Marketing model training method and device, marketing method and device, equipment, medium and program product

Abstract

The embodiment of the invention discloses a marketing model training method and device, a marketing method and device, equipment, a medium and a program product. The marketing model training method comprises the steps of determining a causal relationship between a first historical marketing strategy and a first user behavior; and training a marketing model based on the causal relationship and the user characteristics, so that the trained marketing model can process the causal relationship corresponding to the to-be-recommended marketing strategy to obtain a to-be-reached target user corresponding to the to-be-recommended marketing strategy.

Description

Technical Field

[0001] The present application belongs to the field of artificial intelligence technology, and in particular relates to a marketing model training method, marketing method, device, equipment, medium and program product. Background Art

[0002] At present, marketing strategies in marketing activities often rely on historical data and simple statistical analysis. This method focuses on correlation analysis and has some limitations in practical applications, which restricts the accurate evaluation of the effectiveness of marketing activities and makes it difficult to cope with the rapid changes in the market and user behavior. Summary of the Invention

[0003] The embodiments of the present application provide a marketing model training method, marketing method, apparatus, device, medium and program product.

[0004] The present invention provides a marketing model training method, which includes:

[0005] Determine the causal relationship between the first historical marketing strategy and the first user behavior;

[0006] Based on the causal relationship and user characteristics, the marketing model is trained so that the trained marketing model can process the causal relationship corresponding to the recommended marketing strategy and obtain the target users to be reached corresponding to the recommended marketing strategy.

[0007] In some embodiments, before determining the causal relationship between the first historical marketing strategy and the first user behavior, the method further includes: constructing feature variables, processing variables, and result variables based on the second historical marketing strategy, the second user behavior, and the second user characteristics; constructing a first data set through the feature variables, the processing variables, and the result variables; performing model training based on the first data set to obtain a causal relationship evaluation model; determining the causal relationship between the first historical marketing strategy and the first user behavior includes: determining the causal relationship between the first historical marketing strategy and the first user behavior through the causal relationship evaluation model.

[0008] In some embodiments, before performing model training based on the first data set to obtain a causal relationship evaluation model, the method further includes: constructing a model to be trained based on a causal inference algorithm; the causal inference algorithm includes a meta-learner and / or a decision tree Uplift Tree; performing model training based on the first data set to obtain a causal relationship evaluation model includes: training the model to be trained based on the first data set to obtain a causal relationship evaluation model.

[0009] It can be seen that constructing a causal relationship evaluation model through a causal inference algorithm is conducive to quickly determining the causal relationship between marketing strategies and user behaviors through the causal relationship evaluation model, which is further conducive to improving the accuracy of identifying target users.

[0010] In some embodiments, the Meta-learner includes a T-learner algorithm and an X-learner algorithm. When the causal inference algorithm includes a Meta-learner, constructing the model to be trained based on the causal inference algorithm includes: constructing the model to be trained based on the X-learner algorithm; training the model to be trained based on the first data set includes: obtaining an estimated result based on the T-learner algorithm; the estimated result includes a causal relationship obtained based on the T-learner algorithm; and training the model to be trained based on the estimated result and the first data set.

[0011] It can be seen that combining the T-learner algorithm and the X-learner algorithm for model training is conducive to improving the processing capability of the causal relationship evaluation model and improving the accuracy of the causal relationship.

[0012] In some embodiments, before determining the causal relationship between the first historical marketing strategy and the first user behavior, the method further includes: dividing users based on user data to obtain user groups; the user data includes one or more of the user's transaction frequency, transaction amount, asset size, and risk preference; determining the causal relationship between the first historical marketing strategy and the first user behavior includes: determining the causal relationship between the first historical marketing strategy and the first user behavior in each user group.

[0013] In some embodiments, the trained marketing model can process the causal relationship corresponding to the marketing strategy to be recommended, and obtain the target users to be reached by the marketing strategy to be recommended in each user group.

[0014] The present application also provides a marketing method, which includes:

[0015] The causal relationship corresponding to the recommended marketing strategy is processed by the marketing model to obtain the target users to be reached corresponding to the recommended marketing strategy; the marketing model is trained based on the above-mentioned marketing model training method;

[0016] Based on the marketing strategy to be recommended, the target users are reached.

[0017] The present application also provides a marketing model training device, which includes:

[0018] a determination module, configured to determine a causal relationship between the first historical marketing strategy and the first user behavior;

[0019] The training module is used to train the marketing model based on the causal relationship and user characteristics, so that the trained marketing model can process the causal relationship corresponding to the recommended marketing strategy and obtain the target users to be reached corresponding to the recommended marketing strategy.

[0020] The present application also provides a marketing device, comprising:

[0021] A processing module, configured to process the causal relationship corresponding to the recommended marketing strategy using a marketing model to obtain target users to be reached corresponding to the recommended marketing strategy; the marketing model is trained based on the above-mentioned marketing model training method;

[0022] An interactive module is used to reach the target users based on the marketing strategy to be recommended.

[0023] An embodiment of the present application provides an electronic device, comprising a processor and a memory for storing a computer program that can be run on the processor; wherein,

[0024] The processor is used to run the computer program to execute any one of the above-mentioned marketing model training methods or marketing methods.

[0025] An embodiment of the present application provides a computer storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the above-mentioned marketing model training methods or marketing methods.

[0026] An embodiment of the present application provides a computer program product, including a computer program, which implements any of the above-mentioned marketing model training methods or marketing methods when executed by a processor.

[0027] The embodiments of the present application provide a marketing model training method, marketing method, device, equipment, medium and program product, which conducts model training by determining the causal relationship between marketing strategies and user behaviors, so that the trained marketing model can process the recommended marketing strategies by analyzing the causal relationship, and determine the target users to be reached corresponding to the recommended marketing strategies, which helps to improve the pertinence and marketing effect of the recommended marketing strategies, and further helps to improve the ability to respond to changes in the market and user behaviors. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 A flowchart of a marketing model training method provided in an embodiment of the present application;

[0029] Figure 2A flowchart of a marketing method provided in an embodiment of the present application;

[0030] Figure 3 A schematic diagram of the structure of a marketing model training device provided in an embodiment of the present application;

[0031] Figure 4 A schematic diagram of the structure of a marketing device provided in an embodiment of the present application;

[0032] Figure 5 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0033] In the securities industry, accurate user analysis and marketing strategies are crucial to a company's competitiveness. Traditional marketing strategies often rely on historical data and simple statistical analysis. While this approach can provide some insights, it lacks a deep understanding of the relationship between marketing strategies and user behavior.

[0034] Currently, most marketing strategy analysis methods focus on correlation analysis between marketing strategies and user behavior, which makes it difficult to accurately identify the direct impact of marketing strategies on user behavior. Furthermore, current analysis methods often lack the ability to adapt to complex markets and changes in user behavior, making it difficult to adjust marketing strategies in real time to respond to market changes.

[0035] The currently commonly used representative analysis method is to use traditional statistical analysis methods to identify people who are sensitive to marketing strategies. This method usually includes steps such as data collection, data preprocessing, feature selection, model training, and result evaluation. However, during the research process, this applicant found that this method ignores the importance of the causal relationship between marketing strategies and user behavior, resulting in a lack of accuracy and effectiveness in the formulation of marketing strategies. In other words, the shortcoming of the current analysis method is that it relies on correlation analysis rather than inference of causal relationships, which limits the accurate evaluation of the effectiveness of marketing strategies. In addition, the currently commonly used analysis methods are usually not real-time and adaptable, and are difficult to cope with the rapid changes in the market and user behavior.

[0036] In response to the above-mentioned problems, the embodiments of the present application provide a marketing model training method, marketing method, device, equipment, medium and program product, which can overcome the limitations of current analysis methods based on the inference analysis of causal relationships, identify people who are sensitive to marketing strategies through in-depth analysis of relevant marketing data, and accurately evaluate the impact of different marketing strategies on user behavior, accurately reach the target marketing population, identify the target users to be reached corresponding to the recommended marketing strategy, and provide more scientific and accurate decision-making support for marketing strategies.

[0037] The following is a further detailed description of the embodiments of the present application in conjunction with the accompanying drawings and examples. It should be understood that the embodiments provided herein are merely for explaining the embodiments of the present application and are not intended to limit the embodiments of the present application. In addition, the embodiments provided below are partial embodiments for implementing the present application, rather than providing all embodiments for implementing the present application. In the absence of conflict, the technical solutions described in the embodiments of the present application can be implemented in any combination.

[0038] It should be noted that, in the embodiments of the present application, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a method or apparatus comprising a series of elements includes not only the elements explicitly stated, but also other elements not explicitly listed, or also includes elements inherent to the implementation of the method or apparatus. In the absence of further restrictions, an element defined by the sentence "include..." does not exclude the presence of other related elements (such as steps in the method or units in the apparatus, for example, a unit in the apparatus may be a portion of a circuit, a portion of a processor, a portion of a program or software, etc.) in the method or apparatus comprising the element.

[0039] The marketing model training method and marketing method provided in the embodiments of the present application include a series of steps, but the marketing model training method and marketing method provided in the embodiments of the present application are not limited to the recorded steps. Similarly, the marketing model training device and marketing device provided in the embodiments of the present application include a series of modules, but the device provided in the embodiments of the present application is not limited to including the modules explicitly recorded, and may also include modules that need to be set up to obtain relevant information or perform processing based on information.

[0040] The present application embodiment provides a marketing model training method, such as Figure 1 As shown, Figure 1 A flowchart of a marketing model training method is shown. Figure 1 The marketing model training method shown includes:

[0041] Step 101: Determine a causal relationship between a first historical marketing strategy and a first user behavior.

[0042] The first historical marketing strategy may be any historical marketing strategy within a historical period, and the first user behavior may be any user behavior related to the first historical marketing strategy. By determining the causal relationship between the first historical marketing strategy and the first user behavior, the causal relationship between any historical marketing strategy and any related user behavior can be obtained.

[0043] The first historical marketing strategy can specifically be a marketing strategy already published in the securities industry, or it can be a marketing strategy already published in other industries. Taking the securities industry as an example, the first historical marketing strategy and the corresponding first user behavior can be collected from multiple channels, such as the securities company's internal system, user interaction platform, and market database, based on the first historical marketing strategy. The first user behavior includes, but is not limited to, multi-dimensional data such as the transaction frequency and transaction amount generated by the first user in response to the first historical marketing strategy. In this embodiment, the first user behavior can also include one or more of the following: whether the user is interested in the first historical marketing strategy, whether the user has a purchase intention for the marketing product corresponding to the first historical marketing strategy, whether the user is satisfied with the marketing product corresponding to the first historical marketing strategy, whether the user has a regular purchasing habit, and whether the user actively promotes the marketing product corresponding to the first historical marketing strategy.

[0044] In the process of determining the causal relationship, the causal relationship between the first historical marketing strategy and the first user behavior can be determined based on the first historical marketing strategy and the first user behavior corresponding to the user. After obtaining the raw data such as the first historical marketing strategy and the first user behavior, the collected raw data can be cleaned and preprocessed. For example, duplicate data can be removed, outliers can be processed, and the data format can be standardized. Cleaning and preprocessing the raw data can improve data availability and accuracy, providing a high-quality data foundation for subsequent analysis and prediction.

[0045] The causal relationship between the first historical marketing strategy and the first user behavior can be determined based on a randomized controlled experiment method, or a causal relationship evaluation model can be constructed based on a causal inference algorithm. The causal relationship between the first historical marketing strategy and the first user behavior is determined through the trained causal relationship evaluation model. This embodiment does not limit the specific method for determining the causal relationship.

[0046] Step 102: Based on the causal relationship and user characteristics, the marketing model is trained so that the trained marketing model can process the causal relationship corresponding to the recommended marketing strategy and obtain the target users to be reached corresponding to the recommended marketing strategy.

[0047] Taking the example of a historical marketing strategy targeting the securities or financial industries, user characteristics can be categorized into different user groups, such as high-net-worth individuals, active traders, and conservative investors, by comprehensively analyzing multi-dimensional data such as transaction frequency, transaction amount, asset size, and risk appetite. Detailed user profile reports can be generated for each group, and user characteristics can be determined based on these profiles. As can be seen, user characteristics can be determined based on user behavior. Since causal relationships are determined based on historical marketing strategies and user behavior, there is a certain correspondence between user characteristics and causal relationships.

[0048] The marketing model is trained through causal relationships and user characteristics. When the trained marketing model processes the recommended marketing strategy and obtains the target users to be reached corresponding to the recommended marketing strategy, it can fully combine the causal relationship between the recommended marketing strategy and user behavior, analyze user characteristics, and predict user behavior that conforms to the causal relationship, thereby obtaining the target users corresponding to the recommended marketing strategy, so that the target users can make positive feedback to the recommended marketing strategy, for example, making the target users to be reached users who are interested in the recommended marketing strategy, or making the target users to be reached users who have the intention to purchase the marketing products of the recommended marketing strategy.

[0049] Specifically, data from a specific historical period, such as six months, can be selected to identify causal relationships and the corresponding user characteristics. For different business scenarios and marketing campaigns, a second dataset can be constructed by randomly sampling a portion of the negative samples from a large number of negative samples to balance data distribution and optimize model training efficiency.

[0050] When training the marketing model using the second data set, the performance of the marketing model can be optimized by adjusting key parameters of the marketing model, such as the learning rate and regularization strength. The cross-validation method can be used to evaluate the generalization ability of the marketing model to ensure the stability and accuracy of the marketing model on different data subsets.

[0051] When building marketing models based on different algorithms, you can compare performance metrics like accuracy, recall, and F1 index to select the optimal marketing model for target users. To meet the business needs of different scenarios, you can also train and validate each marketing model separately, selecting the optimal model to reach target users in each scenario.

[0052] After obtaining a trained marketing model, it can be deployed in a real-world business environment. By inputting real-time user characteristics and causal relationships, the marketing model outputs prediction results to support marketing strategies. Simultaneously, based on business feedback and updated user-related data, the model parameters can be regularly optimized and adjusted to maintain the accuracy and adaptability of the marketing model's prediction results.

[0053] The embodiment of the present application provides a marketing model training method, which trains the marketing model by combining the causal relationship between marketing strategy and user behavior. The trained marketing model can effectively improve the personalized service of marketing activities, and can also timely adjust the marketing strategy through the processing results of the marketing model to enhance the market adaptability of the marketing strategy.

[0054] In practical applications, steps 101 to 102 may be implemented based on a processor, and the processor may be at least one of an application specific integrated circuit (ASIC), a digital signal processor (DSP), a digital signal processing device (DSPD), a programmable logic device (PLD), a field programmable gate array (FPGA), a central processing unit (CPU), a controller, a microcontroller, and a microprocessor.

[0055] Before determining the causal relationship between the first historical marketing strategy and the first user behavior based on step 101, in order to further obtain an accurate causal relationship, in some embodiments, feature variables, processing variables, and result variables can be constructed based on the second historical marketing strategy, the second user behavior, and the second user characteristics; a first data set can be constructed through the feature variables, processing variables, and result variables; a model is trained based on the first data set to obtain a causal relationship evaluation model; correspondingly, step 101 can specifically determine the causal relationship between the first historical marketing strategy and the first user behavior through the causal relationship evaluation model.

[0056] This embodiment further provides a method for determining causal relationships. Specifically, a causal relationship evaluation model can be constructed based on a meta-learner. Meta-learners include T-learners, S-learners, X-learners, and R-learners. Meta-learners can learn good model initialization parameters, allowing the causal relationship evaluation model to quickly converge to an optimal solution based on these parameters when faced with a new task using a small amount of training data.

[0057] In this embodiment, the second historical marketing strategy can be any historical marketing strategy within the historical period, the second user behavior can be the behavior of any user corresponding to the second historical marketing strategy, and the second user characteristic can be the characteristic of any user corresponding to the second historical marketing strategy. The second historical marketing strategy can be the same as the first historical marketing strategy, or a different historical marketing strategy from the first historical marketing strategy. The second user behavior and the first user behavior can be behaviors generated by the same user, or behaviors generated by different users.

[0058] The characteristic variable (X) is a variable that describes the attributes of the user. Specifically, the characteristic variable (X) can be determined based on the second user characteristics and the second user behavior. The processing variable (T) is a variable that is actively manipulated by the technician, and its impact on user behavior can be observed. Specifically, the processing variable can be assigned using methods such as A / B testing and multivariate testing. For example, the users can be divided into two groups, one as the treatment group that receives the second historical marketing strategy, and the other as the control group that receives the conventional marketing strategy or does not receive the second historical marketing strategy. The result variable (Y) is a variable that reflects the user's response to the marketing strategy and can evaluate the effectiveness of the marketing strategy. Specifically, the second user's behavior can be tracked to evaluate the result variable of the second historical marketing strategy relative to the user.

[0059] Depending on the processing requirements, a suitable machine learning model can be selected as a basic learner, such as linear regression, eXtreme Gradient Boosting (XGBoost), neural network, etc. In this embodiment, a causal relationship evaluation model can be constructed based on linear regression and LightGBM models as basic learners.

[0060] After constructing a first data set based on the feature variable (X), the processing variable (T), and the outcome variable (Y), a model can be trained based on the first data set and any algorithm in the Meta-learner to obtain a causal relationship evaluation model. The causal relationship between the first historical marketing strategy and the first user behavior can be directly obtained through the causal relationship evaluation model.

[0061] Causal relationships can be used to determine the user's sensitivity to marketing strategies. That is, for marketing strategies, causal relationships can be used to determine which users can give positive feedback to the marketing strategies. This helps to make marketing strategies more personalized, make timely adjustments to market and user changes, and improve the pertinence and effectiveness of marketing strategies.

[0062] Before the model training based on the first data set is performed to obtain the causal relationship evaluation model as given in the above embodiment, in order to improve the prediction effect of the causal relationship evaluation model, in some embodiments, a model to be trained can be constructed based on a causal inference algorithm; the causal inference algorithm includes a Meta-learner and / or a decision tree Uplift Tree; after the model to be trained is constructed, the model training based on the first data set is performed to obtain the causal relationship evaluation model as given in the above embodiment. Specifically, the model to be trained can be trained based on the first data set to obtain the causal relationship evaluation model.

[0063] In this embodiment, a to-be-trained model can be constructed based on a Meta-learner and / or Uplift Tree, and a causal relationship evaluation model can be obtained by training the to-be-trained model. Uplift Tree is a decision tree model used for causal inference, which is used to estimate the causal effect of a treatment variable (e.g., marketing strategy) on an outcome variable (e.g., user behavior) to obtain a causal relationship.

[0064] When constructing a to-be-trained model based on Meta-learner and determining a causal relationship evaluation model, the to-be-trained model can be constructed based on at least one of T-learner, S-learner, X-learner, and R-learner. When constructing a to-be-trained model based on S-learner, the treatment variable is used as a feature, and the data of the treatment group and the control group are combined to obtain a first dataset, which is used to train the to-be-trained model to obtain a causal relationship evaluation model.

[0065] When building a model to be trained based on T-learner, the data in the first dataset can be divided into a treatment group and a control group. The model to be trained is trained using the treatment group and the control group respectively, and then the causal relationship is estimated by comparing the prediction results of the two trained models.

[0066] When building a training model based on X-learner, you can capture the result distribution of the treatment group and the control group through group modeling, use cross-prediction to generate counterfactual results, and then estimate the treatment effect by grouping and weighted integration to obtain a more accurate causal effect evaluation model.

[0067] When building a model to be trained based on R-learner, the prediction results can be embedded in the training process of the model to be trained by constructing a loss function. The intervention effect of the residual term is utilized, combined with cross-validation, to optimize the prediction results, so that the trained causal relationship evaluation model can achieve more accurate causal relationship prediction results.

[0068] After constructing a to-be-trained model using at least one of the T-learner, S-learner, X-learner, and R-learner algorithms to obtain a causal relationship evaluation model, the first historical marketing strategy and the first user behavior can be processed and analyzed based on the causal relationship evaluation model to obtain an individual causal effect (ITE) or a conditional average causal effect (CATE). Causal relationships can be determined using the ITE and CATE. The performance of the causal relationship evaluation model can be evaluated based on its actual processing hit rate.

[0069] When constructing a model to be trained based on Uplift Tree and determining a causal relationship evaluation model, the first data set can be divided into a treatment group and a control group based on the value of the treatment variable. Uplift Tree can be selected as the model to be trained and the model to be trained constructed by Uplift Tree can be trained. The model to be trained modifies the tree's loss function so that the tree splitting criterion maximizes the causal effect between the treatment group and the control group. The causal relationship evaluation model constructed by the trained Uplift Tree can predict CATE based on the feature variables to obtain a causal relationship. The performance of the model constructed by Uplift Tree can be determined by the hit rate of the causal relationship evaluation model.

[0070] When building a model to be trained based on Meta-learner and Uplift Tree, causal relationships can be estimated more accurately based on the advantages of Meta-learner and Uplift Tree. For example, Meta-learner, such as S-Learner, T-Learner or X-Learner, can be used to preliminarily model the data of the treatment group and the control group. CATE is estimated by optimizing the loss function. Based on the CATE estimated by Meta-learner, Uplift Tree is used for further modeling, by recursively performing a binary split on the data in the first data set, maximizing the improvement metric of the child node, such as Kullback-Leibler Divergence (KL) divergence, Euclidean distance, etc., and finally, the estimated value of ITE is output through the leaf node of Uplift Tree.

[0071] By combining Meta-learner and Uplift Tree to build a model to be trained, a causal relationship evaluation model is obtained, which can further improve the accuracy and robustness of the model and further help understand which features have a significant impact on causal effects.

[0072] In the process of constructing the model to be trained based on the causal inference algorithm given in the above embodiments, in order to further improve the prediction effect of the causal relationship evaluation model, in some embodiments, the causal inference algorithm Meta-learner includes a T-learner algorithm and an X-learner algorithm. When the causal inference algorithm includes Meta-learner, the above-mentioned construction of the model to be trained based on the causal inference algorithm can specifically be based on the X-learner algorithm to construct the model to be trained; the above-mentioned training of the model to be trained based on the first data set can specifically be based on the T-learner algorithm to obtain an estimated result; the estimated result includes the causal relationship obtained based on the T-learner algorithm; based on the estimated result and the first data set, the model to be trained is trained.

[0073] In this embodiment, the causal relationship evaluation model can be determined by combining the T-learner algorithm and the X-learner algorithm. When model training is performed based on the combination of the T-learner algorithm and the X-learner algorithm, the model to be trained is constructed based on the X-learner algorithm. The first data set can be divided into a treatment group and a control group based on the T-learner algorithm, and the two models to be trained are trained based on the data in the treatment group and the control group, respectively. By comparing the prediction results corresponding to the treatment group with the prediction results corresponding to the control group, the preliminary causal effect is calculated to obtain the estimated result corresponding to the T-learner algorithm. Here, the two models to be trained based on the data in the treatment group and the control group can be the same model.

[0074] After obtaining the estimated results corresponding to the T-learner algorithm, the trained model corresponding to the treatment group is used to predict the control group data based on the X-learner algorithm to obtain the "counterfactual" treatment effect of the control group. The trained model corresponding to the control group is used to predict the treatment group data to obtain the "counterfactual" treatment effect of the treatment group. The T-learner estimated results are corrected by performing a weighted combination of the "counterfactual" prediction differences between the treatment group and the control group. Here, the weights in the weighted combination process can be calculated based on the propensity score to balance the differences in the covariate distribution between the treatment group and the control group.

[0075] By combining the T-learner and X-learner algorithms for model training, a causal relationship evaluation model is developed. This model improves its processing capabilities and accuracy, making it suitable for handling highly heterogeneous application scenarios and more accurately identifying the causal effects of different users. By applying meta-learning algorithms and gain tree algorithms, the potential causal relationships between marketing strategies and user behaviors are deeply analyzed, helping to identify users who are likely to generate positive feedback on specific marketing campaigns and improve the return on investment (ROI) of marketing campaigns.

[0076] Before determining the causal relationship between the first historical marketing strategy and the first user behavior in the above embodiment, users can also be divided based on user data to obtain user groups; user data includes one or more of the user's transaction frequency, transaction amount, asset size, and risk preference; after obtaining the user grouping, the above determination of the causal relationship between the first historical marketing strategy and the first user behavior can specifically be to determine the causal relationship between the first historical marketing strategy and the first user behavior in each user group.

[0077] User data includes, but is not limited to, one or more of a user's transaction frequency, transaction amount, asset size, and risk preference. Before determining the causal relationship between the first historical marketing strategy and the first user's behavior, users can be segmented into user groups through a comprehensive analysis of multi-dimensional user data. For example, users can be specifically categorized as high-net-worth users, active traders, conservative investors, etc.

[0078] User portraits can be determined by grouping users, user characteristics can be obtained, and based on the method given in the above embodiment, the causal relationship between the first historical marketing strategy and the first user behavior of users in each user group can be determined.

[0079] It can be seen that by collecting data from multiple dimensions, the comprehensiveness and accuracy of user data and marketing strategy data are ensured, which helps to more comprehensively understand market dynamics and user behavior, thereby providing richer information support for marketing strategy decisions.

[0080] After training the marketing model and grouping users based on user data, the trained marketing model can process the causal relationship corresponding to the recommended marketing strategy and obtain the target users to be reached in each user group by the recommended marketing strategy.

[0081] Based on the method given in the above embodiment, after obtaining the trained marketing model, the causal relationship corresponding to the recommended marketing strategy can be processed based on the trained marketing model to obtain the impact of the recommended marketing strategy on the user behavior of each user in each user group, that is, to predict the feedback of each user in each user group on the recommended marketing strategy, such as whether each user in each user group is interested in the recommended marketing strategy, whether they will purchase the marketing product corresponding to the recommended marketing strategy, etc.

[0082] Based on the prediction results of the trained marketing model, the target users to be reached can be determined in each user group. For example, according to the prediction results of the trained marketing model, the predicted user behaviors are divided, and users who are interested in the recommended marketing strategies are determined. Based on the degree of interest of users in the recommended marketing strategies, users are ranked from high to low. In each user group, the top 20% of users are determined as target users to be reached, and the recommended marketing strategies are recommended to the target users, thereby improving users' response enthusiasm to the recommended marketing strategies.

[0083] In specific applications, a prediction model can be determined based on the causal relationship evaluation model and the marketing model. This allows for a deeper understanding of user behavior through multiple features, including basic user information, asset and transaction information, number of days browsed, number of clicks, marketing type of browsed, and marketing type of last click. Causal relationships can be determined using the causal relationship evaluation model within the prediction model, and target users can be identified using the marketing model within the prediction model, providing rich data support for the formulation and updating of marketing strategies. The prediction results of the trained marketing model can be used to determine the user's willingness to purchase the value-added service products corresponding to the recommended marketing strategy.

[0084] Those skilled in the art will understand that in the above-mentioned method of the specific implementation method, the writing order of each step does not mean a strict execution order and does not constitute any limitation on the implementation process. The specific execution order of each step should be determined by its function and possible internal logic.

[0085] The present application also provides a marketing method, such as Figure 2 As shown, Figure 2 The marketing methods shown include:

[0086] Step 201: Process the causal relationship corresponding to the recommended marketing strategy through the marketing model to obtain the target users to be reached corresponding to the recommended marketing strategy.

[0087] The marketing model is obtained by training based on the marketing model training method given in the above embodiment.

[0088] Step 202: Reach target users based on the marketing strategy to be recommended.

[0089] Based on the marketing model trained using the above-described embodiment, the causal relationship between the marketing strategy to be recommended and user behavior is processed to determine the user's purchase intention for the marketing product corresponding to the marketing strategy to be recommended. Users whose purchase intention exceeds a purchase intention threshold are identified as target users, and the target users are reached, i.e., the marketing strategy to be recommended is recommended to the target users. Specifically, the causal relationship between the marketing strategy to be recommended and user behavior can be determined based on the above-described causal relationship evaluation model. Based on the causal relationship, the marketing strategy to be recommended, and user characteristics, the target users to be reached are determined using the marketing model.

[0090] In actual applications, the recommended marketing strategy can also be adjusted based on the user's purchasing intention obtained from the marketing model. By adjusting the recommended marketing strategy, the user's purchasing intention can be increased, or the proportion of users whose purchasing intention is greater than the willingness threshold can be increased, thereby improving the ROI of the recommended marketing strategy.

[0091] In practical applications, steps 201 to 202 may be implemented based on a processor, and the processor may be at least one of an ASIC, a DSP, a DSPD, a PLD, an FPGA, a CPU, a controller, a microcontroller, and a microprocessor.

[0092] Based on the marketing model training method and marketing method given in the above embodiments, the embodiments of the present application provide an intelligent operation engine based on causal inference, which includes a data collection module, a data processing module, a feature extraction module, a causal inference module, a user behavior prediction module and a decision support module.

[0093] Among them, the data collection module is used to collect marketing data from multiple channels such as internal systems, customer interaction platforms, market databases, etc., including multi-dimensional data such as transaction frequency, transaction amount, asset size and risk preference, to ensure the comprehensiveness and accuracy of the collected data.

[0094] The data processing module is used to perform in-depth cleaning and preprocessing on the raw data collected by the data collection module, such as data deduplication, outlier processing, and unified data format, so as to improve the quality of the raw data and lay a solid foundation for subsequent analysis.

[0095] The feature extraction module analyzes multi-dimensional data such as transaction frequency, transaction amount, asset size, and risk appetite to categorize users into different groups, such as high-net-worth individuals, active traders, and conservative investors. Detailed user profile reports are generated for these groups to identify user characteristics, providing a solid data foundation for developing targeted marketing strategies and services.

[0096] The causal inference module is used to analyze the causal relationship between marketing strategies and user behavior based on Meta-learner and Uplift Tree, and identify which marketing strategies have a significant impact on user behavior.

[0097] The user behavior prediction module is used to build a marketing model based on the causal relationship analyzed by the causal inference module. The marketing model is used to predict whether users can generate positive feedback on marketing strategies and provide guidance for marketing activities.

[0098] The decision support module is used to integrate the forecast results of the marketing model and market trends, and filter the top 20% of users in the forecast data ranking for reach.

[0099] The embodiments of the present application provide a marketing model training method and a marketing method that can reduce the need for manual intervention through automated data collection and processing, reduce operating costs, and improve the operational efficiency of financial institutions. By predicting user behavior, when formulating marketing strategies, it is possible to respond to market changes more quickly, adjust marketing strategies in a timely manner, and enhance market adaptability. Through accurate user behavior prediction and decision support, marketers and financial institutions can more effectively allocate operational resources, concentrate resources in areas most likely to produce results, and improve resource utilization efficiency. At the same time, through more accurate marketing strategies and personalized services, user satisfaction and loyalty are improved, which helps to maintain user relationships in the long term.

[0100] Corresponding to the marketing model training method given in the above embodiment, based on the marketing model training method given in the above embodiment, the embodiment of the present application also proposes a marketing model training device, such as Figure 3 As shown, the marketing model training device includes:

[0101] The determination module 301 is used to determine the causal relationship between the first historical marketing strategy and the first user behavior.

[0102] The training module 302 is used to train the marketing model based on the causal relationship and user characteristics, so that the trained marketing model can process the causal relationship corresponding to the recommended marketing strategy and obtain the target users to be reached corresponding to the recommended marketing strategy.

[0103] In practical applications, the determination module 301 and the training module 302 can be implemented based on a processor and a communication device.

[0104] In some embodiments, before determining the causal relationship between the first historical marketing strategy and the first user behavior, the training module 302 is also used to construct feature variables, processing variables, and result variables based on the second historical marketing strategy, the second user behavior, and the second user characteristics; construct a first data set through the feature variables, processing variables, and result variables; perform model training based on the first data set to obtain a causal relationship evaluation model; the training module 302 is specifically used to determine the causal relationship between the first historical marketing strategy and the first user behavior through the causal relationship evaluation model.

[0105] In some embodiments, before performing model training based on the first data set to obtain a causal relationship evaluation model, the training module 302 is also used to construct a model to be trained based on a causal inference algorithm; the causal inference algorithm includes a Meta-learner and / or an Uplift Tree; the training module 302 is specifically used to train the model to be trained based on the first data set to obtain a causal relationship evaluation model.

[0106] In some embodiments, Meta-learner includes a T-learner algorithm and an X-learner algorithm. When the causal inference algorithm includes Meta-learner, the training module 302 is specifically used to construct a model to be trained based on the X-learner algorithm; obtain an estimated result based on the T-learner algorithm; the estimated result includes the causal relationship obtained based on the T-learner algorithm; and train the model to be trained based on the estimated result and the first data set.

[0107] In some embodiments, the marketing model training device also includes a division module. Before determining the causal relationship between the first historical marketing strategy and the first user behavior, the division module is used to divide users based on user data to obtain user groups; the user data includes one or more of the user's transaction frequency, transaction amount, asset size, and risk preference; the determination module 301 is specifically used to determine the causal relationship between the first historical marketing strategy and the first user behavior in each user group.

[0108] Corresponding to the marketing method given in the above embodiment, based on the marketing method given in the above embodiment, the embodiment of the present application also proposes a marketing device, such as Figure 4 As shown, the marketing device includes:

[0109] The processing module 401 is used to process the causal relationship corresponding to the recommended marketing strategy through the marketing model to obtain the target users to be reached corresponding to the recommended marketing strategy; the marketing model is trained based on the above-mentioned marketing model training method.

[0110] The interaction module 402 is used to reach target users based on the marketing strategy to be recommended.

[0111] In practical applications, the processing module 401 and the interaction module 402 can be implemented based on a processor and a communication device.

[0112] It should be noted that the description of the above device embodiment is similar to the description of the above method embodiment, and has similar beneficial effects as the same method embodiment. For technical details not disclosed in the device embodiment of this application, please refer to the description of the method embodiment of this application for understanding.

[0113] It should be noted that, in the embodiment of the present application, if the above method is implemented in the form of a software function module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiment of the present application is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which is stored in a storage medium and includes a number of instructions for enabling a computer device (which can be a terminal, server, etc.) to execute all or part of the methods described in each embodiment of the present application. The aforementioned storage medium includes various media that can store program codes, such as a U disk, a mobile hard disk, a read-only memory (ROM), a magnetic disk or an optical disk. In this way, the embodiment of the present application is not limited to any specific combination of hardware and software.

[0114] An embodiment of the present application also provides an electronic device. Figure 5 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application is shown in FIG. Figure 5 As shown, the electronic device 50 may include:

[0115] The memory 501 is used to store executable instructions.

[0116] The processor 502 is configured to implement any one of the above-mentioned marketing model training methods and marketing methods when executing the executable instructions stored in the memory 501.

[0117] The processor 502 may be at least one of an ASIC, a DSP, a DSPD, a PLD, an FPGA, a CPU, a controller, a microcontroller, and a microprocessor.

[0118] The above-mentioned computer-readable storage medium or memory 501 can be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), a magnetic random access memory (FRAM), a flash memory (Flash Memory), a magnetic surface memory, an optical disc, or a compact disc read-only memory (CD-ROM); it can also be various terminals including one or any combination of the above-mentioned memories, such as mobile phones, computers, tablet devices, personal digital assistants, etc.

[0119] Correspondingly, an embodiment of the present application further provides a computer storage medium, on which computer executable instructions are stored, and the computer executable instructions are used to implement any one of the marketing model training methods and marketing methods provided in the above embodiments.

[0120] Correspondingly, an embodiment of the present application further provides a computer program product, which includes computer-executable instructions, and the computer-executable instructions are used to implement any marketing model training method and marketing method provided in the embodiment of the present application.

[0121] In some embodiments, the functions or modules included in the device provided in the embodiments of the present application can be used to execute the method described in the above method embodiments. The specific implementation can refer to the description of the above method embodiments. For the sake of brevity, it will not be repeated here.

[0122] The above description of the various embodiments tends to emphasize the differences between the various embodiments. The same or similar aspects can be referenced with each other and will not be repeated herein for the sake of brevity.

[0123] The methods disclosed in the various method embodiments provided in this application can be arbitrarily combined without conflict to obtain new method embodiments.

[0124] The features disclosed in the various product embodiments provided in this application can be arbitrarily combined without conflict to obtain new product embodiments.

[0125] The features disclosed in the various method or device embodiments provided in this application can be arbitrarily combined without conflict to obtain new method embodiments or device embodiments.

[0126] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus the necessary general hardware platform, and of course 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 application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for enabling a terminal (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in each embodiment of the present application.

[0127] The embodiments of the present application are described above in conjunction with the accompanying drawings, but the present application is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of this application, ordinary technicians in this field can also make many forms without departing from the purpose of this application and the scope of protection of the claims, all of which are protected by this application.

Claims

1. A marketing model training method, characterized in that: The method comprises: Determine the causal relationship between the first historical marketing strategy and the first user behavior; Based on the causal relationship and user characteristics, the marketing model is trained so that the trained marketing model can process the causal relationship corresponding to the recommended marketing strategy and obtain the target users to be reached corresponding to the recommended marketing strategy.

2. The method according to claim 1, characterized in that Before determining the causal relationship between the first historical marketing strategy and the first user behavior, the method further includes: Based on the second historical marketing strategy, the second user behavior, and the second user characteristics, constructing feature variables, processing variables, and result variables; constructing a first data set through the feature variables, the processing variables, and the result variables; Performing model training based on the first data set to obtain a causal relationship assessment model; Determining the causal relationship between the first historical marketing strategy and the first user behavior includes: The causal relationship between the first historical marketing strategy and the first user behavior is determined through the causal relationship evaluation model.

3. The method according to claim 2, characterized in that Before performing model training based on the first data set to obtain a causal relationship evaluation model, the method further includes: Constructing a model to be trained based on a causal inference algorithm; the causal inference algorithm includes a meta-learner and / or a decision tree uplift tree; The performing model training based on the first data set to obtain a causal relationship assessment model includes: Based on the first data set, the model to be trained is trained to obtain a causal relationship evaluation model.

4. The method according to claim 3, characterized in that The Meta-learner includes a T-learner algorithm and an X-learner algorithm. When the causal inference algorithm includes the Meta-learner, constructing a model to be trained based on the causal inference algorithm includes: Based on the X-learner algorithm, a model to be trained is constructed; The training of the model to be trained based on the first data set includes: Obtaining an estimated result based on the T-learner algorithm; the estimated result including a causal relationship obtained based on the T-learner algorithm; The model to be trained is trained based on the estimation result and the first data set.

5. The method according to claim 1, wherein Before determining the causal relationship between the first historical marketing strategy and the first user behavior, the method further includes: Dividing users into user groups based on user data, wherein the user data includes one or more of the user's transaction frequency, transaction amount, asset size, and risk preference; Determining the causal relationship between the first historical marketing strategy and the first user behavior includes: A causal relationship between the first historical marketing strategy and the first user behavior in each user group is determined.

6. The method according to claim 5, characterized in that The trained marketing model can process the causal relationship corresponding to the recommended marketing strategy to obtain the target users to be reached by the recommended marketing strategy in each user group.

7. A marketing method, characterized in that The method comprises: Processing the causal relationship corresponding to the recommended marketing strategy through a marketing model to obtain target users to be reached corresponding to the recommended marketing strategy; the marketing model is trained based on the method according to any one of claims 1 to 6; Based on the marketing strategy to be recommended, the target users are reached.

8. A marketing model training device, characterized in that: The device comprises: a determination module, configured to determine a causal relationship between the first historical marketing strategy and the first user behavior; The training module is used to train the marketing model based on the causal relationship and user characteristics, so that the trained marketing model can process the causal relationship corresponding to the recommended marketing strategy and obtain the target users to be reached corresponding to the recommended marketing strategy.

9. A marketing device, characterized in that: The device comprises: A processing module, configured to process the causal relationship corresponding to the marketing strategy to be recommended using a marketing model to obtain target users to be reached corresponding to the marketing strategy to be recommended; the marketing model is trained based on the method according to any one of claims 1 to 6; An interactive module is used to reach the target users based on the marketing strategy to be recommended.

10. An electronic device, characterized in that: The electronic device comprises a processor and a memory for storing a computer program that can be run on the processor; wherein, The processor is configured to run the computer program to perform the method according to any one of claims 1 to 6 or claim 7.

11. A computer storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 6 or claim 7 is implemented.

12. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the computer program implements the method according to any one of claims 1 to 6 or claim 7.

Images