[0032] In this step, the data sets corresponding to multiple business scenarios can be encoded, for example, embedding encoding can be performed to convert them into a format suitable for neural network processing. This is to convert the original data into a format that is more suitable for neural network processing. Embedding encoding is a technology for converting data in machine learning and deep learning. It is mainly used to convert discrete, high-dimensional features (such as words in text, or user and product IDs) into continuous, low-dimensional vector representations. This encoding method captures the relationship and semantic information between features through vectors in mathematical space, so that the model can better understand and process the data.

[0033] Step S206, input the encoded data sets corresponding to the multiple business scenarios into the target scenario shared network layer in the target neural network model, and output the processed data corresponding to the multiple business scenarios.

[0034] In this step, the coded data sets corresponding to each of the multiple business scenarios are input into the target scenario shared network layer of the target neural network model, which includes the target scenario shared network layer and the target scenario exclusive network layer. The coded data of all business scenarios are input into the scenario shared network layer. This network layer is designed to capture and learn common features across scenarios, such as users' basic behavior patterns and preference trends. These features are not specific to a single scenario, but are helpful for predictions of multiple scenarios. The scenario shared network layer processes the input data through a multi-layer neural network structure. Each layer may contain different neural network modules (such as convolutional layers, fully connected layers, recurrent neural networks, etc.) to extract and fuse features in the data. These layers may use conditional gating mechanisms to dynamically determine which features are passed to subsequent scenario exclusive layers and which features are shared. After the scenario shared network layer is processed, the output is a feature representation that incorporates cross-scenario information, which is further passed to the exclusive network layer of each scenario. The dedicated network layer for each scene will perform deeper processing and analysis based on the features output by the scene’s shared network layer and the characteristics of the scene itself to optimize the prediction task for the specific scene.

[0035] The encoded data is input into the scene sharing network layer, which consists of multiple layers of neural networks. The conditional gating mechanism is used to selectively extract and transmit information between the multiple layers. When the data is transmitted in the scene sharing network layer, it needs to be processed by the conditional gating mechanism for selective routing and filtering. The conditional gating mechanism is implemented by a learnable weight learning network. The conditional gating mechanism is used to dynamically adjust the flow of information between different levels. The conditional gating mechanism is implemented by a learnable weight learning network. Through the conditional gating mechanism, the model can selectively activate or inhibit certain information paths according to the specific characteristics and context of the current data. This not only improves the flexibility and accuracy of information processing, but also allows the model to effectively integrate knowledge from different scenarios while maintaining professionalism.

[0036] Step S208, input the processed data corresponding to each of the multiple business scenarios into the corresponding scenario-specific network layer in the target neural network model to obtain the prediction result of the target user, wherein the prediction result represents the life cycle value of the target user, and the target neural network model includes the scenario-specific network layer corresponding to each of the multiple business scenarios.

[0037] In this step, after the scene-sharing network layer is processed, the processed data corresponding to each of the output multiple business scenarios are further passed to the dedicated network layer of each scene. The dedicated network layer of each scene will perform deeper processing and analysis based on the characteristics of the scene-sharing network layer output and the characteristics of the scene itself to optimize the prediction task of the specific scene, and finally obtain the prediction result of the target user. In the scene-specific layer, the data of each business scenario will be further optimized, for example, through specially designed neural network modules, such as attention mechanisms, specific activation functions or regularization techniques, to enhance the understanding of scene-specific information. This layer converts the output data of the scene-sharing layer into a representation that is more specialized in the scene. After the scene-specific layer processes the data, it will make a prediction and calculate the difference between the prediction and the actual result, that is, the loss. Different scenarios may use different loss functions, such as the ZILN loss function. These loss values ​​will be fed back to the model to adjust the weights of the network layer and optimize the performance of the model. Through repeated iterations, the neural network model continuously adjusts its weights to optimize the performance of the scene-sharing network layer and the scene-specific layer until the preset training termination condition is reached. After training, the model can be used to predict the life cycle value of new users in multiple business scenarios.

[0038] When data is transmitted from the low-level network to the high-level network, it will be filtered by the weight learning network (gated network) before being passed upward. Specifically, the gating network is divided into a scene-sharing gating network and a scene-specific gating network. The scene-sharing gating network receives data from the low-level scene-sharing network and all scene-specific networks, and sends it to the high-level scene-sharing network and all scene-specific networks; the scene-specific network only receives data from the low-level scene-sharing network and the corresponding low-level scene-specific network, and only sends it to the high-level scene-sharing network and the corresponding high-level scene transmission network. This multi-level shared layer design allows the model to better extract features and integrate information when processing large amounts of heterogeneous data, greatly improving the model's adaptability and prediction performance to new scenarios. In addition, by fine-tuning the interaction between the shared and exclusive layers, the model can improve the utilization of information and the generalization ability of the model without sacrificing scene specificity.

[0039] Among them, the data processed by the shared layer is transmitted to the corresponding scene-specific layer respectively. In each exclusive layer, the loss is calculated according to the scene characteristics. In this step, the data flows from the multi-level shared layer to the respective scene-specific layers. These layers are designed for each specific business scenario to ensure that the model can be finely optimized for the unique needs of each scenario. The data received by each scene-specific layer comes from the shared layer at the top of the multi-level shared layer and the corresponding scene-specific layer. These data have been preliminarily screened and weighted through the conditional gating mechanism to ensure that the incoming data contains the most relevant information in the corresponding scene.

[0040] Through the above process, the scenario-shared network layer and the scenario-specific network layer work together to enable the model to learn and extract useful information from multiple business scenarios, optimize the accuracy of LTV prediction, and maintain a full understanding of the characteristics of each scenario. This design method fully utilizes the cross-scenario commonality and scenario-specificity of the data, thereby improving the generalization ability and prediction performance of the model.

[0041] Through the above steps, the purpose of setting up combined prediction of scenario sharing layer and scenario exclusive layer is achieved, thereby realizing the technical effect of improving prediction accuracy, and further solving the technical problem that in the field of life cycle value prediction, there are multiple business scenarios, and a single model is often difficult to capture the complex relationship across scenarios, resulting in insufficient prediction accuracy.

[0042] As an optional embodiment, before inputting the encoded data sets corresponding to each of the multiple business scenarios into the target scenario shared network layer in the target neural network model, it includes: obtaining the historical data sets corresponding to each of the multiple business scenarios; dividing the historical data sets corresponding to each of the multiple business scenarios into a training set and a test set; using the training set to train a preset initial neural network model to obtain a trained neural network model, wherein the initial neural network model includes an initial scenario shared network layer and an initial scenario-specific network layer corresponding to each of the multiple business scenarios; using the test set to adjust the trained neural network model to obtain a target neural network model.

[0043] Optionally, historical data can be collected from each business scenario, which may include user behavior, transaction records, etc. The acquisition of historical data sets is based on actual business operations and data analysis requirements, and is extracted from the databases of various business systems. The historical data sets of each scenario are divided into training sets and test sets. Usually, the data sets are randomly split according to a certain ratio (such as 70% training and 30% testing) to ensure that the data distribution in the training set and the test set is similar. This step is very important for the evaluation and verification of the model. The training set is used to learn the model parameters, while the test set is used to independently verify the performance of the model. The preset initial neural network model is trained using the training set data. The initial model contains an initial scenario shared network layer and an initial scenario-specific network layer corresponding to each of the multiple business scenarios. During the training process, the model learns the patterns in the data and adjusts the weights and parameters through forward propagation and back propagation algorithms to minimize the prediction loss on the training set. When training the initial model, the training set data of each business scenario is encoded separately and then input into the scenario shared network layer and the corresponding scenario-specific network layer. Through the training of the neural network, the model learns the shared features of the scene and the specific features of each scene, optimizing the prediction ability for each task. After the training process is completed, the trained neural network model can be validated using the test set. This involves inputting the test set data into the model to evaluate the model's performance on unseen data. If the model's performance does not meet expectations, it may be necessary to adjust the model structure, parameters, or training process, such as adjusting the learning rate, increasing or decreasing the number of network layers, changing the loss function or optimizer, etc., to improve the model's prediction accuracy. In the validation phase, if the model performs poorly on the test set, hyperparameter tuning may be required to find a better model configuration. Hyperparameters are parameters that need to be manually set before model training, such as the number of network layers, the number of neurons, the learning rate, the regularization coefficient, etc., which have a significant impact on the performance of the model. Through continuous training, validation, and tuning, a neural network model that performs best on the test set of all business scenarios is determined. This model is the target neural network model, which includes optimized scenario-shared network layers and scenario-specific network layers, and can process data across scenarios and make accurate LTV predictions.

[0044] This series of steps ensures that the model can fully learn the characteristics of all business scenarios during the training phase and verify the generalization ability of the model during the testing phase, ultimately obtaining a neural network model that can effectively predict the user lifetime value in multiple scenarios.

[0045] As an optional embodiment, a training set is used to train a preset initial neural network model to obtain a trained neural network model, wherein the training set includes training sets corresponding to multiple business scenarios, including: inputting the training sets corresponding to each of the multiple business scenarios into the initial scenario shared network layer, and training the initial scenario shared network layer to obtain the trained scene shared network layer and outputting processed data sets corresponding to each of the multiple business scenarios; inputting the processed data sets corresponding to each of the multiple business scenarios into the corresponding initial scenario-specific network layer, training the initial scenario-specific network layer corresponding to each of the multiple business scenarios, and obtaining the trained scene-specific network layer corresponding to each of the multiple business scenarios; determining the trained neural network model based on the trained scene shared network layer and the trained scene-specific network layer corresponding to each of the multiple business scenarios.

[0046] Optionally, the training set is used to train the preset initial neural network model, and then the trained neural network model is obtained. The training set data corresponding to each of the multiple business scenarios can be first input into the initial scenario shared network layer. This network layer is responsible for identifying and extracting common features across scenarios, such as the user's basic behavior pattern, credit score, etc., which are meaningful to all business scenarios. The initial scenario shared network layer is trained by the back propagation algorithm, and the parameters of the shared layer are adjusted to optimize the extraction of common features. This step is the core of multi-scenario learning, which helps the model share knowledge between different scenarios and improves the overall learning efficiency and prediction performance. Then the data set processed by the scenario shared network layer is input into the initial scenario exclusive network layer corresponding to each business scenario. Each exclusive network layer focuses on learning the features and patterns of its specific scenario. The initial scenario exclusive network layer corresponding to each of the multiple business scenarios is trained, and the parameters of the exclusive network layer are adjusted through iterative optimization to improve the prediction accuracy in specific scenarios, such as purchase prediction in e-commerce scenarios, default risk assessment in lending scenarios, etc. When both the scene-sharing network layer and the scene-specific network layer are trained, the trained scene-sharing network layer and the trained scene-specific network layers corresponding to multiple business scenarios are combined to build the entire trained neural network model. The trained model integrates the optimization results of the scene-sharing layer and the scene-specific layer to achieve efficient use of cross-scene features and deep learning of scene-specific knowledge, thereby achieving higher prediction accuracy and generalization ability in multi-scenario LTV prediction tasks.

[0047] Through this series of steps, the initial neural network model evolves into an advanced model that can effectively process and predict the user lifetime value in multiple scenarios by learning data from different business scenarios. This model can not only capture common information across scenarios, but also achieve deep optimization in each specific business scenario, providing strong technical support for accurate LTV prediction.

[0048] As an optional embodiment, the processed data sets corresponding to each of the multiple business scenarios are respectively input into the corresponding initial scenario-specific network layer, and the initial scenario-specific network layer corresponding to each of the multiple business scenarios is trained to obtain the trained scenario-specific network layer corresponding to each of the multiple business scenarios, including: setting a loss function in the initial scenario-specific network layer corresponding to each of the multiple business scenarios; inputting the processed data sets corresponding to each of the multiple business scenarios into the corresponding initial scenario-specific network layer to obtain the loss value corresponding to the scenario-specific network layer corresponding to each of the multiple business scenarios; determining the sum of the loss values ​​of the initial scenario-specific network layer corresponding to each of the multiple business scenarios based on preset weights; adjusting the parameters corresponding to the initial scenario-specific network layer corresponding to each of the multiple business scenarios based on the sum of the loss values ​​of the initial scenario-specific network layer corresponding to each of the multiple business scenarios to obtain the trained scenario-specific network layer corresponding to each of the multiple business scenarios.

[0049] Optionally, for the initial scenario-specific network layer in each business scenario, a suitable loss function needs to be set to measure the gap between the model prediction value and the actual value. The choice of loss function should be based on the specific prediction task and data characteristics of each scenario. For example, an e-commerce scenario may use a binary cross entropy loss (for predicting whether a user will purchase) plus a lognormal regression loss (for predicting the user's lifetime value LTV); while a lending scenario may use a similar loss function with default prediction. The data set processed by the scenario shared network layer is input into the initial scenario-specific network layer corresponding to each business scenario. The dedicated network layer will make predictions based on the input data and output the user's behavior or lifetime value prediction results in that scenario. For each business scenario, the model's prediction results are compared with the actual data, and the loss value is calculated according to the preset loss function. The loss function gives a quantitative measure of the model error, which will be used as the goal of optimizing the model parameters. Based on the preset weights, the sum of the loss values ​​of the initial scenario-specific network layers corresponding to multiple business scenarios is determined. The setting of weights usually reflects the relative importance of each scenario in the business or the relative amount of data, allowing the model to pay more attention to certain scenarios during training to achieve global optimization. Based on the total loss value of multiple scenarios (i.e., weighted sum loss), the parameters of each scenario-specific network layer are adjusted through the back-propagation algorithm. The optimization algorithm (such as gradient descent, Adam, etc.) will update the weights and biases of the network layer according to the gradient of the loss value to minimize the loss function. The above process is repeated in each training iteration until the model converges or the preset training stop condition is reached. The convergence of the model usually means that the total loss value no longer decreases significantly, which indicates that the model parameters have been adjusted to a state that can better fit the training data. After the training is completed, the initial scenario-specific network layer parameters of each business scenario will be optimized and converted into a trained scenario-specific network layer. These network layers can more accurately predict the user lifetime value (LTV) or other related tasks in the corresponding business scenario, showing higher prediction accuracy and business adaptability.

[0050] In the multi-scenario LTV prediction model, by setting a loss function for each scenario-specific network layer and performing weighted summation, it is possible to ensure that the model can comprehensively consider the importance of all scenarios during training, while allowing deep optimization of specific scenarios. This training method fully utilizes the advantages of multi-task learning, promotes knowledge sharing between different business scenarios and improves the generalization ability of the model, laying a solid foundation for the application of the model in actual business environments.

[0051] As an optional embodiment, before determining the sum of the loss values ​​of the initial scenario-specific network layers corresponding to multiple business scenarios based on preset weights, it includes: obtaining the priorities corresponding to each of the multiple business scenarios; and setting weights based on the priorities corresponding to each of the multiple business scenarios.

[0052] Optionally, the priorities corresponding to multiple business scenarios can be set or directly obtained according to actual conditions. For example, for a company, the e-commerce scenario may directly contribute to its main source of income, while the lending scenario may involve higher financial risks and customer credit assessment. The priority of each scenario can be determined based on its contribution to the value of the enterprise. According to the priority, the weights in the training process are set for each business scenario. Scenarios with higher weights mean that their prediction results have a greater impact on the final performance of the model, and the model will focus more on optimizing the prediction accuracy of these scenarios. The weights can be set manually or dynamically adjusted through some automatic tuning mechanism to reflect changes in business goals and fluctuations in the importance of scenarios. During the model training phase, the loss function of each business scenario is multiplied by its corresponding weight, and then weighted summed to obtain the total loss of multiple scenarios. This weighted summation ensures that the model not only focuses on the accurate prediction of all scenarios during the training process, but also pays more attention to scenarios with higher weights, which helps the model to prioritize the key business needs of the enterprise in practical applications. The model will update its parameters based on the weighted total loss. The optimization algorithm (such as Adam, SGD, etc.) will adjust the weights and biases of the network layers according to the gradient of the loss of each scenario and the set weights to minimize the total loss function, thereby improving the predictive performance and business adaptability of the overall model.

[0053] Through the above steps, the model can not only learn and predict the user lifetime value in multiple business scenarios, but also automatically adjust its prediction focus according to the priority of the business scenario, thereby improving the practical value and effect of the model in the actual business environment.

[0054] As an optional embodiment, the target neural network model includes multiple neural network layers, wherein any one of the multiple neural network layers includes a target scenario shared network layer and target scenario-specific network layers corresponding to multiple business scenarios.

[0055] Optionally, the target neural network model may include multiple neural network layers, wherein each layer may include a target scenario shared network layer and a target scenario exclusive network layer corresponding to each of the multiple business scenarios. The levels of the multi-layer neural network layers may be different, and data will be transmitted from the low-level neural network to the high-level neural network. This multi-level shared layer design allows the model to better extract features and integrate information when processing large amounts of heterogeneous data, greatly improving the model's adaptability and predictive performance to new scenarios. In addition, by fine-tuning the interaction between shared and exclusive layers, the model can improve information utilization and the generalization ability of the model without sacrificing scenario specificity.

[0056] The following is a specific example of training a target neural network model. Figure 3is a training flow chart of a target neural network model in a method for predicting life cycle value according to an optional embodiment of the present invention, such as Figure 3 As shown:

[0057] S1. Multi-scenario data preparation: Prepare data sets for different business scenarios (such as e-commerce and lending). The data for each scenario includes basic user information and data specific to the business scenario. This step is the basis for model building and involves detailed preparation of data sets for different business scenarios. The data set needs to contain two parts of information: basic user information and business scenario-specific data.

[0058] S2, Data Encoding: Embedding the data of all scenarios and converting it into a format suitable for neural network processing. This step involves embedding the data collected from various business scenarios in order to convert the original data into a format more suitable for neural network processing. Embedding encoding is an effective data preprocessing technology.

[0059] S3, multi-level shared layer processing: The encoded data is input into the shared layer, which is composed of multiple layers of neural networks. The network is divided into a scene-sharing network and a scene-exclusive network. The multiple layers selectively extract and transmit information through a conditional gating mechanism (Conditional Gating), and the conditional gating mechanism is implemented by a learnable weight learning network. In this step, the encoded data is input into a complex multi-layer neural network structure, which is one of the core technologies of the present invention. The shared layer has multiple layers, each of which is composed of multiple sub-networks, including a scene-sharing network and a scene-exclusive network; when data is transmitted in the shared layer, it needs to be selected by the conditional gating mechanism for routing and filtering.

[0060] S4, Scenario-specific layer processing: The data processed by the multi-level shared layers are transmitted to the corresponding scenario-specific layers. In each dedicated layer, the loss is calculated according to the scenario characteristics. In this step, the data flows from the multi-level shared layers to the respective scenario-specific layers, which are designed for each specific business scenario to ensure that the model can be finely optimized for the unique needs of each scenario.

[0061] S5. Loss calculation: The loss of each scenario is calculated using the ZILN loss function, and the total loss is the weighted sum of the losses of each scenario. By setting different weights for different scenarios, the total loss of multiple scenarios is calculated by weighted summation. In this step, the weighted loss function is used to weight the ZILN losses of different scenarios to obtain the total loss of multiple scenarios, where the weights are adjusted according to the importance of the business, which can further improve the business adaptability and effectiveness of the model.

[0062] S6. Model training and prediction: Finally, the model training is completed through the optimization algorithm based on the model framework to ensure the best scenario adaptability and prediction accuracy. The model training and verification are completed through the selection of optimization algorithms and parameter tuning; after that, the model will be deployed in actual business scenarios for LTV prediction.

[0063] It should be noted that, for the above-mentioned method embodiments, for the sake of simplicity, they are all described as a series of action combinations, but those skilled in the art should know that the present invention is not limited by the described action sequence, because according to the present invention, certain steps can be performed in other sequences or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present invention.

[0064] Through the description of the above implementation methods, those skilled in the art can clearly understand that the life cycle value prediction method according to the above embodiment can be implemented by means of software plus a necessary general hardware platform, and of course it can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. The computer software product is stored in a storage medium (such as ROM / RAM, a disk, or an optical disk), and includes a number of instructions for a terminal device (which can be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0065] According to an embodiment of the present invention, a life cycle value prediction device for implementing the above life cycle value prediction method is also provided. Figure 4 is a structural block diagram of a life cycle value prediction device provided according to an embodiment of the present invention. Figure 4 As shown, the life cycle value prediction device includes: an acquisition module 42, an encoding module 44, a first input module 46 and a second input module 48. The life cycle value prediction device is described below.

[0066] The acquisition module 42 is used to acquire data sets corresponding to the target user in multiple business scenarios, wherein the data sets include user information and business scenario information.

[0067] The encoding module 44 is connected to the acquisition module 42, and is used to encode the data sets corresponding to the multiple business scenarios to obtain the encoded data sets corresponding to the multiple business scenarios.

[0068] The first input module 46 is connected to the encoding module 44, and is used to input the encoded data sets corresponding to the multiple business scenarios into the target scenario shared network layer in the target neural network model, and output the processed data corresponding to the multiple business scenarios.

[0069] The second input module 48 is connected to the first input module 46, and is used to input the processed data corresponding to each of the multiple business scenarios into the corresponding scenario-specific network layers in the target neural network model to obtain the prediction results of the target users, wherein the prediction results represent the life cycle value of the target users, and the target neural network model includes scenario-specific network layers corresponding to each of the multiple business scenarios.

[0070] It should be noted that the acquisition module 42, encoding module 44, first input module 46 and second input module 48 correspond to steps S202 to S208 in the embodiment, and the examples and application scenarios implemented by the multiple modules and the corresponding steps are the same, but are not limited to the contents disclosed in the above embodiment. It should be noted that the above modules, as part of the device, can be run in the computer terminal 10 provided in the embodiment.

[0071] An embodiment of the present invention may provide a computer device. Optionally, in this embodiment, the computer device may be located in at least one network device among multiple network devices of a computer network. The computer device includes a memory and a processor.

[0072] Among them, the memory can be used to store software programs and modules, such as the program instructions / modules corresponding to the life cycle value prediction method and device in the embodiment of the present invention. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, that is, realizing the above-mentioned life cycle value prediction method. The memory may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory may further include a memory remotely arranged relative to the processor, and these remote memories can be connected to the computer terminal via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0073] The processor can call the information and application programs stored in the memory through the transmission device to perform the following steps: obtain the data sets corresponding to the target users in multiple business scenarios, wherein the data sets include user information and business scenario information; encode the data sets corresponding to the multiple business scenarios to obtain the encoded data sets corresponding to the multiple business scenarios; input the encoded data sets corresponding to the multiple business scenarios into the target scenario shared network layer in the target neural network model, and output the processed data corresponding to the multiple business scenarios; input the processed data corresponding to the multiple business scenarios into the corresponding scenario-specific network layers in the target neural network model to obtain the prediction results of the target users, wherein the prediction results represent the life cycle value of the target users, and the target neural network model includes scenario-specific network layers corresponding to the multiple business scenarios.

[0074] An embodiment of the present invention is used to provide a method for predicting life cycle value, by obtaining data sets corresponding to target users in multiple business scenarios, wherein the data sets include user information and business scenario information; encoding the data sets corresponding to the multiple business scenarios to obtain encoded data sets corresponding to the multiple business scenarios; inputting the encoded data sets corresponding to the multiple business scenarios into the target scenario shared network layer in the target neural network model, and outputting processed data corresponding to the multiple business scenarios; inputting the processed data corresponding to the multiple business scenarios into the corresponding scenario-specific network layers in the target neural network model, respectively, to obtain a prediction result for the target user, wherein the prediction result represents the life cycle value of the target user, and the target neural network model includes scenario-specific network layers corresponding to the multiple business scenarios, thereby achieving the purpose of setting a combined prediction of a scenario shared layer and a scenario-specific layer, thereby realizing the technical effect of improving prediction accuracy, and further solving the technical problem that there are multiple business scenarios in the field of life cycle value prediction, and a single model is often difficult to capture the complex relationship across scenarios, resulting in insufficient prediction accuracy.

[0075] A person of ordinary skill in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing the hardware related to the terminal device through a program, and the program can be stored in a non-volatile storage medium, and the storage medium may include: a flash drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.

[0076] The embodiment of the present invention further provides a non-volatile storage medium. Optionally, in this embodiment, the non-volatile storage medium can be used to store the program code executed by the life cycle value prediction method provided by the above embodiment.

[0077] Optionally, in this embodiment, the non-volatile storage medium may be located in any computer terminal in a computer terminal group in a computer network, or in any mobile terminal in a mobile terminal group.

[0078] Optionally, in this embodiment, the non-volatile storage medium is configured to store program code for executing the following steps: obtaining data sets corresponding to each target user in multiple business scenarios, wherein the data sets include user information and business scenario information; encoding the data sets corresponding to each of the multiple business scenarios to obtain encoded data sets corresponding to each of the multiple business scenarios; inputting the encoded data sets corresponding to each of the multiple business scenarios into the target scenario shared network layer in the target neural network model, and outputting processed data corresponding to each of the multiple business scenarios; inputting the processed data corresponding to each of the multiple business scenarios into the corresponding scenario-specific network layers in the target neural network model, respectively, to obtain a prediction result for the target user, wherein the prediction result represents the life cycle value of the target user, and the target neural network model includes a scenario-specific network layer corresponding to each of the multiple business scenarios.

[0079] An embodiment of the present invention also provides a computer program product, including a computer program. Optionally, in this embodiment, when the computer program is executed by a processor, it can achieve: obtaining data sets corresponding to target users in multiple business scenarios, wherein the data sets include user information and business scenario information; encoding the data sets corresponding to each of the multiple business scenarios to obtain encoded data sets corresponding to each of the multiple business scenarios; inputting the encoded data sets corresponding to each of the multiple business scenarios into the target scenario shared network layer in the target neural network model, and outputting processed data corresponding to each of the multiple business scenarios; inputting the processed data corresponding to each of the multiple business scenarios into the corresponding scenario-specific network layers in the target neural network model, respectively, to obtain prediction results for the target users, wherein the prediction results represent the life cycle value of the target users, and the target neural network model includes scenario-specific network layers corresponding to each of the multiple business scenarios.

[0080] The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages or disadvantages of the embodiments.

[0081] In the above embodiments of the present invention, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0082] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of the units can be a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0083] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0084] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0085] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a non-volatile storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, server or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, magnetic disk or optical disk and other media that can store program codes.

[0086] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for predicting life cycle value, characterized in that: include: Acquire data sets corresponding to target users in multiple business scenarios, wherein the data sets include user information and business scenario information; Encoding the data sets corresponding to the multiple business scenarios to obtain the encoded data sets corresponding to the multiple business scenarios; Inputting the encoded data sets corresponding to each of the multiple business scenarios into the target scenario shared network layer in the target neural network model, and outputting the processed data corresponding to each of the multiple business scenarios; The processed data corresponding to each of the multiple business scenarios are respectively input into the corresponding scenario-specific network layer in the target neural network model to obtain a prediction result for the target user, wherein the prediction result represents the life cycle value of the target user, and the target neural network model includes a scenario-specific network layer corresponding to each of the multiple business scenarios.

2. The method according to claim 1, characterized in that Before inputting the encoded data sets corresponding to the multiple business scenarios into the target scenario shared network layer in the target neural network model, the method includes: Obtain historical data sets corresponding to multiple business scenarios; Dividing the historical data sets corresponding to each of the multiple business scenarios into a training set and a test set; Using the training set to train a preset initial neural network model to obtain a trained neural network model, wherein the initial neural network model includes an initial scenario shared network layer and initial scenario-specific network layers corresponding to each of the multiple business scenarios; The trained neural network model is adjusted using the test set to obtain the target neural network model.

3. The method according to claim 2, characterized in that The training set is used to train the preset initial neural network model to obtain a trained neural network model, wherein the training set includes training sets corresponding to each of the multiple business scenarios, including: Inputting the training sets corresponding to the multiple business scenarios into the initial scene sharing network layer, and training the initial scene sharing network layer to obtain a trained scene sharing network layer and outputting processed data sets corresponding to the multiple business scenarios; Inputting the processed data sets corresponding to each of the multiple business scenarios into the corresponding initial scenario-specific network layer respectively, training the initial scenario-specific network layer corresponding to each of the multiple business scenarios, and obtaining the trained scenario-specific network layer corresponding to each of the multiple business scenarios; The trained neural network model is determined according to the trained scenario-shared network layer and the trained scenario-specific network layers corresponding to each of the multiple business scenarios.

4. The method according to claim 3, characterized in that The step of inputting the processed data sets corresponding to the multiple business scenarios into the corresponding initial scenario-specific network layers, training the initial scenario-specific network layers corresponding to the multiple business scenarios, and obtaining the trained scenario-specific network layers corresponding to the multiple business scenarios, comprises: Setting a loss function in an initial scenario-specific network layer corresponding to each of the multiple business scenarios; Inputting the processed data sets corresponding to each of the multiple business scenarios into the corresponding initial scenario-specific network layer respectively, and obtaining the loss values ​​corresponding to the scenario-specific network layer corresponding to each of the multiple business scenarios; Based on the preset weights, determine the sum of the loss values ​​of the initial scenario-specific network layers corresponding to each of the multiple business scenarios; Based on the sum of the loss values ​​of the initial scenario-specific network layers corresponding to each of the multiple business scenarios, the parameters corresponding to the initial scenario-specific network layers corresponding to each of the multiple business scenarios are adjusted to obtain the trained scenario-specific network layers corresponding to each of the multiple business scenarios.

5. The method according to claim 4, characterized in that Before determining the sum of the loss values ​​of the initial scenario-specific network layers corresponding to the multiple business scenarios based on the preset weights, the method includes: Obtaining the priorities corresponding to the multiple business scenarios; The weight is set based on the priorities corresponding to each of the multiple business scenarios.

6. The method according to any one of claims 1 to 5, characterized in that: The target neural network model includes multiple layers of neural network layers, wherein any one of the multiple layers of neural network layers includes the target scenario shared network layer and the target scenario-specific network layers corresponding to each of the multiple business scenarios.

7. A life cycle value prediction device, characterized in that: include: An acquisition module, used to acquire data sets corresponding to target users in multiple business scenarios, wherein the data sets include user information and business scenario information; An encoding module, used to encode the data sets corresponding to the multiple business scenarios to obtain the encoded data sets corresponding to the multiple business scenarios; A first input module, used to input the encoded data sets corresponding to each of the multiple business scenarios into the target scenario shared network layer in the target neural network model, and output the processed data corresponding to each of the multiple business scenarios; The second input module is used to input the processed data corresponding to each of the multiple business scenarios into the corresponding scenario-specific network layer in the target neural network model to obtain the prediction result of the target user, wherein the prediction result represents the life cycle value of the target user, and the target neural network model includes a scenario-specific network layer corresponding to each of the multiple business scenarios.

8. A non-volatile storage medium, characterized in that: The non-volatile storage medium includes a stored program, wherein when the program is running, the device where the non-volatile storage medium is located is controlled to execute the life cycle value prediction method described in any one of claims 1 to 6.

9. A computer device, characterized in that: include: Memory and processor, The memory stores a computer program; The processor is used to execute the computer program stored in the memory, and when the computer program is running, the processor executes the life cycle value prediction method according to any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the life cycle value prediction method described in any one of claims 1 to 6 is implemented.

Images