Marketing method and system combined with third-party platform, electronic equipment and storage medium

Abstract

The invention provides a marketing method and system combined with a third-party platform, electronic equipment and a storage medium. The method comprises the following steps: constructing a first prediction model on a target platform; the first prediction model is a model obtained by encrypting first behavior data of the target platform by adopting a privacy computing technology and training according to the encrypted first behavior data; acquiring parameters of a second prediction model of the third-party platform, and optimizing the first prediction model according to the parameters of the second prediction model; the second prediction model is obtained by training the third-party platform according to the encrypted second behavior data; using the optimized first prediction model to predict behaviors and preferences of the user to obtain a prediction result; a marketing strategy is optimized according to the prediction result; namely, by obtaining the model parameters of the third-party platform, the user is comprehensively understood, favorable data support is provided for formulating the marketing strategy, and the effectiveness of the strategy and the resource use efficiency are improved.

Description

Technical Field

[0001] The present invention belongs to the field of data processing technology, and more specifically, particularly relates to a marketing method and system, electronic equipment, and storage medium combined with a third-party platform. Background Art

[0002] In the current financial market, the diversification of marketing models has become a key means for banks to enhance their competitiveness. However, with the increasing number of third-party channels, banks have encountered unprecedented challenges in obtaining user information.

[0003] At present, many banks rely mainly on their own channels and data resources in the marketing process. However, when users interact through third-party channels, banks often cannot directly obtain detailed information about these users. This limitation in information acquisition makes it impossible for banks to conduct comprehensive and in-depth analysis of user behaviors and preferences.

[0004] This situation has brought significant negative impacts on the marketing strategies of banks. First, due to the inability to fully understand users, banks lack sufficient data support when formulating marketing strategies, which limits the effectiveness of strategies and the degree of optimization of resource allocation. Summary of the invention

[0005] In view of this, the purpose of the present invention is to provide a marketing method and system, electronic device, and storage medium combined with a third-party platform, which are used to obtain model parameters of the third-party platform, comprehensively understand users, provide favorable data support for formulating marketing strategies, and improve the effectiveness of strategies and resource utilization efficiency.

[0006] The first aspect of the present application discloses a marketing method in combination with a third-party platform, comprising:

[0007] Constructing a first prediction model on the target platform; the first prediction model is a model obtained by encrypting the first behavior data of the target platform using privacy computing technology and training based on the encrypted first behavior data; the first prediction model is used to predict the user's behavior and preference based on the user's behavior data;

[0008] Obtaining parameters of a second prediction model of a third-party platform, and optimizing the first prediction model according to the parameters of the second prediction model; the second prediction model is obtained by training the third-party platform according to the encrypted second behavior data;

[0009] The optimized first prediction model is used to predict the user's behavior and preference to obtain a prediction result;

[0010] Optimize marketing strategies based on the prediction results.

[0011] Optionally, the using the optimized first prediction model to predict the user's behavior and preference to obtain a prediction result includes:

[0012] The optimized first prediction model is used to predict the behavior and preferences of users who have not redeemed their rights on the target platform to obtain prediction results.

[0013] Optionally, constructing a first prediction model on the target platform includes:

[0014] Acquire first user behavior data of the user on the target platform;

[0015] A logistic regression model is selected to be trained using a gradient descent method, and parameter aggregation is performed to obtain an initial prediction model of the target platform;

[0016] The secondary prediction model of the target platform is iteratively trained to obtain a final prediction model of the target platform; the secondary prediction model of the target platform is a model obtained by synchronizing parameters with the primary prediction model of the third-party platform and continuing to train according to the synchronized parameters.

[0017] Optionally, before obtaining the parameters of the second prediction model on the third-party platform, the method further includes: constructing the second prediction model on the third-party platform; the constructing the second prediction model on the third-party platform includes:

[0018] Obtaining second user behavior data of the user on the third-party platform;

[0019] A logistic regression model is selected and trained using a gradient descent method, and parameter aggregation is performed to obtain an initial prediction model of the third-party platform;

[0020] The secondary prediction model of the third-party platform is iteratively trained to obtain a final prediction model of the third-party platform; the secondary prediction model of the third-party platform is a model obtained by synchronizing parameters with the initial prediction model of the target platform and continuing to train according to the synchronized parameters.

[0021] Optionally, the first behavior data includes: redemption records and transaction records of the user;

[0022] The second behavior data includes: rights and interests field records and browsing records.

[0023] Optionally, the privacy computing technology includes: homomorphic encryption and secure multi-party computing.

[0024] The second aspect of the present application discloses a marketing system combined with a third-party platform, including:

[0025] A construction module, used to construct a first prediction model on a target platform; the first prediction model is a model obtained by encrypting the first behavior data of the target platform using privacy computing technology and training based on the encrypted first behavior data; the first prediction model is used to predict the user's behavior and preference based on the user's behavior data;

[0026] An exchange optimization module, used to obtain parameters of a second prediction model of a third-party platform, and optimize the first prediction model according to the parameters of the second prediction model; the second prediction model is obtained by training the third-party platform according to the encrypted second behavior data;

[0027] The prediction module is used to use the optimized first prediction model to predict the user's behavior and preference to obtain a prediction result; and optimize the marketing strategy according to the prediction result.

[0028] Optionally, the prediction module is used to use the optimized first prediction model to predict the user's behavior and preference, and when the prediction result is obtained, it is specifically used to:

[0029] The optimized first prediction model is used to predict the behavior and preferences of users who have not redeemed their rights on the target platform to obtain prediction results.

[0030] The third aspect of the present application discloses an electronic device, comprising:

[0031] one or more processors;

[0032] a storage device having one or more programs stored thereon;

[0033] When one or more programs are executed by one or more processors, the one or more processors implement the marketing method combined with a third-party platform as described in any one of the first aspects of the present application.

[0034] A fourth aspect of the present application discloses a storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the marketing method combined with a third-party platform as described in any one of the first aspect of the present application is implemented.

[0035] It can be seen from the above technical scheme that the present invention provides a marketing method combined with a third-party platform, including: constructing a first prediction model on the target platform; the first prediction model is a model obtained by encrypting the first behavior data of the target platform using privacy computing technology, and training based on the encrypted first behavior data; the first prediction model is used to predict the user's behavior and preference based on the user's behavior data; obtaining the parameters of the second prediction model of the third-party platform, and optimizing the first prediction model based on the parameters of the second prediction model; the second prediction model is obtained by training the third-party platform based on the encrypted second behavior data; using the optimized first prediction model, predicting the user's behavior and preference to obtain a prediction result; optimizing the marketing strategy based on the prediction result; that is, by obtaining the model parameters of the third-party platform, a comprehensive understanding of the user is achieved, providing favorable data support for the formulation of marketing strategies, improving the effectiveness of the strategy and resource utilization efficiency, and addressing the data privacy and analysis accuracy issues faced when conducting marketing activities through a third-party platform. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0037] Figure 1 is a flow chart of a marketing method combined with a third-party platform provided by an embodiment of the present invention;

[0038] Figure 2 is a flow chart of another marketing method combined with a third-party platform provided by an embodiment of the present invention;

[0039] Figure 3 is a flow chart of another marketing method combined with a third-party platform provided by an embodiment of the present invention;

[0040] Figure 4 is a schematic diagram of a marketing system combined with a third-party platform provided by an embodiment of the present invention;

[0041] Figure 5 It is a schematic diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0042] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0043] In the present application, the term "comprise", "include" or any other variant thereof is intended to cover non-exclusive inclusion, so that the process, method, article or equipment including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method, article or equipment. In the absence of more restrictions, the elements defined by the sentence "comprise one ..." do not exclude the existence of other identical elements in the process, method, article or equipment including the elements. In addition, the terms "first", "second", "third", "fourth", etc. (if present) in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and do not have to be used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable in appropriate circumstances, so that the embodiments of the present application described here can be implemented in an order other than those illustrated or described here, for example.

[0044] The present application provides a marketing method combined with a third-party platform to solve the problem in the prior art that, due to the inability to fully understand users, banks lack sufficient data support when formulating marketing strategies, resulting in the effectiveness of strategies and the degree of optimization of resource allocation being limited, and, when evaluating the effectiveness of equity investment, banks find it difficult to accurately identify the characteristics of users who have not redeemed their equity through internal channels, which further aggravates the blindness of marketing decisions; in addition, this information asymmetry also makes it impossible for banks to fully tap into potential customer groups, reducing the return on marketing investment, thereby restricting the bank's advantage in market competition.

[0045] Explanation of relevant terms:

[0046] Federated learning, a distributed machine learning method, allows multiple parties to collaborate on training models without sharing their original data. It performs machine learning while protecting user privacy by passing model parameters (not data) between participants.

[0047] Privacy computing aims to protect the privacy and security of data during data processing and analysis. Common technologies include homomorphic encryption and secure multi-party computing, which allow data to be processed and analyzed in an encrypted state, thereby protecting the data content from being leaked.

[0048] Homomorphic encryption, a special encryption method, allows computational operations (such as addition and multiplication) to be performed directly on encrypted data without first decrypting the data. This method is very useful for performing complex calculations while ensuring data privacy.

[0049] Secure multi-party computation is a technology that allows multiple parties to perform computations together without exposing their input data. It is often used to collaboratively solve common computational problems without revealing the privacy information of each party.

[0050] User portrait is a comprehensive description of user characteristics, including user behavior, preferences, purchase history, etc. In marketing and recommendation systems, user portraits are used to understand customers and provide more accurate services or products.

[0051] Machine learning is an artificial intelligence technology that enables computer systems to use data to automatically learn and improve. It uses algorithms to analyze data, discover patterns in the data, and use these patterns to make predictions about new data.

[0052] Predictive models, a type of model in machine learning, are used to predict future events or behaviors based on existing data. In the present invention, predictive models are used to predict whether users will respond to specific marketing activities.

[0053] See also Figure 1 , combined with third-party platform marketing methods, including:

[0054] S101. Construct a first prediction model on a target platform.

[0055] The first prediction model is a model obtained by encrypting the first behavior data of the target platform using privacy computing technology and training based on the encrypted first behavior data; the first prediction model is used to predict the user's behavior and preferences based on the user's behavior data.

[0056] On the target platform, we first build the first prediction model. This step is crucial because it will form the basis for subsequent predictions of user behavior and preferences. In order to ensure the security and privacy of user data, advanced privacy computing technology can be used to encrypt the first behavior data collected by the target platform. These encrypted data not only protect the privacy of users, but also retain the validity and availability of the data. Next, use these encrypted behavior data to train the first prediction model. After careful design and multiple iterations, this model can eventually accurately predict their behavior and preferences based on their behavior data.

[0057] S102: Obtain parameters of a second prediction model of a third-party platform, and optimize the first prediction model according to the parameters of the second prediction model.

[0058] The second prediction model is trained by the third-party platform based on the encrypted second behavior data.

[0059] In order to further improve the accuracy and wide applicability of the prediction model, we actively cooperated with third-party platforms. These third-party platforms hold a large amount of user behavior data resources, and have carefully created a second prediction model using these valuable data. Following strict security and compliance processes, we successfully obtained the core parameters of these second prediction models. In other words, companies that can go online on third-party platforms to jump to mobile banking marketing activities with one click, conduct public domain diversion, and carry out cross-domain marketing can carry out precise collaborative marketing on the premise that the privacy of their respective user data is protected.

[0060] Next, we used these parameters to deeply optimize the first prediction model. This integration process aims to fully absorb the strengths of different data sources and models, and achieve complementary advantages and performance improvement between models.

[0061] After this optimization, the first prediction model has been revitalized. It can more accurately capture the unique behaviors and preferences of different users in different scenarios, and provide users with a more considerate and personalized service experience.

[0062] In other words, the training process of the first prediction model is to enable the target platform and the third-party platform to collaborate through the federated learning mechanism to train an efficient machine learning model without directly sharing the original user data. This innovative approach not only ensures the privacy and security of user data, but also opens up a way for the target platform to integrate the data advantages of the third-party platform, thereby improving the accuracy of user portraits and the effectiveness of marketing strategies. Using federated learning technology, the target platform can indirectly use the user data of the third-party platform to optimize and refine the prediction model and improve its prediction accuracy.

[0063] S103: Use the optimized first prediction model to predict the user's behavior and preference to obtain a prediction result.

[0064] After successfully completing the deep optimization of the model, the prediction process for specific user behaviors and personalized preferences was officially launched. In order to obtain first-hand information on the user's current behavior status, the user's latest behavior data was captured and integrated in real time, including but not limited to the user's browsing history, click behavior, purchase history, and interaction status on the platform.

[0065] Subsequently, these filtered and sorted data are input into the first optimized prediction model. With its advanced algorithm architecture and efficient computing power, the model can quickly and accurately predict the specific behaviors that users may take in the future based on these detailed behavioral data, such as the categories of goods that users may browse, the interactive activities that they may participate in, and the purchase intentions that they may generate.

[0066] At the same time, the model can also provide in-depth insights into users' personalized preferences, such as user preferences for product styles, price ranges, and promotional activities. These prediction results not only provide a window for in-depth understanding of user needs and market dynamics, but also provide valuable reference for optimizing service experience and improving user satisfaction.

[0067] By accurately predicting users' specific behaviors and personalized preferences, we can more accurately push products and services that meet user needs, thereby improving users' shopping experience and satisfaction, and further consolidating and expanding market share.

[0068] S104. Optimize marketing strategies based on prediction results.

[0069] Based on the accurate prediction results, we start to deeply optimize the marketing strategy. Through a comprehensive and in-depth understanding of user behavior and preferences, we can more accurately target the target user group and formulate a series of highly targeted marketing plans accordingly.

[0070] For example, it can intelligently recommend products or services that suit users’ shopping preferences, thereby greatly improving users’ willingness to buy and satisfaction. At the same time, it can also cleverly adjust the timing of marketing campaigns based on users’ active time, ensuring that preferential information is accurately delivered when users are most likely to pay attention and participate.

[0071] These carefully designed optimization measures not only significantly improved the conversion rate and effectiveness of marketing activities, but also invisibly enhanced user satisfaction and loyalty. Because users feel the care and concern, they are more willing to establish long-term and stable cooperative relationships, which brings a steady stream of customers and word-of-mouth effects.

[0072] In other words, based on the insights provided by the first prediction model, the target platform can formulate more accurate and personalized marketing strategies, optimize resource allocation, and improve the overall effectiveness of marketing activities.

[0073] In other words, through federated learning technology, cross-channel data fusion and analysis are achieved, enabling banks to obtain more comprehensive insights into user behavior, thereby significantly improving the accuracy and depth of cross-channel data analysis. With the output results of the federated learning model, banks can formulate more accurate marketing strategies, achieve efficient use of resource allocation, and further optimize the overall performance of marketing activities and improve business results.

[0074] More specifically, using forecast results to optimize marketing strategies is a systematic and in-depth process that involves multiple links and levels. The following is a detailed description of how to use forecast results to optimize marketing strategies:

[0075] First, accurately locate the target user group.

[0076] User segmentation: Through predictive analysis, users can be segmented based on multiple dimensions such as purchasing intention, behavioral characteristics, acceptance, income, age, financial experience, etc. This more detailed grouping helps to identify the unique needs and preferences of different user groups, thereby formulating more targeted marketing strategies.

[0077] Predictive Scoring: Scoring prospects, leads, and customers based on their likelihood of purchasing allows you to prioritize high-value prospects. This helps optimize resource allocation and put more effort into user groups with greater potential.

[0078] Second, develop more targeted marketing plans.

[0079] Personalized content push: Through predictive analysis, you can understand which types of content are more suitable for certain potential customers. Based on the user's personalized needs and preferences, you can customize personalized content push, improve the quality of communication with users, and increase the possibility of sales conversion.

[0080] Optimize the push time of marketing activities: According to the user's active time, the push time of marketing activities can be adjusted to ensure that the preferential information is accurately delivered when the user is most likely to pay attention and participate, thereby improving the conversion rate and effect of marketing activities.

[0081] Cross-selling and up-selling strategies: Predictive analysis can be used to identify additional products or services that users may be interested in. Through cross-selling and up-selling strategies, users’ purchase amounts and purchase frequency can be increased, improving overall sales performance.

[0082] Next, improve user satisfaction and loyalty.

[0083] Customer Lifetime Value Prediction: Through predictive analysis, the lifetime value of each customer can be evaluated. This helps to formulate more reasonable customer maintenance strategies and improve customer satisfaction and loyalty.

[0084] Predicting customer churn: Predictive analysis can identify customer groups that are likely to churn. By taking timely retention measures, such as providing coupons and improving service quality, the customer churn rate can be reduced and the customer base can be kept stable.

[0085] Optimize pricing strategy: Predictive analysis can also help develop a more reasonable pricing strategy. By analyzing the user's purchase history and price sensitivity, the best price point can be determined to improve the competitiveness and profitability of the product.

[0086] Finally, continuously monitor and adjust your marketing strategy.

[0087] Tracking the effectiveness of marketing campaigns: Using predictive analytics, you can continuously monitor the effectiveness of marketing campaigns.

[0088] By analyzing data indicators such as conversion rate, click-through rate, purchase volume, etc., the effectiveness of marketing strategies can be evaluated and timely adjusted and optimized.

[0089] Iterate and optimize the forecasting model: As time goes by and the market environment changes, the forecasting model may need to be continuously iterated and optimized.

[0090] Models can be trained and adjusted based on new data and information to improve the accuracy and reliability of predictions.

[0091] In other words, using prediction results to optimize marketing strategies is a complex process involving multiple links and levels. By accurately positioning the target user group, formulating more targeted marketing plans, improving user satisfaction and loyalty, and continuously monitoring and adjusting marketing strategies, it is possible to optimize and upgrade marketing strategies and improve marketing effectiveness and market competitiveness.

[0092] Throughout the entire process, we always pay attention to data security and privacy protection to ensure that users' personal information is not leaked or abused. At the same time, we continue to optimize and improve the prediction model to adapt to the ever-changing market environment and user needs.

[0093] In this embodiment, a first prediction model is constructed on the target platform; the first prediction model is a model obtained by encrypting the first behavior data of the target platform using privacy computing technology and training based on the encrypted first behavior data; the first prediction model is used to predict the user's behavior and preference based on the user's behavior data; the parameters of the second prediction model of the third-party platform are obtained, and the first prediction model is optimized based on the parameters of the second prediction model; the second prediction model is obtained by training the third-party platform based on the encrypted second behavior data; the user's behavior and preference are predicted using the optimized first prediction model to obtain a prediction result; based on the prediction result, the marketing strategy is optimized; that is, by obtaining the model parameters of the third-party platform, the user is fully understood, favorable data support is provided for the formulation of marketing strategies, the effectiveness of the strategy and the efficiency of resource utilization are improved, and the data privacy and analysis accuracy issues faced when conducting marketing activities through the third-party platform are addressed.

[0094] It should be noted that in the current banking marketing practice, facing the challenges of data sharing and user analysis when cooperating with third-party platforms, common strategies include: banks and third-party platforms share basic user information and behavioral data based on established agreements. These data are the key to analyzing user preferences and building marketing strategies. Banks purchase user data sets directly from third-party data suppliers for market analysis and accurate customer identification. Faced with data sharing barriers, banks only perform simplified user behavior analysis based on limited basic data (such as click-through rate and conversion rate). However, detailed user data sharing may violate privacy protection regulations and increase the risk of data leakage. Third-party platforms are unwilling to share complete user data for various reasons, which limits banks' comprehensive insight into user behavior. The lack of data may cause banks' analysis results to deviate from reality and make it difficult to accurately capture users' true preferences and behavior patterns.

[0095] To solve the above problems, this application combines federated learning and privacy computing technologies: by cooperating with third-party platforms to train prediction models, there is no need to directly exchange user data, which not only protects user privacy, but also enables banks to indirectly use third-party data resources. During the model training and prediction stage, the security and privacy protection of user data are ensured, and the accuracy of data analysis is improved while meeting regulatory requirements. Using the optimized federated learning model, banks can more accurately predict user behavior and preferences, build detailed user portraits, and provide strong data support for marketing strategies.

[0096] In summary, this application not only breaks through the limitations of traditional methods in terms of data privacy and analysis accuracy, but also creates a new paradigm for data analysis and user portrait construction, significantly improving the effectiveness and efficiency of marketing activities.

[0097] It is also worth noting that in the field of cooperation between bank marketing and third-party channels, some solutions focus on data sharing architecture and customer behavior analysis. These solutions attempt to find a balance between data privacy protection and precision marketing. However, they still face certain limitations in data privacy protection and analysis accuracy. For example, API-based data sharing framework: Under this framework, banks and third-party platforms use application programming interfaces (APIs) to achieve data interaction. This approach usually involves transmitting some user information (such as behavior records or statistical overviews) to banks through secure API channels for user analysis and marketing strategy planning. Centralized data processing model: This model integrates data from different third-party channels into the bank's own data warehouse, aiming to provide a more comprehensive data overview. However, its effectiveness is often limited by the scope of data that third-party channels are willing to open. Application of traditional machine learning models: Banks use standard machine learning models to analyze data provided by cooperative channels, aiming to predict user behavior and optimize marketing strategies. These models are usually trained and deployed on the bank's internal servers. However, they can be inefficient and lack accuracy when processing large-scale, cross-channel data, especially when the dataset is incomplete.

[0098] This embodiment effectively solves the problems of data privacy protection, data sharing limitations, data integration complexity and real-time performance by introducing federated learning and privacy computing technologies, thereby enabling more secure and efficient cross-channel data analysis and user portrait construction.

[0099] Optionally, the optimized first prediction model is used to predict the user's behavior and preference to obtain prediction results, including:

[0100] The optimized first prediction model is used to predict the behavior and preferences of users who have not redeemed their rights on the target platform to obtain prediction results.

[0101] For those users who received benefits on a third-party platform but did not redeem them on the target platform, the optimized model is used to predict user portraits to more accurately identify potential customer groups.

[0102] The bank uses its own user transaction data, especially the information of users who have redeemed their benefits at the bank, to train a prediction model. This model aims to predict whether users will choose to redeem their benefits and further explore their other potential behavior patterns. For those user groups who have received benefits on third-party platforms but have not redeemed them at the bank, the bank will use the optimized model for in-depth analysis. This analysis process aims to reveal the unique characteristics and preferences of these users so that the bank can understand their behavior patterns more accurately and provide strong support for subsequent marketing strategies and service optimization.

[0103] Specifically, using the optimized first prediction model, we conducted in-depth prediction analysis of the behavior patterns and preference tendencies of users who have not yet redeemed their rights on the target platform. This prediction process relies on the model's powerful data processing capabilities and precise algorithm logic to carefully capture and analyze the user's behavior trajectory and potential needs.

[0104] Through this prediction, we obtained rich prediction results, which not only revealed the possible behavioral trends of users, but also deeply explored their personalized preferences. These valuable prediction data provided valuable insights, enabling a more accurate understanding of the real needs and expectations of these users, thus providing strong support for the subsequent marketing strategy formulation.

[0105] The improvement of this predictive capability not only breaks the information asymmetry, but also enables more accurate reach and service to these users, improving user satisfaction and loyalty, thereby promoting the sustainable and healthy development of banking business.

[0106] It should be pointed out that in the process of evaluating the effect of equity investment, banks face a significant challenge: it is difficult to accurately describe the characteristics of users who have not redeemed their equity through internal channels. This dilemma undoubtedly increases the uncertainty of marketing decisions, making strategy formulation like sailing in the fog. Furthermore, this information asymmetry seriously hinders banks from deeply exploring potential customer groups. Due to the inability to accurately identify and reach these users, banks' marketing investments often fail to obtain the expected returns, which not only wastes valuable resources, but also weakens banks' dominant position in the fierce market competition.

[0107] However, in this embodiment, for those users who have not redeemed their rights on the target platform, advanced prediction technology is used to conduct in-depth analysis of their behaviors and preferences. This move breaks down information barriers, allowing banks to more clearly understand the real needs and preferences of these users. Through this prediction capability, banks can formulate more accurate and effective marketing strategies to ensure that every marketing investment can accurately reach the target users, thereby significantly improving the return on marketing investment. This not only helps banks consolidate and expand their market share, but also lays a solid foundation for their future sustainable development.

[0108] Optionally, building a first prediction model on the target platform includes:

[0109] First, the first user behavior data of the user on the target platform is obtained; secondly, the logistic regression model is selected and trained using the gradient descent method, and parameter aggregation is performed to obtain the initial prediction model of the target platform; then, the secondary prediction model of the target platform is iteratively trained to obtain the final prediction model of the target platform; the secondary prediction model of the target platform is: synchronized with the parameters of the initial prediction model of the third-party platform, and the model is obtained by continuing to train according to the synchronized parameters.

[0110] Specifically, first, the first user behavior data of users on the target platform is comprehensively collected. These data cover various user interaction behaviors, browsing records, consumption habits, etc., providing a solid foundation for subsequent model training. More specifically, the first user behavior data is data from banks, including users' exchange records, transaction history, and account information. After collecting the data, the bank platform cleans, standardizes, and extracts features locally to ensure data consistency and quality.

[0111] Next, the logistic regression model was selected as the prediction tool and trained using the gradient descent method. During the training process, the model parameters were continuously adjusted to minimize the prediction error, and the training results were integrated into the initial prediction model of the target platform through parameter aggregation technology. This model has the initial ability to predict user behavior and preferences. Logistic regression is a model widely used in binary classification problems, suitable for predicting whether a user will redeem a certain benefit.

[0112] However, we did not stop there. In order to further improve the prediction accuracy of the model, the initial prediction model was iteratively trained to obtain the final prediction model of the target platform. In this iterative process, a cooperation mechanism with a third-party platform was introduced. Specifically, the parameters of the initial prediction model were synchronized with the prediction model of the third-party platform, and then training continued on the target platform based on these synchronized parameters. This approach not only fully utilized the prediction experience of the third-party platform, but also enabled the model to absorb more external information, thereby more accurately predicting the user's behavior and preferences on the target platform.

[0113] Specifically, Figure 2 As shown, it shows the training and application process of a prediction model:

[0114] (1) Data preparation.

[0115] Data from banks: including users’ exchange records, transaction history and account information.

[0116] Third-party platform data: covers user behavior data on third-party channels, such as benefits collection and browsing history.

[0117] (2) Model initialization: Initialize the weights (w) and biases (b) of the logistic regression model locally.

[0118] (3) Local model training.

[0119] Update the parameters of the logistic regression model using gradient descent on the local data.

[0120] The gradient descent formula, that is, the weight update formula is:

[0121]

[0122] Where α is the learning rate and J(w,b) is the log-likelihood loss function.

[0123] The log-likelihood loss function is:

[0124]

[0125] Where m is the number of samples, y (i) is the actual label, is the predicted probability.

[0126] (4) Parameter aggregation.

[0127] A central server is used to aggregate model parameters from all parties.

[0128] A simple averaging method is usually used to aggregate the model parameters uploaded by various parties.

[0129] (5) Model update and feedback.

[0130] The updated model parameters are distributed back to each participant.

[0131] Each party uses the new model parameters to train on local data and enter the next round of iteration.

[0132] (6) Iterative training.

[0133] Continue iterating training until the model's performance on the validation set reaches a preset standard or the iteration limit is reached.

[0134] (7) Model application.

[0135] The trained prediction model is used to predict the user's redemption probability for specific rights and interests and to predict other related behaviors.

[0136] Through the above process, federated learning technology enables banks and third-party platforms to jointly train logistic regression models while strictly protecting the privacy of user data. This innovative approach significantly improves the accuracy of cross-channel data analysis, providing strong support for banks to optimize their marketing strategies and enhance marketing effectiveness.

[0137] Through this series of construction and optimization steps, an efficient and accurate prediction model was established on the target platform, providing strong support for subsequent precision marketing and personalized services.

[0138] Optionally, before obtaining the parameters of the second prediction model on the third-party platform, the method further includes: constructing the second prediction model on the third-party platform.

[0139] Construct a second prediction model on a third-party platform, including:

[0140] First, the second user behavior data of the user on the third-party platform is obtained; secondly, the logistic regression model is selected to be trained using the gradient descent method, and parameter aggregation is performed to obtain the initial prediction model of the third-party platform; then, the secondary prediction model of the third-party platform is iteratively trained to obtain the final prediction model of the third-party platform; the secondary prediction model of the third-party platform is: synchronized with the parameters of the initial prediction model of the target platform, and the model obtained by continuing to train according to the synchronized parameters.

[0141] Before formally obtaining the second prediction model parameters on the third-party platform, a complete second prediction model is first built on the third-party platform. This process is specifically divided into the following steps:

[0142] First, we extensively collect the second user behavior data of users on third-party platforms. These data record in detail the key information such as various operations, browsing history, transaction records, etc. of users on third-party platforms, providing valuable materials for subsequent model training. The second user behavior data is third-party platform data, covering the behavior data of users on third-party platforms, such as rights collection and browsing records. The third-party platform cleans, standardizes and extracts features of the data locally to ensure the consistency and quality of the data.

[0143] Next, we selected the logistic regression model as the prediction tool and applied the gradient descent method for training. During the training process, we continuously adjusted and optimized the model parameters to minimize the prediction error, and integrated the training results into the first prediction model of the third-party platform through parameter aggregation technology. This model has the initial ability to predict user behavior and preferences on third-party platforms.

[0144] However, in order to further improve the prediction performance of the model, training was not stopped. Instead, the initial prediction model was iteratively trained to obtain the final prediction model for the third-party platform. In this iterative process, an innovative strategy was adopted: the parameters of the initial prediction model of the target platform were synchronized with the initial prediction model of the third-party platform. Based on these synchronized parameters, training continued on the third-party platform, resulting in a more accurate and stable prediction model. This approach not only fully utilizes the complementarity between the target platform and the third-party platform, but also enables the model to absorb more cross-platform information, thereby more accurately predicting user behaviors and preferences on different platforms.

[0145] The training process of the third-party platform is similar to that of the target platform mentioned above, and will not be described in detail here.

[0146] Through the above steps, an efficient and accurate prediction model was successfully built on the third-party platform, laying a solid foundation for subsequent parameter acquisition and model optimization.

[0147] In other words, with the help of federated learning technology, banks can collaborate with third-party platforms to train machine learning models locally without directly transmitting user data. Under the framework of federated learning, data privacy is effectively protected because only model parameters (such as weights and biases) are circulated between the participants, thus ensuring that the user's original data will not be leaked.

[0148] The forecasting model will be regularly iterated and optimized based on new data and market feedback to maintain its forecast accuracy.

[0149] In summary, on the basis of protecting user data privacy, distributed data resources are effectively utilized, and the data analysis efficiency and accuracy of the target platform's cross-channel marketing activities are improved. Through this method, the target platform can more accurately locate the target customer group, optimize the allocation of marketing resources, and improve the overall effect of marketing activities.

[0150] In this embodiment, federated learning and privacy computing technologies are used to promote the target platform and the third party to jointly train the model under the premise of protecting data security, and realize cross-platform data security interaction and sharing of model parameters. This method enables banks to collaborate with third parties without directly sharing original user data, train their respective local models, and optimize their respective models by updating model parameters. The use of privacy computing technology ensures security and privacy protection during cross-platform data transmission. At the same time, the introduction of third-party data enhances the comprehensiveness and integrity of user behavior data, thereby improving the accuracy of user portraits, laying a solid foundation for the implementation of precision marketing strategies.

[0151] Optionally, the first behavior data includes: the user's redemption record and transaction record;

[0152] The second behavior data includes: rights and interests field records and browsing records.

[0153] The first user behavior data mainly covers the user's redemption records and transaction records on the target platform. These records reflect the user's actual redemption behavior and transaction history on the platform in detail, providing an important basis for analyzing the user's consumption habits and preferences; of course, the first behavior data can also include other data, which will not be detailed here.

[0154] The second user behavior data mainly focuses on the user's rights and interests records and browsing records on third-party platforms. Rights and interests records reveal the user's interest and participation in different types of rights and interests, while browsing records reflect the user's information acquisition and focus on the platform. These data can provide valuable information for in-depth understanding of the user's rights and interests preferences and browsing habits; of course, the second behavior data can also include other data, which will not be listed here one by one.

[0155] By integrating and analyzing these two types of behavioral data, we can have a more comprehensive understanding of user behaviors and preferences on different platforms, thereby providing strong support for subsequent predictive model construction and marketing strategy formulation.

[0156] Optionally, privacy computing technologies include homomorphic encryption and secure multi-party computing.

[0157] Apply privacy computing technologies, such as homomorphic encryption or secure multi-party computing, to ensure the security and privacy of user data during model training and prediction.

[0158] Privacy computing technology mainly covers two core areas: homomorphic encryption and secure multi-party computing.

[0159] Homomorphic encryption is an advanced encryption method that allows specific mathematical operations to be performed on encrypted data and obtain an encrypted result that is the same as the plaintext operation without first decrypting the data. This feature ensures the security of data during processing, allowing data to be analyzed and used in an encrypted state, thereby effectively protecting the privacy of users.

[0160] Secure multi-party computation is a technology that allows multiple participants to jointly compute a function without exposing their input data. It ensures that each participant can only obtain the final calculation result and cannot know the input data of other participants, thus achieving secure data sharing and collaborative computing.

[0161] By combining homomorphic encryption and secure multi-party computing, privacy computing technology provides more comprehensive and effective security protection for data processing and analysis. It can make full use of data resources and promote the development of data science and artificial intelligence while protecting user privacy.

[0162] It should be noted that in the model training and prediction stages, federated learning incorporates homomorphic encryption and secure multi-party computing technologies to ensure absolute security and privacy protection of data during transmission and processing. Under the framework of federated learning, the two pillars of privacy computing technology - homomorphic encryption and secure multi-party computing (SMC) - aim to protect data privacy while promoting the effective use of data.

[0163] In the federated learning scenario, each participant (target platform and third platform) first uses homomorphic encryption technology to encrypt local data. The encrypted data is then used to train the local model, or sent to the central server for aggregate analysis when necessary. Thanks to the characteristics of homomorphic encryption, all operations are performed directly on the encrypted data, ensuring that sensitive information is not leaked at any stage. Taking the Paillier encryption system as an example, its encryption process can be expressed as:

[0164] E(m)=g m ·r n mod n 2

[0165] Among them, m represents the plaintext message, g and n form part of the public key, and r is a randomly generated number. This encryption mechanism ensures the privacy of data while supporting specific mathematical operations in the encryption domain.

[0166] In federated learning, each participant is responsible for calculating part of the model update, such as gradients or model parameters. These calculation results are shared between participants through a carefully designed security protocol without disclosing their original data or specific calculation details. Taking additive secret sharing as an example, a value x is split into multiple parts:

[0167] x=x1+x2+…+x n

[0168] Among them, x i is allocated to the i-th party. Each participant only knows the share x held by himself. i , but knows nothing about other shares or the sum x. This mechanism effectively protects the privacy of data and promotes collaborative computing among multiple parties.

[0169] In summary, privacy computing technology provides an innovative way for federated learning to perform data analysis and model training without directly accessing or sharing the original data. This method not only greatly protects data privacy, but also significantly improves the security of data analysis, laying a solid foundation for the widespread application of federated learning.

[0170] like Figure 3 As shown, the following is an example of the application of privacy computing in the marketing method combined with a third-party platform:

[0171] Suppose Bank A wants to analyze the behavior patterns of users who receive benefits on third-party platform B but do not redeem them at the bank. Bank A has its own customer data, while third-party platform B has data on user behavior, but due to privacy protection policies, platform B is reluctant to directly share detailed information about users.

[0172] 1) Initialization phase:

[0173] Bank A and third-party platform B each select a specific machine learning model (such as logistic regression) to analyze user behavior. In order to protect data privacy, both parties agree to use privacy computing technology, such as homomorphic encryption, to encrypt and process their respective data.

[0174] 2) Data encryption:

[0175] Bank A encrypts its customer data, including users’ exchange records and transaction history. Meanwhile, Platform B encrypts its user behavior data, such as benefits collection and browsing history.

[0176] 3) Local model training:

[0177] On their respective servers, Bank A and Platform B use encrypted data to train the selected model. During the training process, the data remains encrypted at all times, which means that sensitive information will not be leaked even during the training process.

[0178] 4) Model parameter exchange:

[0179] After training is complete, Bank A and Platform B exchange the model parameters through a secure protocol. These parameters are encrypted and do not leak any sensitive information. Parameter exchange allows both parties to obtain a more comprehensive model that reflects the user behavior characteristics of both the bank and the third-party platform.

[0180] 5) Model fusion and optimization:

[0181] After receiving each other's model parameters, Bank A and Platform B each fuse these parameters locally and optimize their own models. The optimized models can more accurately predict user behavior and preferences.

[0182] 6) Application and analysis:

[0183] Bank A uses the optimized model to analyze the user group that has not redeemed the benefits at the bank to discover potential market opportunities or optimize marketing strategies. Since the data is always encrypted throughout the process, the security of user privacy is guaranteed.

[0184] It can be seen that the application of privacy computing in this patent enables banks to effectively conduct cross-channel data analysis without directly accessing or sharing sensitive user data, while strictly protecting user privacy.

[0185] In this embodiment, relying on federated learning and privacy computing technology, the sharing of model parameters between banks and third-party platforms is achieved instead of directly exchanging raw data, thereby improving the efficiency of cross-channel data utilization while strictly protecting the privacy and security of user data, and effectively reducing the risk of data leakage. At the same time, by integrating the data resources of banks and third-party platforms, the model constructed by the present invention can gain a deeper insight into user behavior and preferences, providing banks with more accurate user behavior analysis. This further enhances the ability of banks to predict user needs and behaviors, enabling them to formulate marketing strategies that are more in line with user needs. In addition, based on the accurate analysis of user behavior, this application helps banks achieve the optimal allocation of marketing resources. By accurately locking in the user groups most likely to respond, banks can ensure the efficient execution of marketing activities, which not only significantly improves the input-output ratio of marketing activities, but also significantly reduces the costs caused by ineffective marketing. This change not only optimizes marketing strategies, but also profoundly affects the decision-making quality of product development and customer service.

[0186] Another embodiment of the present application provides a marketing system combined with a third-party platform.

[0187] See also Figure 4 , the marketing system combined with the third-party platform includes:

[0188] Construction module 101 is used to construct a first prediction model on the target platform; the first prediction model is a model obtained by encrypting the first behavior data of the target platform using privacy computing technology and training based on the encrypted first behavior data; the first prediction model is used to predict the user's behavior and preference based on the user's behavior data.

[0189] The exchange optimization module 102 is used to obtain the parameters of the second prediction model of the third-party platform and optimize the first prediction model according to the parameters of the second prediction model; the second prediction model is trained by the third-party platform according to the encrypted second behavior data.

[0190] The prediction module 103 is used to use the optimized first prediction model to predict the user's behavior and preference to obtain a prediction result; and optimize the marketing strategy according to the prediction result.

[0191] Optionally, the prediction module 103 is used to use the optimized first prediction model to predict the user's behavior and preference, and when the prediction result is obtained, it is specifically used to:

[0192] The optimized first prediction model is used to predict the behavior and preferences of users who have not redeemed their rights on the target platform to obtain prediction results.

[0193] For details on the specific working process and principle of each of the above modules, please refer to the marketing method combined with the third-party platform provided in the above embodiment, which will not be described here one by one. It depends on the actual situation and is within the scope of protection of this application.

[0194] In this embodiment, the construction module 101 constructs a first prediction model on the target platform; the first prediction model is a model obtained by encrypting the first behavior data of the target platform using privacy computing technology and training based on the encrypted first behavior data; the first prediction model is used to predict the user's behavior and preference based on the user's behavior data; the exchange optimization module 102 obtains the parameters of the second prediction model of the third-party platform, and optimizes the first prediction model based on the parameters of the second prediction model; the second prediction model is obtained by training the third-party platform based on the encrypted second behavior data; the prediction module 103 uses the optimized first prediction model to predict the user's behavior and preference to obtain a prediction result; according to the prediction result, the marketing strategy is optimized; that is, by obtaining the model parameters of the third-party platform, the user is fully understood, and favorable data support is provided for the formulation of marketing strategies, thereby improving the effectiveness of the strategy and the efficiency of resource utilization.

[0195] Another embodiment of the present application provides a storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the marketing method combined with a third-party platform as in any one of the above embodiments is implemented.

[0196] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, 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 foregoing.

[0197] It should be noted that the computer-readable medium disclosed above 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 disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that may be used by or in combination with an instruction execution system, device or device. In the present disclosure, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, in which a computer-readable program code is carried. This propagated data signal may take a variety of forms, including but not limited to an electromagnetic signal, an optical signal, or any suitable combination of the above. The computer readable signal medium may also be any computer readable medium other than a computer readable storage medium, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer readable medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (radio frequency), etc., or any suitable combination of the above.

[0198] The computer-readable medium may be included in the electronic device, or may exist independently without being incorporated into the electronic device.

[0199] Another embodiment of the present invention provides an electronic device, such as Figure 5 As shown, including:

[0200] One or more processors 201.

[0201] The storage device 202 stores one or more programs.

[0202] When one or more programs are executed by one or more processors 201, one or more processors 201 implement a marketing method combining a third-party platform as in any one of the above embodiments.

[0203] In particular, according to an embodiment of the present disclosure, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present disclosure includes a computer program product, which includes a computer program carried on a non-transitory computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart.

[0204] Although the subject matter has been described in language specific to structural features and / or methodological logical actions, it should be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or actions described above. On the contrary, the specific features and actions described above are merely example forms of implementing the claims.

[0205] Although several specific implementation details are included in the above discussion, these should not be interpreted as limiting the scope of the present disclosure. Certain features described in the context of a separate embodiment can also be implemented in a single embodiment in combination. On the contrary, the various features described in the context of a single embodiment can also be implemented in multiple embodiments individually or in any suitable sub-combination.

[0206] The above description is only a preferred embodiment of the present disclosure and an explanation of the technical principles used. Those skilled in the art should understand that the scope of disclosure involved in the present disclosure is not limited to the technical solutions 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 above disclosed concept. For example, the above features are replaced with the technical features with similar functions disclosed in the present disclosure (but not limited to) by each other to form a technical solution.

[0207] The features recorded in the various embodiments in this specification can be replaced or combined with each other, and the same and similar parts between the various embodiments can refer to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system or system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can refer to the partial description of the method embodiment. The system and system embodiments described above are only schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative work.

[0208] Professionals may further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in the above description according to function. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians may use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0209] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A marketing method combining a third-party platform, characterized in that: include: Building a first prediction model on the target platform; The first prediction model is a model obtained by encrypting the first behavior data of the target platform using privacy computing technology and training based on the encrypted first behavior data; the first prediction model is used to predict the user's behavior and preference based on the user's behavior data; Obtaining parameters of a second prediction model of a third-party platform, and optimizing the first prediction model according to the parameters of the second prediction model; the second prediction model is obtained by training the third-party platform according to the encrypted second behavior data; The optimized first prediction model is used to predict the user's behavior and preference to obtain a prediction result; Optimize marketing strategies based on the prediction results.

2. The marketing method combining a third-party platform according to claim 1, characterized in that: The optimized first prediction model is used to predict the user's behavior and preference to obtain a prediction result, including: The optimized first prediction model is used to predict the behavior and preferences of users who have not redeemed their rights on the target platform to obtain prediction results.

3. The marketing method combining a third-party platform according to claim 1, characterized in that: The step of constructing a first prediction model on the target platform includes: Acquire first user behavior data of the user on the target platform; A logistic regression model is selected to be trained using a gradient descent method, and parameter aggregation is performed to obtain an initial prediction model of the target platform; The secondary prediction model of the target platform is iteratively trained to obtain a final prediction model of the target platform; the secondary prediction model of the target platform is a model obtained by synchronizing parameters with the primary prediction model of the third-party platform and continuing to train according to the synchronized parameters.

4. The marketing method combining a third-party platform according to claim 1, characterized in that: Before obtaining the parameters of the second prediction model on the third-party platform, the method further includes: constructing the second prediction model on the third-party platform; the constructing the second prediction model on the third-party platform includes: Obtaining second user behavior data of the user on the third-party platform; A logistic regression model is selected and trained using a gradient descent method, and parameter aggregation is performed to obtain an initial prediction model of the third-party platform; The secondary prediction model of the third-party platform is iteratively trained to obtain a final prediction model of the third-party platform; the secondary prediction model of the third-party platform is a model obtained by synchronizing parameters with the initial prediction model of the target platform and continuing to train according to the synchronized parameters.

5. The marketing method combining a third-party platform according to claim 1, characterized in that: The first behavior data includes: the user's redemption records and transaction records; The second behavior data includes: rights and interests field records and browsing records.

6. The marketing method combining a third-party platform according to claim 1, characterized in that: The privacy computing technology includes: homomorphic encryption and secure multi-party computing.

7. A marketing system combined with a third-party platform, characterized in that: include: A construction module, used for constructing a first prediction model on a target platform; The first prediction model is a model obtained by encrypting the first behavior data of the target platform using privacy computing technology and training based on the encrypted first behavior data; the first prediction model is used to predict the user's behavior and preference based on the user's behavior data; An exchange optimization module, used to obtain parameters of a second prediction model of a third-party platform, and optimize the first prediction model according to the parameters of the second prediction model; the second prediction model is obtained by training the third-party platform according to the encrypted second behavior data; The prediction module is used to use the optimized first prediction model to predict the user's behavior and preference to obtain a prediction result; and optimize the marketing strategy according to the prediction result.

8. The marketing system combined with a third-party platform according to claim 7, characterized in that: The prediction module is used to use the optimized first prediction model to predict the user's behavior and preference, and when the prediction result is obtained, it is specifically used to: The optimized first prediction model is used to predict the behavior and preferences of users who have not redeemed their rights on the target platform to obtain prediction results.

9. An electronic device, characterized in that: include: one or more processors; a storage device having one or more programs stored thereon; When one or more programs are executed by one or more processors, the one or more processors implement the marketing method combined with a third-party platform as described in any one of claims 1 to 6.

10. A storage medium, characterized in that: A computer program is stored thereon, wherein when the computer program is executed by a processor, the marketing method combined with a third-party platform as described in any one of claims 1-6 is implemented.

Images