Data prediction method and device based on multi-stage model, equipment and medium

Abstract

The invention discloses a data prediction method and device based on a multi-stage model, equipment and a medium, and the method comprises the steps: obtaining business data, carrying out the feature extraction of the business data, and obtaining a shared feature; respectively inputting the shared features into at least two stage prediction models which are connected in sequence, so that the later stage prediction model performs stage prediction based on the shared features and the output features of the previous stage prediction model; and taking the prediction result of each stage prediction model as the prediction result corresponding to each stage, and predicting the conversion condition of the user in each stage through the business data, so that the subsequent customer activity promotion behavior is only performed on the silent customer with the conversion possibility, thereby improving the pertinence and effectiveness of customer activity promotion.

Description

Technical Field

[0001] The present disclosure generally relates to the field of data prediction technology, and specifically relates to a data prediction method, device, equipment and medium based on a multi-stage model. Background Art

[0002] In the online financial credit industry, the main ways to acquire customers include advertising and marketing and awakening silent customers. As the credit industry market matures, the importance of the silent customer awakening (also known as customer activation) customer acquisition model has gradually become more prominent.

[0003] Silent customers can be roughly divided into two categories: one is historically registered users but new customers who have not applied, and the other is old customers who have completed the transaction through email and meet the definition of institutional churn. Customer activation is usually achieved by directly reaching all users through outbound calls, SMS, and mini programs. However, such operations not only waste the cost of reaching out, but also easily cause negative emotions among customers who have no conversion needs. Summary of the invention

[0004] In view of the above-mentioned defects or deficiencies in the prior art, it is desired to provide a data prediction method, apparatus, device and medium based on a multi-stage model, which predicts the conversion status of users at each stage through business data, so that subsequent customer activation actions are only performed on silent customers with the possibility of conversion, thereby improving the pertinence and effectiveness of customer activation.

[0005] In a first aspect, the embodiments of the present application provide a data prediction method, apparatus, device and medium based on a multi-stage model, including:

[0006] Acquire business data, perform feature extraction on the business data, and obtain shared features;

[0007] Inputting the shared features into at least two stage prediction models connected in sequence, respectively, so that the subsequent stage prediction model performs stage prediction based on the shared features and the output features of the previous stage prediction model;

[0008] The prediction results of the prediction models in each stage are used as the prediction results corresponding to each stage.

[0009] In some embodiments, the business data includes multiple types of business data, and the feature extraction of the business data to obtain the shared features includes:

[0010] Obtain a feature extraction model corresponding to each type of business data;

[0011] Inputting the multiple types of business data into the corresponding feature extraction model respectively to obtain business data features corresponding to each type of business data;

[0012] The business data features are aligned and then spliced ​​to obtain the shared features.

[0013] In some embodiments, the multiple types of business data include unstructured sequence feature data, and the method further includes:

[0014] Get the word table corresponding to the sequence features;

[0015] The sequence feature data corresponding to each behavior is generated based on the time sequence of the user behavior and the word table.

[0016] In some embodiments, the at least two stage prediction models are connected in sequence according to a stage dependency relationship.

[0017] In some embodiments, it also includes:

[0018] During the training process, loss weights are added to the at least two stage prediction models based on the priorities of the at least two stage prediction models.

[0019] In some embodiments, the business data used for training includes stage label data corresponding to the at least two stage prediction models, and the stage label data uses the conversion results of the corresponding stages as corresponding labels.

[0020] In a second aspect, an embodiment of the present application provides a data prediction device based on a multi-stage model, comprising:

[0021] An acquisition module, used to acquire business data, perform feature extraction on the business data, and obtain shared features;

[0022] A prediction module, used for inputting the shared features into at least two stage prediction models connected in sequence, respectively, so that the subsequent stage prediction model performs stage prediction based on the shared features and the output features of the previous stage prediction model;

[0023] The result module is used to use the prediction results of the prediction models of each stage as the prediction results corresponding to each stage.

[0024] In a third aspect, an embodiment of the present application 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 method described in the embodiment of the present application when executing the program.

[0025] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described in the embodiment of the present application.

[0026] In a fifth aspect, an embodiment of the present application provides a computer program product, including a computer program, characterized in that when the computer program is executed by a processor, it implements the method described in the embodiment of the present application.

[0027] The present application proposes a data prediction method, device, equipment and medium based on a multi-stage model, which can use shared features to make predictions in each stage of prediction, effectively improve the utilization rate of feature data in the prediction process of each stage, thereby improving the prediction accuracy, and use the analysis results of the previous stage to make predictions in the staged prediction process of each credit business stage, thereby further improving the accuracy of the current stage prediction.

[0028] Additional aspects and advantages of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Other features, objects and advantages of the present application will become more apparent by reading the detailed description of non-limiting embodiments made with reference to the following drawings:

[0030] Figure 1 A schematic diagram of a process of a data prediction method based on a multi-stage model provided in an embodiment of the present application is shown;

[0031] Figure 2 A schematic diagram of the structure of a multi-stage model provided in an embodiment of the present application is shown;

[0032] Figure 3 A schematic diagram of the structure of a data prediction device based on a multi-stage model provided in an embodiment of the present application is shown;

[0033] Figure 4 A schematic diagram of the structure of a computer system of an electronic device or server suitable for implementing an embodiment of the present application is shown. DETAILED DESCRIPTION

[0034] The present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It is to be understood that the specific embodiments described herein are only used to explain the relevant invention, rather than to limit the invention. It is also necessary to explain that, for ease of description, only the parts related to the invention are shown in the accompanying drawings.

[0035] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0036] With the continuous development of online financial credit business, relevant technologies for user analysis using deep learning models have been developed. For example, the activation marketing model based on traditional machine learning algorithms can achieve user differentiation prediction by statistical analysis of historical data. However, since the traditional machine learning algorithm model is relatively simple and mainly based on statistical features, that is, the model input is in the form of tabular data, the model fitting ability is general, and the prediction accuracy of whether the customer responds to the conversion is low. Moreover, the screening of features using traditional machine learning algorithm models relies more on expert experience, and can only extract a small number of features with high differentiation ability from a large amount of raw data, and a large number of features with slightly lower differentiation ability are discarded, resulting in low utilization of data features and low prediction accuracy.

[0037] Based on this, the present application proposes a data prediction method, device, equipment and medium based on a multi-stage model, which can effectively improve the accuracy of the prediction.

[0038] In order to further illustrate the technical solution provided by the embodiment of the present application, this is described in detail below in conjunction with the accompanying drawings and specific implementation methods. Although the embodiment of the present application provides the method operation instruction steps shown in the following embodiments or drawings, more or less operation instruction steps may be included in the method based on routine or no creative labor. In the steps where there is no necessary causal relationship logically, the execution order of these steps is not limited to the execution order provided by the embodiment of the present application. The method may be executed in the order of the method shown in the embodiment or drawings or in parallel during the actual processing process or when the device is executed.

[0039] It should be noted that the acquisition or use of data in the embodiments of the present application requires the user's consent. The relevant data can only be obtained after the user's authorization and permission, and the acquisition or use of the data complies with relevant legal provisions.

[0040] Please refer to Figure 1 , Figure 1 FIG. 1 is a flow chart of a data prediction method based on a multi-stage model provided in an embodiment of the present application. Figure 1 As shown, the method includes:

[0041] Step 101, obtain business data, extract features from the business data, and obtain shared features.

[0042] It should be noted that business data refers to user historical behavior data, including the user's online credit business behavior, including but not limited to registration, application, credit use, etc., and also includes the user's behavior on the credit app, including but not limited to app installation, opening, uninstallation, etc.

[0043] Furthermore, the business data includes multiple types of business data, and feature extraction is performed on the business data to obtain shared features, specifically including: obtaining a feature extraction model corresponding to each type of business data, inputting the multiple types of business data into the corresponding feature extraction models respectively, to obtain business data features corresponding to each type of business data; and then aligning and splicing the features of each business data to obtain shared features.

[0044] That is to say, the embodiment of the present application provides different types of feature extraction models for different types of business data, and extracts features from business data through matching feature extraction models, thereby improving the accuracy and reliability of feature extraction of business data. Moreover, the use of multiple types of business data effectively enriches the types of business data, which can effectively improve the fitting ability of the post-prediction model and improve the prediction accuracy.

[0045] Specifically, in the embodiment of the present application, business data includes structured data and unstructured data. Among them, structured data is data that can be attributed by tables, or numerical and categorical data, such as user historical behavior data, etc. Unstructured data is character data, such as sequence feature data of users installing, opening, and uninstalling credit apps generated based on word tables.

[0046] Among them, the word table is an app word table. Specifically, credit apps related to credit demand and qualification assessment are screened to generate an app word table. Then, the apps included in the word table are screened according to the time sequence of user installation, opening and uninstallation operations to form respective input lists, that is, app installation sequence features, app opening sequence features and app uninstallation sequence features are generated. Finally, the sequence features are truncated or padded according to the maximum length of the sequence feature data to obtain the final app installation sequence feature data, app opening sequence feature data and app uninstallation sequence feature data.

[0047] In a possible embodiment, if Figure 2 As shown, the embodiment of the present application utilizes a multi-layer perceptron (MLP) to perform feature extraction on structured data to obtain business data features corresponding to the structured data, and utilizes an encoder model to perform feature extraction on unstructured data to obtain business data features corresponding to the unstructured data, and then aligns and concatenates the business data features corresponding to the structured data and the business data features corresponding to the unstructured data to obtain shared features.

[0048] Preferably, the structured data includes numerical data and categorical data. The numerical data and categorical data are first input into the embedding layer to be converted into data feature vectors and then input into a multi-layer perceptron (MLP) for feature extraction to obtain business data features corresponding to the structured data. The unstructured data is feature extracted using a Transformer-Encoder model to obtain business data features corresponding to the unstructured data. In some embodiments, in order to further improve the accuracy of feature extraction, different Transformer-Encoder models can be trained for different user operations on the credit app, such as a Transformer-Encoder model corresponding to the app installation sequence feature data, a Transformer-Encoder model corresponding to the app opening sequence feature data, and a Transformer-Encoder model corresponding to the app uninstall sequence feature data.

[0049] Step 102, respectively input the shared features into at least two stage prediction models connected in sequence, so that the subsequent stage prediction model performs stage prediction based on the shared features and the output features of the previous stage prediction model.

[0050] It should be noted that the embodiment of the present application inputs the shared features into at least two stage prediction models connected in sequence, so that each stage prediction model can perform prediction analysis based on the shared features, effectively reducing the gradient vanishing problem caused by feature transfer encountered in the later stage prediction model.

[0051] The present application further connects at least two stage prediction models in sequence to effectively enrich the types of input feature data of the prediction model in the later stage, that is, increases the output features of the prediction model in the previous stage, thereby effectively improving the accuracy and reliability of the prediction results of the prediction model in the later stage.

[0052] In a feasible embodiment, at least two stage prediction models are connected in sequence according to a stage dependency relationship.

[0053] It should be understood that in the online credit business industry, there is a sequential relationship between some key nodes. For example, application behavior requires registration behavior as a prerequisite, credit granting behavior requires application behavior as a prerequisite, and credit use behavior requires credit granting behavior as a prerequisite. In other words, only after the user has registered can he apply, that is, it is necessary to predict the user's application behavior. If the user is not registered, similarly, only after the user has applied can it be predicted whether the user will be granted credit. Only after the user is granted credit can it be further predicted whether the user will use the credit.

[0054] Based on this, at least two stage prediction models are connected in sequence according to the stage dependency relationship, so that the stage prediction model that needs to rely on the results of the previous behavior can make stage predictions for the current stage based on the prediction results of the previous stage prediction model, thereby improving the accuracy of the current stage prediction for users who have implemented or are predicted to implement the previous stage behavior, and reducing the current stage prediction model's predictions for users who have not implemented the previous stage behavior or are predicted not to implement the previous stage behavior, so as to reduce the amount of prediction data in the current stage and in subsequent stages.

[0055] In a feasible embodiment, the output features of the prediction model in the previous stage include but are not limited to the prediction results corresponding to the previous stage and the output features of the prediction model in the previous stage before classification, that is, including but not limited to feature information used for predicting classification and predicting classification results.

[0056] In a feasible embodiment, the stage prediction model is a binary classification model, such as a multi-layer perceptron MLP, etc. At least two stage prediction models may be the same or different, so long as they can predict the corresponding task stages. This application does not make any specific limitations.

[0057] It should also be understood that, for the first stage prediction model in at least two stage prediction models, the input features do not include the output features of the previous stage prediction model. Therefore, the first stage prediction model predicts the first stage user behavior based on the shared features, and inputs the prediction results and features into the next stage prediction model.

[0058] Step 103: The prediction results of the prediction models at each stage are used as the prediction results corresponding to each stage.

[0059] That is to say, in addition to inputting the prediction results into the prediction model of the next stage, the prediction model of each stage also directly outputs the prediction results of the current stage, so as to send recapture contact information to users with positive prediction results (prediction results are implementation behavior) based on the prediction results, thereby improving the activation rate of silent customers. At the same time, no recapture contact information is sent to users with negative prediction results (prediction results are non-implementation of behavior), thereby reducing the frequency of contact with these silent customers, avoiding unnecessary disturbance to silent users who have no needs, and improving user experience.

[0060] Therefore, the present application proposes a data prediction method based on a multi-stage model, which can use shared features for prediction in each stage, effectively improve the utilization rate of feature data in the prediction process of each stage, thereby improving the prediction accuracy, and use the analysis results of the previous stage for prediction in the staged prediction process of each credit business stage, thereby further improving the accuracy of the current stage prediction.

[0061] In a feasible embodiment, during the training process, the user's historical activation marketing response data is first obtained, and then the response data is subjected to feature derivation and processing.

[0062] It should be noted that in the embodiment of the present application, due to the extremely low user response reach rate in the silent customer activation scenario, in order to avoid model fitting anomalies, the embodiment of the present application configures the unconverted (negative) sample size according to the successfully converted (positive) sample size. Exemplarily, all positive samples are obtained, and then based on the ratio of 10% of positive samples, negative samples are extracted from the full amount of negative samples so that the negative sample ratio is 90%. After obtaining the target number of negative samples, the sample data is cleaned and filled, and then feature derivation is performed according to expert experience to generate the final sample set.

[0063] In a feasible embodiment, positive samples can be determined in stages. For example, customers logging into the credit app within N days of marketing promotion are regarded as positive samples in the login stage, and vice versa as negative samples; logged-in users applying within N days of marketing promotion are regarded as positive samples in the application stage, and vice versa as negative samples; credit approval of applied users within N days of marketing promotion is regarded as positive samples in the credit stage, and vice versa as negative samples; credit users using credit within N days of marketing promotion are regarded as positive samples in the credit use stage, and vice versa as negative samples.

[0064] Therefore, the present application can configure positive samples in stages, can effectively increase the number of positive samples, and can improve the pertinence of samples in each stage, avoiding the negative sample labels of untrusted users from affecting the prediction results in the application stage, thereby improving the activation rate in each stage.

[0065] In a feasible embodiment, during the model training process, a binary classification function may be used as a loss function, such as BceLoss (Binary Cross Entropy Loss).

[0066] Optionally, training can be performed in sequence according to the model execution order, for example, the feature extraction model is trained first, and after the feature extraction model converges, the first-stage prediction model is trained, and after the first-stage prediction model converges, the second-stage prediction model is trained, and so on, to ensure the accuracy and reliability of the first execution model before training the subsequent model, thereby avoiding the prior model and the subsequent model forming an overall training result, which leads to a decrease in the prediction accuracy output by the prior model and the subsequent model respectively.

[0067] Optionally, the entire model can be trained as a whole. In this case, the loss weights corresponding to the model at each stage can be set according to the importance of each stage, and the parameters can be updated based on the gradient descent algorithm until the final model converges.

[0068] In a feasible embodiment, the trained model can be deployed across platforms in the form of a Docker container to improve the generalization ability of the model.

[0069] In summary, the present application proposes a data prediction method based on a multi-stage model, which can utilize shared features for prediction in each stage, effectively improve the utilization rate of feature data in the prediction process of each stage, thereby improving the prediction accuracy, and utilize the analysis results of the previous stage for prediction in the staged prediction process of each credit business stage, thereby further improving the accuracy of the current stage prediction.

[0070] It should be noted that although the operations of the method of the present invention are described in a particular order in the drawings, this does not require or imply that the operations must be performed in this particular order or that all illustrated operations must be performed to achieve desired results.

[0071] Figure 3 A schematic diagram of the structure of a data prediction device based on a multi-stage model provided in one embodiment of the present application is shown.

[0072] like Figure 3 As shown, the data prediction device 10 based on the multi-stage model includes:

[0073] An acquisition module 11 is used to acquire business data, perform feature extraction on the business data, and obtain shared features;

[0074] A prediction module 12, configured to input the shared features into at least two sequentially connected stage prediction models, respectively, so that the subsequent stage prediction model performs stage prediction based on the shared features and the output features of the previous stage prediction model;

[0075] The result module 13 is used to use the prediction results of the prediction models of each stage as the prediction results corresponding to each stage.

[0076] In some embodiments, the business data includes multiple types of business data, and the acquisition module 11 is further used to:

[0077] Obtain a feature extraction model corresponding to each type of business data;

[0078] Inputting the multiple types of business data into the corresponding feature extraction model respectively to obtain business data features corresponding to each type of business data;

[0079] The business data features are aligned and then spliced ​​to obtain the shared features.

[0080] In some embodiments, the acquisition module 11 is further used to:

[0081] Get the word table corresponding to the sequence features;

[0082] The sequence feature data corresponding to each behavior is generated based on the time sequence of the user behavior and the word table.

[0083] In some embodiments, the at least two stage prediction models are connected in sequence according to a stage dependency relationship.

[0084] In some embodiments, the data prediction device 10 based on the multi-stage model further includes a training module for:

[0085] During the training process, loss weights are added to the at least two stage prediction models based on the priorities of the at least two stage prediction models.

[0086] In some embodiments, the business data used for training includes stage label data corresponding to the at least two stage prediction models, and the stage label data uses the conversion results of the corresponding stages as corresponding labels.

[0087] It should be understood that the modules or modules described in the data prediction device 10 based on the multi-stage model are similar to those described in the reference Figure 2 The various steps in the described method correspond to each other. Therefore, the operations and features described above for the method are also applicable to the data prediction device 10 based on the multi-stage model and the modules contained therein, and will not be repeated here. The data prediction device 10 based on the multi-stage model can be pre-implemented in the browser or other security applications of the electronic device, and can also be loaded into the browser or its security application of the electronic device by downloading or the like. The corresponding modules in the data prediction device 10 based on the multi-stage model can cooperate with the modules in the electronic device to implement the solution of the embodiment of the present application.

[0088] For the several modules or units mentioned in the above detailed description, this division is not mandatory. In fact, according to the embodiments of the present disclosure, the features and functions of two or more modules or units described above can be embodied in one module or unit. On the contrary, the features and functions of one module or unit described above can be further divided into multiple modules or units to be embodied.

[0089] Reference below Figure 4 , Figure 4 A schematic diagram of the structure of a computer system of an electronic device or server suitable for implementing an embodiment of the present application is shown.

[0090] like Figure 4As shown, the computer system includes a central processing unit (CPU) 401, which can perform various appropriate actions and processes according to the program stored in the read-only memory (ROM) 402 or the program loaded from the storage part 408 into the random access memory (RAM) 403. In the RAM 403, various programs and data required for the operation instructions of the system are also stored. The CPU 401, the ROM 402, and the RAM 403 are connected to each other through a bus 404. An input / output (I / O) interface 405 is also connected to the bus 404.

[0091] The following components are connected to the I / O interface 405: an input section 406 including a keyboard, a mouse, etc.; an output section 407 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 408 including a hard disk, etc.; and a communication section 409 including a network interface card such as a LAN card, a modem, etc. The communication section 409 performs communication processing via a network such as the Internet. A drive 410 is also connected to the I / O interface 405 as needed. A removable medium 411, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 410 as needed, so that a computer program read therefrom is installed into the storage section 408 as needed.

[0092] In particular, according to an embodiment of the present application, the above reference flow chart Figure 2 The described process can be implemented as a computer software program. For example, an embodiment of the present application includes a computer program product, which includes a computer program carried on a computer readable medium, and the computer program includes a program code for executing the method shown in the flow chart. In such an embodiment, the computer program includes a program code for executing the method shown in the flow chart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 409, and / or installed from the removable medium 411. When the computer program is executed by the central processing unit (CPU) 401, the above-mentioned functions defined in the system of the present application are executed.

[0093] It should be noted that the computer-readable medium shown in the present application may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present application, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, device or device. In the present application, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable program code. This propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer-readable signal media may also be any computer-readable medium such as a computer-readable storage medium that can send, propagate or transmit a program for use by or in conjunction with an instruction execution system, device or device. The program code contained on the computer-readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wire, optical cable, RF, etc., or any suitable combination of the above.

[0094] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, functions and operating instructions of the system, method and computer program product according to various embodiments of the present application. In this regard, each box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the aforementioned module, program segment or a part of a code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some implementations as replacements, the functions marked in the box can also occur in a sequence different from that marked in the accompanying drawings. For example, the boxes represented by two connections can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flow chart, and the combination of the boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operating instruction, or can be implemented with a combination of dedicated hardware and computer instructions.

[0095] The units or modules involved in the embodiments described in the present application may be implemented by software or by hardware. The described units or modules may also be arranged in a processor. For example, they may be described as: a processor includes an acquisition module, a prediction module, and a result module. Among them, the names of these units or modules do not constitute a limitation on the units or modules themselves under certain circumstances. For example, the acquisition module may also be described as "acquiring business data, performing feature extraction on the business data, and obtaining shared features".

[0096] As another aspect, the present application further provides a computer-readable storage medium, which may be included in the electronic device described in the above embodiment, or may exist independently without being assembled into the electronic device. The above computer-readable storage medium stores one or more programs, and when the above programs are used by one or more processors to execute the data prediction method based on the multi-stage model described in the present application.

[0097] The above description is only a preferred embodiment of the present application and an explanation of the technical principles used. Those skilled in the art should understand that the scope of disclosure involved in the present application is not limited to the technical solution formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the aforementioned disclosed concept. For example, the above features are replaced with the technical features with similar functions disclosed in this application (but not limited to) by each other to form a technical solution.

Claims

1. A data prediction method based on a multi-stage model, characterized in that: include: Acquire business data, perform feature extraction on the business data, and obtain shared features; Inputting the shared features into at least two stage prediction models connected in sequence, respectively, so that the subsequent stage prediction model performs stage prediction based on the shared features and the output features of the previous stage prediction model; The prediction results of the prediction models in each stage are used as the prediction results corresponding to each stage.

2. The data prediction method based on the multi-stage model according to claim 1, characterized in that: The business data includes multiple types of business data, and the feature extraction of the business data to obtain shared features includes: Obtain a feature extraction model corresponding to each type of business data; Inputting the multiple types of business data into the corresponding feature extraction model respectively to obtain business data features corresponding to each type of business data; The business data features are aligned and then spliced ​​to obtain the shared features.

3. The data prediction method based on the multi-stage model according to claim 2 is characterized in that: The multiple types of business data include unstructured sequence feature data, and the method further includes: Get the word table corresponding to the sequence features; The sequence feature data corresponding to each behavior is generated based on the time sequence of the user behavior and the word table.

4. The data prediction method based on the multi-stage model according to claim 1, characterized in that: The at least two stage prediction models are connected in sequence according to the stage dependency relationship.

5. The data prediction method based on the multi-stage model according to claim 1, characterized in that: Also includes: During the training process, loss weights are added to the at least two stage prediction models based on the priorities of the at least two stage prediction models.

6. The data prediction method based on the multi-stage model according to claim 5, characterized in that: The business data used for training includes stage label data corresponding to the at least two stage prediction models, and the stage label data uses the conversion results of the corresponding stages as corresponding labels.

7. A data prediction device based on a multi-stage model, characterized in that: include: An acquisition module, used to acquire business data, perform feature extraction on the business data, and obtain shared features; A prediction module, used for inputting the shared features into at least two stage prediction models connected in sequence, respectively, so that the subsequent stage prediction model performs stage prediction based on the shared features and the output features of the previous stage prediction model; The result module is used to use the prediction results of the prediction models of each stage as the prediction results corresponding to each stage.

8. 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 multi-stage model-based data prediction method as described in any one of claims 1 to 6 is implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the data prediction method based on a multi-stage model as described in any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the data prediction method based on a multi-stage model described in any one of claims 1 to 6 is implemented.

Images