A marketing resource allocation method and apparatus

Abstract

The application provides a marketing resource allocation method and device, which can be used in the field of artificial intelligence technology, and the method comprises the following steps: through a conversion estimation model, the core feature set, the stored product basic attribute, the marketing binary classification label and the purchase binary classification label are subjected to probability prediction to generate a basic purchase probability and a marketing intervention purchase probability, and the marketing sensitivity level of a user is determined; based on the marketing sensitivity level of the user, a marketing resource allocation scheme is constructed, precise calculation and grading based on individual causal incremental effect are realized in the marketing resource allocation, the sensitivity of the user to the marketing intervention is accurately quantified, and effective stratification of the user group is realized; meanwhile, the built-in dynamic features and model iteration mechanism guarantee the robustness and adaptability of the algorithm in the continuously changing business environment from the system level, form an intelligent decision-making closed loop that can be self-optimized, ensure the dynamic adaptability and long-term effectiveness of the marketing strategy, and optimize the user experience while improving the conversion effect.

Description

Technical Field

[0001] This invention relates to the field of computer technology, particularly to the field of artificial intelligence technology, and especially to a marketing resource allocation method and apparatus. Background Technology

[0002] In traditional financial product and service marketing, resources are typically allocated based on static characteristics such as users' historical transaction records, asset size, and demographic attributes. Rule engines or simple predictive models are used to filter out "potential buyers" for resource allocation. However, this approach often has significant limitations: on the one hand, it tends to concentrate marketing resources on users with high purchasing intent or who already meet the purchasing criteria, leading to an overemphasis on "naturally converted" users and failing to effectively market to "sensitive users" who are genuinely influenced by marketing activities to make purchases. On the other hand, due to the lack of quantitative evaluation of the incremental effects of marketing interventions, traditional models struggle to distinguish whether user purchasing behavior stems from marketing stimuli or their own needs, resulting in misallocation and waste of marketing resources and a low overall return on investment (ROI).

[0003] With the development of artificial intelligence and big data technologies, although some studies have attempted to introduce machine learning models for user response prediction, they still generally face problems such as redundant feature systems, poor model interpretability, and insufficient dynamic adaptability. In particular, there is a lack of a systematic causal inference mechanism to identify the real gains of marketing, making it difficult to continuously achieve efficient resource allocation and optimized user experience in complex and ever-changing financial marketing scenarios. Summary of the Invention

[0004] One objective of this invention is to provide a marketing resource allocation method. By constructing a complete technical framework of "feature engineering screening - two-layer probability prediction - sensitivity quantification," it achieves accurate measurement and grading based on individual causal incremental effects in marketing resource allocation, precisely quantifying users' sensitivity to marketing interventions and effectively segmenting user groups. Simultaneously, its built-in dynamic feature and model iteration mechanism ensures the robustness and adaptability of the algorithm in continuously changing business environments at the system level, forming a self-optimizing intelligent decision-making closed loop. This ensures the dynamic adaptability and long-term effectiveness of marketing strategies, while improving conversion rates and optimizing user experience. Another objective of this invention is to provide a marketing resource allocation device. A further objective is to provide a computer-readable medium. A final objective is to provide a computer device.

[0005] To achieve the above objectives, this invention discloses a marketing resource allocation method, comprising: Obtain the core feature set of users from historical marketing campaigns; The trained conversion prediction model performs probability prediction on the core feature set, stored product basic attributes, marketing binary tags, and purchase binary tags to generate basic purchase probability and marketing intervention purchase probability. Determine the user's marketing sensitivity level based on the base purchase probability and the purchase probability after marketing intervention; Based on users' marketing sensitivity levels, a marketing resource allocation plan is constructed.

[0006] Preferably, the core feature set of users from historical marketing campaigns is obtained, including: Collect multi-dimensional user data from historical marketing campaigns; By linking and integrating multi-dimensional data according to user identifiers and timestamps, a standardized dataset is obtained. Standardized data is processed using feature engineering to generate a core feature set.

[0007] Preferably, the standardized data undergoes feature engineering to generate a core feature set, including: Extract features from standardized data to generate standardized features; The core feature set is selected from the standardized features using an ensemble tree model.

[0008] Preferably, the marketing category labels include either "marketed" labels or "unmarketed" labels; The trained conversion prediction model performs probability predictions on the core feature set, stored product basic attributes, marketing binary classification tags, and purchase binary classification tags to generate basic purchase probabilities and marketing intervention purchase probabilities, including: By using a conversion prediction model, probability predictions are made on marketing tags, core feature sets, basic product attributes, and purchase binary tags to generate marketing intervention purchase probabilities. By using a conversion prediction model, probability predictions are made for non-marketing tags, core feature sets, basic product attributes, and purchase binary tags to generate basic purchase probabilities.

[0009] Preferably, the user's marketing sensitivity level is determined based on the base purchase probability and the marketing intervention purchase probability, including: The difference between the base purchase probability and the purchase probability after marketing intervention is defined as user sensitivity. Based on preset sensitivity thresholds, user marketing sensitivity levels are categorized according to user sensitivity.

[0010] Preferably, the method further includes: The steps of collecting multi-dimensional user data from historical marketing campaigns are repeated at preset time intervals.

[0011] Preferably, after constructing a marketing resource allocation plan based on users' marketing sensitivity levels, it also includes: To obtain the actual purchase rate of users with high marketing sensitivity; If the predicted deviation between the actual purchase rate and the corresponding marketing intervention purchase probability is greater than the preset deviation threshold, the step of collecting multi-dimensional data of users from historical marketing campaigns will be repeated.

[0012] Preferably, after constructing a marketing resource allocation plan based on users' marketing sensitivity levels, it also includes: Collect multi-dimensional user data from the latest historical marketing campaigns; Based on the latest multi-dimensional user data from historical marketing campaigns, the integrated tree model and conversion prediction model are iteratively updated to generate the updated integrated tree model and conversion prediction model.

[0013] The present invention also discloses a marketing resource allocation device, comprising: The core feature set acquisition unit is used to acquire the core feature set of users in historical marketing campaigns. The probability prediction unit is used to perform probability prediction on the core feature set, stored product basic attributes, marketing binary labels and purchase binary labels through the trained conversion prediction model, and generate basic purchase probability and marketing intervention purchase probability. The sensitivity level determination unit is used to determine the user's marketing sensitivity level based on the basic purchase probability and the marketing intervention purchase probability; The resource allocation unit is used to build marketing resource allocation schemes based on users' marketing sensitivity levels.

[0014] The present invention also discloses a computer-readable medium having a computer program stored thereon, which, when executed by a processor, implements the method described above.

[0015] The present invention also discloses a computer device, including a memory and a processor, wherein the memory is used to store information including program instructions, and the processor is used to control the execution of the program instructions, wherein the processor executes the program to implement the method described above.

[0016] The present invention also discloses a computer program product, including a computer program / instruction, which, when executed by a processor, implements the method described above.

[0017] This invention acquires the core feature set of users from historical marketing campaigns; through a trained conversion prediction model, it performs probability prediction on the core feature set, stored product basic attributes, marketing binary tags, and purchase binary tags to generate basic purchase probability and marketing intervention purchase probability; based on the basic purchase probability and marketing intervention purchase probability, it determines the user's marketing sensitivity level; based on the user's marketing sensitivity level, it constructs a marketing resource allocation scheme. By constructing a complete technical framework of "feature engineering screening - two-layer probability prediction - sensitivity quantification," it achieves accurate calculation and grading based on individual causal incremental effects in marketing resource allocation, accurately quantifies the user's sensitivity to marketing intervention, and achieves effective stratification of user groups; at the same time, its built-in dynamic feature and model iteration mechanism ensures the robustness and adaptability of the algorithm in continuously changing business environments from a system level, forming a self-optimizing intelligent decision-making closed loop, ensuring the dynamic adaptability and long-term effectiveness of marketing strategies, and optimizing user experience while improving conversion results. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 A flowchart of a marketing resource allocation method provided in an embodiment of the present invention; Figure 2 A flowchart illustrating another marketing resource allocation method provided in this embodiment of the invention; Figure 3 This is a schematic diagram of the structure of a marketing resource allocation device provided in an embodiment of the present invention; Figure 4 This is a schematic diagram of the structure of a computer device provided in an embodiment of the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] It should be noted that the marketing resource allocation method and apparatus disclosed in this application can be used in the field of artificial intelligence technology, or in any field other than artificial intelligence technology. The application field of the marketing resource allocation method and apparatus disclosed in this application is not limited.

[0022] To facilitate understanding of the technical solution provided in this application, the relevant content of the technical solution will be explained below. A general marketing gain model is constructed based on historical business data of financial institutions and comprehensive user behavior data. The core innovations include three aspects: First, a multi-dimensional feature engineering framework for the entire business is proposed. From the dimensions of user asset size, business holding period, transaction frequency, risk preference, and past marketing response records, 12 categories of 50 key features are extracted. By filtering the importance of model features, the core feature subset that is sensitive to financial marketing is accurately identified, thus solving the identification bias problem caused by feature redundancy in traditional models.

[0023] Second, a two-layer gain prediction architecture is designed. The bottom layer uses a conversion prediction model (such as XGBoost, NN model, etc.) to build a basic user business willingness prediction module. The upper layer introduces a causal algorithm to quantify the incremental impact of marketing intervention on different users (Uplift), breaking through the limitation of traditional models that can only predict "whether to conduct business" but cannot determine "whether the marketing is effective".

[0024] Third, a dynamic feature update mechanism will be developed, combining real-time user data from financial institutions (such as account fund changes, financial product browsing behavior, and business consultation records) to automatically iterate feature weights and model parameters, ensuring rapid adaptation to market fluctuations and changes in user needs. Ultimately, this will enable marketing resources to be precisely allocated to high-uplift users, improving the conversion rate and revenue of marketing across all categories of financial services, reducing disruption to non-marketing-sensitive users, and simultaneously optimizing user experience and brand favorability. This will allow for the accurate identification and optimized allocation of resources to reach sensitive users in financial product and service marketing.

[0025] The following uses a marketing resource allocation device as an example to illustrate the implementation process of the marketing resource allocation method provided in this embodiment of the invention. It is understood that the execution subject of the marketing resource allocation method provided in this embodiment of the invention includes, but is not limited to, a marketing resource allocation device.

[0026] Figure 1 A flowchart of a marketing resource allocation method provided in an embodiment of the present invention is shown below. Figure 1 As shown, the method includes: Step 101: Obtain the core feature set of users from historical marketing campaigns.

[0027] In this embodiment of the invention, a multi-dimensional data acquisition module collects user attribute data, behavioral interaction data, and marketing activity data from the historical business database of financial institutions. This raw data is correlated and aligned using user identifiers and timestamps as key dimensions to form standardized data. Through a feature engineering processing module, based on business logic and domain knowledge, an initial feature pool is constructed from the standardized data. To overcome the curse of dimensionality and noise interference, an ensemble tree model (such as LightGBM) is used for feature importance assessment and selection. The ensemble tree model automatically selects the core feature set with the highest discriminative power by evaluating the contribution of each feature to predicting historical marketing responses. This process achieves the transformation from massive amounts of raw data to a high-information-density, low-redundancy core feature set.

[0028] This step transforms multi-source, heterogeneous raw data into a core feature set that accurately characterizes users' marketing sensitivity potential through systematic data integration and automated feature selection based on machine learning. This solves the problems of traditional methods relying on human experience, feature redundancy, and poor stability, providing high-quality, interpretable input for subsequent accurate predictions and forming the foundation for improving the overall system's discriminative performance.

[0029] Step 102: Using the trained conversion prediction model, perform probability prediction on the core feature set, stored product basic attributes, marketing binary classification tags, and purchase binary classification tags to generate basic purchase probability and marketing intervention purchase probability.

[0030] In this embodiment of the invention, historical data is used, wherein each sample includes: a core feature set, basic product attributes stored in the database, marketing binary tags, and purchase binary tags. The marketing binary tags indicate whether the user actually received marketing outreach in historical activities, and the purchase binary tags indicate whether the user ultimately purchased the product. The marketing binary tags include a "marketed" tag (1) or a "not marketed" tag (0), and the purchase binary tags include a "purchased" tag (1) or a "not purchased" tag (0).

[0031] This invention utilizes causal inference to estimate the purchase probability of the same user in two parallel scenarios. A conversion prediction model (such as XGBoost) is selected as the base learner. The training objective is to learn the mapping relationship from "features + marketing status" to "purchase outcome". To obtain unbiased estimates, training is conducted on historical randomized experimental data.

[0032] In this embodiment of the invention, the core feature set and product attributes of a user are input into the model, and its marketing binary classification label is set as the non-marketing label (0). The purchase probability output by the model is the user's natural purchase intention without external marketing intervention, that is: the basic purchase probability (P0).

[0033] In this embodiment of the invention, the same characteristics of the same user are input into the model, but the marketing binary classification label is set as the marketing label (1). The purchase probability output by the model is the user's estimated purchase intention after being subject to marketing intervention, that is: marketing intervention purchase probability (P1).

[0034] This step overcomes the limitation of traditional response models, which can only predict "whether a user will buy," by performing counterfactual reasoning twice on the same user, quantifying the net impact of marketing intervention on the probability of purchase. This provides a direct and quantitative technical basis for accurately measuring the "incremental value" of marketing in the future.

[0035] Step 103: Determine the user's marketing sensitivity level based on the basic purchase probability and the marketing intervention purchase probability.

[0036] In this embodiment of the invention, the difference between the base purchase probability and the marketing intervention purchase probability is defined as user sensitivity. Based on the distribution of user sensitivity and business objectives, a sensitivity threshold is set for user segmentation. High marketing sensitivity level: User sensitivity exceeds the first sensitivity threshold α. This type of user marketing intervention can bring significant positive incremental results and is a core target that marketing resources should prioritize.

[0037] No / Low Marketing Sensitivity Level: Users' sensitivity is less than the second sensitivity threshold β. Marketing to these users is ineffective and may even cause resentment; resources should be avoided for them.

[0038] Medium Marketing Sensitivity Level: User sensitivity falls between the first sensitivity threshold α and the second sensitivity threshold β. Appropriate resources may be allocated, or more detailed strategy testing may be conducted.

[0039] This step represents a leap from "probability prediction" to "causal effect classification." By classifying users based on their sensitivity, it is possible to clearly identify the true beneficiaries of marketing campaigns and proactively avoid reaching ineffective groups. This lays the foundation for refined and differentiated marketing strategies and directly drives ROI improvement.

[0040] Step 104: Based on users' marketing sensitivity levels, construct a marketing resource allocation plan.

[0041] In this embodiment of the invention, users with high marketing sensitivity are prioritized for high-cost but high-reach channels, such as one-on-one contact with product managers, exclusive app pop-ups, or VIP privilege channels; users with marketing sensitivity are allocated cost-effective standardized channels, such as targeted SMS, social media community operations, or light advertising recommendations; and users with no or low marketing sensitivity are protected by adding them to a do-not-disturb list to save marketing costs while improving the overall user experience.

[0042] Furthermore, after the allocation plan is implemented, the dynamic update control module will collect a new round of marketing feedback data (such as the actual purchase rate of users at each level). If it is found that the actual conversion rate of highly sensitive users deviates too much from the predicted value, or if the market environment changes significantly, the module will automatically trigger iterative updates of feature engineering and prediction models, so that the entire system forms a closed loop of "decision-execution-feedback-optimization".

[0043] This step automates and automates the allocation of marketing resources, ensuring that limited marketing budgets and channel resources are allocated to the most productive user groups, maximizing the incremental benefits of the overall marketing campaign. Simultaneously, the closed-loop optimization mechanism guarantees that the allocation strategy continuously evolves with user behavior and market dynamics, maintaining its effectiveness and adaptability in the long term.

[0044] The technical solution provided in this invention involves obtaining a core feature set of users from historical marketing activities; using a trained conversion prediction model, probabilistic predictions are made on the core feature set, stored product basic attributes, marketing binary classification tags, and purchase binary classification tags to generate a basic purchase probability and a marketing intervention purchase probability; based on the basic purchase probability and the marketing intervention purchase probability, the user's marketing sensitivity level is determined; based on the user's marketing sensitivity level, a marketing resource allocation scheme is constructed. By constructing a complete technical framework of "feature engineering screening - two-layer probability prediction - sensitivity quantification," accurate calculation and grading based on individual causal incremental effects are achieved in marketing resource allocation, accurately quantifying the user's sensitivity to marketing intervention and realizing effective stratification of user groups; simultaneously, its built-in dynamic feature and model iteration mechanism ensures the robustness and adaptability of the algorithm in a continuously changing business environment from a system level, forming a self-optimizing intelligent decision-making closed loop, ensuring the dynamic adaptability and long-term effectiveness of the marketing strategy, and optimizing the user experience while improving conversion results.

[0045] Figure 2 A flowchart of another marketing resource allocation method provided in an embodiment of the present invention is shown below. Figure 2 As shown, the method includes: Step 201: Collect multi-dimensional data of users from historical marketing campaigns.

[0046] In this embodiment of the invention, each step is performed by a marketing resource allocation device.

[0047] In this embodiment of the invention, multi-dimensional data includes, but is not limited to, basic attribute data, asset data, transaction behavior data, marketing interaction data, and real-time dynamic data. Basic attribute data includes, but is not limited to, authorized personal information such as age and occupation; asset data includes, but is not limited to, asset size and amount of wealth management holdings; transaction behavior data includes, but is not limited to, the frequency, amount, cycle, and product type of historical financial transactions such as purchases, redemptions, and transfers; marketing interaction data includes, but is not limited to, past marketing response records, marketing intervention records, and service contact records; real-time dynamic data includes, but is not limited to, account fund changes and product browsing time.

[0048] Transaction behavior data and marketing interaction data are key time-series data reflecting user interests and intentions; marketing response records are direct user feedback on marketing activities, such as immediate response behaviors like link clicks, coupon redemption, and dwell time on activity pages, as well as whether a purchase conversion ultimately occurs (i.e., the basis for generating subsequent purchase binary categories). Specifically, this includes in-depth interaction behaviors such as browsing time, number of clicks, search keywords, and use of financial calculators on specific financial product detail pages through channels such as financial institution applications (APPs); marketing intervention records specifically include the user list for each marketing campaign, the time of reach, the channels used (such as SMS, APP push, telephone, product manager face-to-face interviews), and whether the user actually received confirmation of the reach (i.e., the basis for generating subsequent marketing binary categories); service contact records are non-transactional interaction information such as communication records with product managers and telephone customer service consultation content.

[0049] In this embodiment of the invention, the data format of multi-dimensional data covers structured data (such as asset amount values) and semi-structured data (such as marketing response logs).

[0050] This invention provides a high-quality, high-coverage data foundation for the entire solution by designing a systematic multi-dimensional data acquisition framework, ensuring the richness of subsequent feature extraction and the reliability of model training, and providing a prerequisite technical guarantee for achieving accurate user insights and scientific incremental evaluation.

[0051] Step 202: Link and integrate the multi-dimensional data according to the user ID and timestamp dimensions to obtain a standardized dataset.

[0052] In this embodiment of the invention, multi-dimensional data often suffers from problems such as inconsistent formats, asynchronous time, and loose relationships. To achieve effective analysis, user ID-timestamp is used as a unique identifier, and scattered data is integrated into a standardized dataset through data joining (JOIN), alignment, deduplication, and cleaning operations.

[0053] For example, it involves linking a user's transaction records at a specific point in time, marketing outreach records received before and after that point, and a snapshot of the user's attributes at that point. Furthermore, it includes data standardization operations, such as standardizing the units of numerical variables, standardizing categorical variables (e.g., One-Hot coding), and handling missing and outlier values, ultimately generating a standardized dataset with controllable quality, consistent fields, and clear time sequence.

[0054] This invention solves the problem of fusion and consistency of multi-source heterogeneous data, providing a high-quality, directly applicable structured data foundation for subsequent analysis and modeling. Through standardization, data noise and the influence of units are eliminated, improving the stability and efficiency of subsequent feature engineering and model training.

[0055] Step 203: Perform feature engineering on the standardized data to generate a core feature set.

[0056] In this embodiment of the invention, step 203 specifically includes: Step 2031: Extract features from the standardized data to generate standardized features.

[0057] In this embodiment of the invention, based on business understanding and analysis requirements, data features are extracted from a standardized dataset and the extracted data features are determined as standardized features. The standardized features aim to characterize user status and behavior patterns from different perspectives.

[0058] For example, based on business logic, 2,500 initial features across 12 categories are constructed, including: asset-related features (such as the average amount of wealth management holdings in the past 3 months), behavioral features (such as the number of times wealth management products are viewed each month), and response-related features (such as the participation rate in historical marketing campaigns). Feature encoding is used to standardize the feature format; for example, the text label "risk preference" is converted to numerical encoding.

[0059] Step 2032: Select the core feature set from the standardized features using the ensemble tree model.

[0060] In this embodiment of the invention, an ensemble tree model (such as Gradient Boosting Decision Tree GBDT) is used, with whether users purchased (or other key business indicators) in historical marketing activities as the target variable. All features in a pre-built standardized feature pool are used as input to train the model, resulting in a trained ensemble tree model.

[0061] Specifically, based on the feature importance evaluation mechanism built into the trained ensemble tree model, the importance of each feature is scored and ranked. According to preset rules (such as selecting the top N features by importance score, or selecting feature combinations with a cumulative importance exceeding a certain percentage), the most predictive and discriminative feature subset, i.e., the core feature set, is automatically selected from the standardized features. This process achieves automatic feature dimensionality reduction and selection.

[0062] It is worth noting that the number of features in the core feature set can be set according to actual needs, and this embodiment of the invention does not limit this. As an optional solution, the core feature set includes 50 key features.

[0063] This invention transforms raw data into a core feature set with high information density, low redundancy, and strong correlation with marketing response through systematic feature extraction and automated feature selection based on machine learning. This significantly improves the quality of model input, enhances model interpretability, and reduces computational complexity, laying a solid foundation for subsequent accurate prediction.

[0064] Step 204: Using the conversion prediction model, perform probability prediction on the marketing tags, core feature set, basic product attributes, and purchase binary tags to generate the marketing intervention purchase probability.

[0065] In this embodiment of the invention, a pre-trained conversion prediction model is invoked. This model has learned the probabilistic relationship between user purchase behavior given user characteristics, product attributes, and whether or not marketing intervention is involved. The conversion prediction model is a classification model built based on XGBoost or a deep neural network.

[0066] Specifically, for each user to be evaluated, their core feature set and relevant product basic attributes are input into the conversion prediction model, and the user's marketing binary label is assigned as "marketed status". Based on these inputs, the conversion prediction model outputs a probability value, namely: marketing intervention purchase probability (P1), which represents the predicted likelihood that the user will ultimately purchase the product after the marketing intervention is implemented.

[0067] Step 205: Using the conversion prediction model, perform probability prediction on the non-marketing tag, core feature set, basic product attributes, and purchase binary tag to generate basic purchase probabilities.

[0068] Specifically, for each user to be evaluated, the core feature set and relevant basic product attributes are input into the conversion prediction model, and the user's marketing binary label is assigned as "no marketing intervention". Based on these inputs, the conversion prediction model outputs a probability value, namely: the basic purchase probability (P0), which represents the predicted likelihood that the user will ultimately purchase the product without any marketing intervention.

[0069] This invention indirectly quantifies the marketing incremental effect by using the same conversion prediction model to make two predictions for the same user under different intervention states. By integrating the idea of ​​causal inference into machine learning prediction, it can effectively distinguish whether a user's purchasing behavior is due to marketing stimulation or their own needs, breaking through the technical bottleneck of traditional response models that can only predict "whether" but cannot answer "whether it is because of marketing".

[0070] Step 206: Determine the difference between the base purchase probability and the marketing intervention purchase probability as the user sensitivity.

[0071] In this embodiment of the invention, user sensitivity = marketing intervention purchase probability (P1) - base purchase probability (P0). User sensitivity measures the net increase in the user's purchase probability due to marketing intervention. A positive value indicates that the marketing may bring benefits, a negative value indicates that the marketing may be ineffective or even cause aversion, and a zero value indicates that the marketing has no effect.

[0072] This invention transforms two probability prediction values ​​into a single, intuitive, and sortable indicator that directly reflects each user's sensitivity to marketing activities, providing a core metric for refined user segmentation.

[0073] Step 207: According to the preset sensitivity threshold, classify the user marketing sensitivity level based on the user's sensitivity.

[0074] In this embodiment of the invention, the user marketing sensitivity levels include high sensitivity level, medium sensitivity level, and no / low sensitivity level.

[0075] In this embodiment of the invention, business operations personnel set one or more sensitivity thresholds based on historical experience, business objectives, or by analyzing the distribution of user sensitivity (such as quantiles). Based on these sensitivity thresholds, users are categorized into different marketing sensitivity levels. Specifically, users with sensitivity greater than a first sensitivity threshold α are classified as having a high marketing sensitivity level; users with sensitivity less than a second sensitivity threshold β are classified as having no / low marketing sensitivity; and users with sensitivity between the first and second sensitivity thresholds α and β are classified as having a medium marketing sensitivity level.

[0076] It is worth noting that the first sensitivity threshold α and the second sensitivity threshold β can be set according to actual needs, and the embodiments of the present invention do not limit this. As an optional solution, the first sensitivity threshold α is 0.2 and the second sensitivity threshold β is 0.

[0077] This invention enables automated and standardized segmentation of user groups, transforming continuous sensitivity values ​​into discrete business tags, making the formulation of differentiated marketing strategies simple and feasible.

[0078] Step 208: Based on the user's marketing sensitivity level, construct a marketing resource allocation plan.

[0079] In this embodiment of the invention, based on the predefined user levels, a predefined resource allocation strategy library is automatically matched to generate a specific marketing resource allocation execution plan. The marketing resource allocation plan can be specific to channels, content, timing, frequency, and budget.

[0080] As an alternative, for users with high sensitivity levels, allocate high-cost but high-conversion premium resources, such as one-on-one communication with product managers, exclusive coupons, and priority access; for users with medium sensitivity levels, allocate standardized and scalable medium resources, such as precise app push notifications, email marketing, and targeted advertising; for users with no or low sensitivity levels, add them to the "Do Not Disturb" list and do not allocate proactive marketing resources, or only conduct very infrequent brand maintenance communication.

[0081] This invention automates and intelligently decides the allocation of marketing resources, ensuring that resources are precisely targeted to the most responsive groups, maximizing the overall return on investment (ROI), while reducing interference with insensitive users and optimizing the user experience.

[0082] Step 209: Obtain the actual purchase rate of users with high marketing sensitivity.

[0083] In this embodiment of the invention, after a marketing campaign has been running for a period of time (e.g., the campaign ends or a preset time interval is reached), the user group marked as having a high marketing sensitivity level and who actually received marketing outreach is retrieved from the business database. The proportion of users among them who ultimately made a purchase is then calculated, i.e., the actual purchase rate. This serves as a real-world verification of the model's predictive performance.

[0084] Step 210: Determine whether the predicted deviation between the actual purchase rate and the corresponding marketing intervention purchase probability is greater than the preset deviation threshold. If yes, proceed to step 201; otherwise, proceed to step 209.

[0085] In this embodiment of the invention, the difference between the actual purchase rate and the marketing intervention purchase rate is defined as the prediction deviation. If the prediction deviation is greater than the deviation threshold, it indicates that the predictive performance of the conversion prediction model may have significantly declined and cannot accurately reflect current user behavior. At this time, the system automatically triggers the model refresh process, which starts from step 201 again to collect the latest and more comprehensive historical marketing campaign data to initiate a new round of feature engineering, model training, and evaluation processes, thereby updating the core feature set and conversion prediction model. If the prediction deviation is less than or equal to the deviation threshold, it indicates that the predictive performance of the conversion prediction model is relatively accurate, and step 209 is continued for continuous monitoring.

[0086] It is worth noting that the deviation threshold can be set according to actual needs, and this embodiment of the invention does not limit it. As an optional solution, the deviation threshold is 10%.

[0087] This invention establishes a feedback control closed loop based on performance monitoring. By continuously comparing predicted values ​​with actual values, the system can autonomously detect model degradation and trigger a retraining mechanism in a timely manner, ensuring that the marketing decision-making system maintains accuracy and reliability in the long term and adapts to the ever-changing business environment.

[0088] Furthermore, in addition to performance-biased trigger-based updates, the system also features a periodic update mechanism. Regardless of the model's current performance, the system automatically re-executes the entire process from data collection at fixed time intervals (e.g., weekly, bi-weekly, or monthly). This ensures the system can regularly absorb the latest business data and continuously capture gradual changes in user behavior and market trends. Specifically, it re-executes the collection of multi-dimensional user data from historical marketing campaigns at preset time intervals. As an optional feature, the time interval is 7 days.

[0089] This invention provides a stable and periodic model maintenance and update rhythm to prevent the model's performance from gradually declining due to long-term lack of updates, which together with triggered updates constitute a dual protection mechanism.

[0090] Furthermore, multi-dimensional user data from the latest historical marketing campaigns is collected. Based on this data, the ensemble tree model and conversion prediction model are iteratively updated to generate updated versions. Specifically, the ensemble tree model used for feature importance assessment is retrained with new data, selecting a new set of core features more relevant to the current marketing response and eliminating older features whose importance has decreased. The conversion prediction model is retrained with new data (including the newly generated core feature set), optimizing its parameters to better reflect recent intervention-outcome relationships and improve prediction accuracy. The performance of the updated model is evaluated on a new validation set. If performance meets the criteria (user sensitivity prediction error less than 5%), the old online model is automatically or, after confirmation, replaced, and the updated ensemble tree model and conversion prediction model are generated and deployed.

[0091] This invention enables end-to-end dynamic iteration and self-evolution of the entire analysis and decision-making chain. It can not only fix problems, but also actively learn new patterns, so that the marketing resource allocation strategy is always in the best state, forming an adaptive intelligent decision-making system.

[0092] This invention has been tested and verified in real-world scenarios across all categories of business in financial institutions, achieving significant technical and business results: First, the model accuracy has been significantly improved. Relying on multi-dimensional feature engineering and a two-layer gain prediction architecture, it can accurately identify core user groups that are sensitive to financial outreach, effectively control the prediction bias of the Uplift value, and significantly improve the conversion effect of outreach across all businesses compared with traditional outreach models.

[0093] Second, the efficiency of resource allocation has been optimized and upgraded. By tilting outreach resources towards high-value and sensitive users, the consumption of ineffective resources has been significantly reduced, the return on investment in outreach has been improved, and the waste of resources on users with no outreach needs has been avoided.

[0094] Third, user experience and brand value are improved simultaneously, significantly reducing unnecessary disturbances to non-sensitive users, reducing user service complaints, simplifying business processing procedures, and significantly improving user service satisfaction.

[0095] The relevant technologies have been applied in various financial business scenarios such as wealth management, credit, and insurance, providing a replicable and scalable technological paradigm for the digital transformation of financial institutions.

[0096] It is worth noting that the acquisition, storage, use, and processing of data in the technical solution of this application all comply with relevant laws and regulations. The user information in the embodiments of this application was obtained through legal and compliant means, and the acquisition, storage, use, and processing of user information have been authorized and agreed upon by the users.

[0097] It is worth noting that the information collected in this application is information and data authorized by the user or fully authorized by all parties, and the collection, storage, use, processing, transmission, provision, disclosure and application of the relevant data all comply with the relevant laws, regulations and standards of the relevant countries and regions, necessary confidentiality measures have been taken, and they do not violate public order and good morals. Corresponding operation portals are provided for users to choose to authorize or refuse.

[0098] It is worth noting that the technical solution provided in this application provides users with a corresponding operation entry point, allowing users to choose to agree to or reject the automated decision-making result; if the user chooses to reject, the process will proceed to the expert decision-making process.

[0099] The technical solution of the marketing resource allocation method provided in this invention involves obtaining the core feature set of users from historical marketing activities; using a trained conversion prediction model, probabilistic predictions are made on the core feature set, stored product basic attributes, marketing binary classification tags, and purchase binary classification tags to generate basic purchase probability and marketing intervention purchase probability; based on the basic purchase probability and marketing intervention purchase probability, the user's marketing sensitivity level is determined; based on the user's marketing sensitivity level, a marketing resource allocation scheme is constructed. By constructing a complete technical framework of "feature engineering screening - two-layer probability prediction - sensitivity quantification," accurate calculation and grading based on individual causal incremental effects are achieved in marketing resource allocation, accurately quantifying the user's sensitivity to marketing intervention and achieving effective stratification of user groups; simultaneously, its built-in dynamic feature and model iteration mechanism ensures the robustness and adaptability of the algorithm in continuously changing business environments from a system level, forming a self-optimizing intelligent decision-making closed loop, ensuring the dynamic adaptability and long-term effectiveness of marketing strategies, and optimizing user experience while improving conversion results.

[0100] Figure 3 This is a schematic diagram of a marketing resource allocation device provided in an embodiment of the present invention. This device is used to execute the aforementioned marketing resource allocation method, such as... Figure 3 As shown, the device includes: a core feature set acquisition unit 11, a probability prediction unit 12, a sensitivity level determination unit 13, and a resource allocation unit 14.

[0101] The core feature set acquisition unit 11 is used to acquire the core feature set of users in historical marketing campaigns.

[0102] The probability prediction unit 12 is used to perform probability prediction on the core feature set, stored product basic attributes, marketing binary labels and purchase binary labels through the trained conversion prediction model, and generate basic purchase probability and marketing intervention purchase probability.

[0103] Sensitivity Level Determination Unit 13 is used to determine the user's marketing sensitivity level based on the basic purchase probability and the marketing intervention purchase probability.

[0104] Resource allocation unit 14 is used to construct a marketing resource allocation scheme based on the user's marketing sensitivity level.

[0105] In this embodiment of the invention, the core feature set acquisition unit 11 is specifically used to collect multi-dimensional data of users in historical marketing activities; to associate and integrate the multi-dimensional data according to the user identifier and timestamp dimensions to obtain a standardized dataset; and to perform feature engineering processing on the standardized data to generate a core feature set.

[0106] In this embodiment of the invention, the core feature set acquisition unit 11 is specifically used to extract features from standardized data and generate standardized features; and to select the core feature set from the standardized features through an ensemble tree model.

[0107] In this embodiment of the invention, the marketing binary category label includes a marketed label or a non-marketed label; the probability prediction unit 12 is specifically used to perform probability prediction on the marketed label, core feature set, product basic attributes and purchase binary category label through a conversion prediction model to generate a marketing intervention purchase probability; and to perform probability prediction on the non-marketed label, core feature set, product basic attributes and purchase binary category label through a conversion prediction model to generate a basic purchase probability.

[0108] In this embodiment of the invention, the sensitivity level determination unit 13 is specifically used to determine the difference between the basic purchase probability and the marketing intervention purchase probability as the user sensitivity; and to classify the user marketing sensitivity level according to the preset sensitivity threshold.

[0109] In this embodiment of the invention, the device further includes: the core feature set acquisition unit 11 is also used to re-execute the step of collecting multi-dimensional data of users in historical marketing activities at preset time intervals.

[0110] In this embodiment of the invention, the device further includes: an actual purchase rate acquisition unit 15.

[0111] The actual purchase rate acquisition unit 15 is used to acquire the actual purchase rate of users with high marketing sensitivity. If the predicted deviation between the actual purchase rate and the corresponding marketing intervention purchase probability is greater than the preset deviation threshold, the core feature set acquisition unit 11 is triggered to re-execute the step of collecting multi-dimensional data of users in historical marketing activities.

[0112] In this embodiment of the invention, the device further includes: a data acquisition unit 16 and an iterative update unit 17.

[0113] The data collection unit 16 is used to collect multi-dimensional data of users from the latest historical marketing campaigns.

[0114] The iterative update unit 17 is used to iteratively update the integrated tree model and conversion prediction model based on the latest multi-dimensional data of users in historical marketing activities, and generate the updated integrated tree model and conversion prediction model.

[0115] In this embodiment of the invention, the core feature set of users in historical marketing activities is obtained; through a trained conversion prediction model, probability prediction is performed on the core feature set, stored product basic attributes, marketing binary classification tags, and purchase binary classification tags to generate basic purchase probability and marketing intervention purchase probability; based on the basic purchase probability and marketing intervention purchase probability, the user's marketing sensitivity level is determined; based on the user's marketing sensitivity level, a marketing resource allocation scheme is constructed. By constructing a complete technical framework of "feature engineering screening - two-layer probability prediction - sensitivity quantification", the scheme achieves accurate calculation and grading based on individual causal incremental effects in marketing resource allocation, accurately quantifies the user's sensitivity to marketing intervention, and achieves effective stratification of user groups; at the same time, its built-in dynamic feature and model iteration mechanism ensures the robustness and adaptability of the algorithm in a continuously changing business environment from the system level, forming a self-optimizing intelligent decision-making closed loop, ensuring the dynamic adaptability and long-term effectiveness of the marketing strategy, and optimizing the user experience while improving conversion results.

[0116] The systems, devices, modules, or units described in the above embodiments can be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer device, specifically, a computer device can be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smartphone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or any combination of these devices.

[0117] This invention provides a computer device including a memory and a processor. The memory is used to store information including program instructions, and the processor is used to control the execution of the program instructions. When the program instructions are loaded and executed by the processor, they implement the steps of the above-described marketing resource allocation method. For a detailed description, please refer to the above-described marketing resource allocation method embodiments.

[0118] The following is for reference. Figure 4 It shows a schematic diagram of the structure of a computer device 600 suitable for implementing the embodiments of this application.

[0119] like Figure 4 As shown, the computer device 600 includes a central processing unit (CPU) 601, which can perform various appropriate tasks and processes based on programs stored in read-only memory (ROM) 602 or programs loaded from storage section 608 into random access memory (RAM) 603. The RAM 603 also stores various programs and data required for the operation of the computer device 600. The CPU 601, ROM 602, and RAM 603 are interconnected via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0120] The following components are connected to I / O interface 605: an input section 606 including a keyboard, mouse, etc.; an output section 607 including a cathode ray tube (CRT), liquid crystal feedback (LCD), etc., and speakers, etc.; a storage section 608 including a hard disk, etc.; and a communication section 609 including a network interface card such as a LAN card, modem, etc. The communication section 609 performs communication processing via a network such as the Internet. A drive 610 is also connected to I / O interface 605 as needed. A removable medium 611, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on drive 610 as needed so that computer programs read from it can be installed in storage section 608 as needed.

[0121] In particular, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program tangibly embodied on a machine-readable medium, the computer program including program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 609, and / or installed from removable medium 611.

[0122] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information by any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0123] For ease of description, the above devices are described separately by function as various units. Of course, in implementing this application, the functions of each unit can be implemented in one or more software and / or hardware.

[0124] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0125] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0126] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0127] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0128] The acquisition, storage, use, and processing of data in this application all comply with the relevant provisions of national laws and regulations.

[0129] It should be noted that in the embodiments of this application, certain software, components, models and other existing solutions in the industry may be mentioned. These should be regarded as exemplary and are only intended to illustrate the feasibility of implementing the technical solution of this application. However, it does not mean that the applicant has used or necessarily used the solution.

[0130] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0131] This application can be described in the general context of computer-executable instructions, such as program modules, that are executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform a specific task or implement a specific abstract data type. This application can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0132] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to interchangeably. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments.

[0133] The above description is merely an embodiment of this application and is not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A method for allocating marketing resources, characterized in that, The method includes: Obtain the core feature set of users from historical marketing campaigns; The conversion prediction model, trained by the model, performs probability prediction on the core feature set, stored product basic attributes, marketing binary tags, and purchase binary tags to generate basic purchase probability and marketing intervention purchase probability. Based on the basic purchase probability and the purchase probability after marketing intervention, the user's marketing sensitivity level is determined; Based on the user's marketing sensitivity level, a marketing resource allocation plan is constructed.

2. The marketing resource allocation method according to claim 1, characterized in that, The core feature set of users obtained from historical marketing campaigns includes: Collect multi-dimensional user data from historical marketing campaigns; The multi-dimensional data is linked and integrated according to user identifier and timestamp dimensions to obtain a standardized dataset; The standardized data is subjected to feature engineering to generate a core feature set.

3. The marketing resource allocation method according to claim 2, characterized in that, The step of performing feature engineering on the standardized data to generate a core feature set includes: Feature extraction is performed on the standardized data to generate standardized features; The core feature set is selected from the standardized features using an ensemble tree model.

4. The marketing resource allocation method according to claim 1, characterized in that, The marketing binary category labels include either "marketed" labels or "not marketed" labels; The trained conversion prediction model performs probability predictions on the core feature set, stored product basic attributes, marketing binary classification tags, and purchase binary classification tags to generate basic purchase probabilities and marketing intervention purchase probabilities, including: The conversion prediction model is used to predict the probability of marketing intervention purchase based on marketing tags, core feature sets, basic product attributes, and purchase binary tags. The conversion prediction model is used to predict the probability of non-marketing tags, core feature sets, basic product attributes, and purchase binary tags, and to generate basic purchase probabilities.

5. The marketing resource allocation method according to claim 1, characterized in that, The step of determining the user's marketing sensitivity level based on the basic purchase probability and the marketing intervention purchase probability includes: The difference between the base purchase probability and the marketing intervention purchase probability is defined as user sensitivity. Based on preset sensitivity thresholds, user marketing sensitivity levels are categorized according to user sensitivity.

6. The marketing resource allocation method according to claim 2, characterized in that, The method further includes: The steps for collecting multi-dimensional user data from historical marketing campaigns are repeated at preset time intervals.

7. The marketing resource allocation method according to claim 2, characterized in that, After constructing the marketing resource allocation scheme based on the user's marketing sensitivity level, the following is also included: To obtain the actual purchase rate of users with high marketing sensitivity; If the predicted deviation between the actual purchase rate and the corresponding marketing intervention purchase probability is greater than a preset deviation threshold, the step of collecting multi-dimensional data of users in historical marketing activities is repeated.

8. The marketing resource allocation method according to claim 3, characterized in that, After constructing the marketing resource allocation scheme based on the user's marketing sensitivity level, the following is also included: Collect multi-dimensional user data from the latest historical marketing campaigns; Based on the latest multi-dimensional user data from historical marketing campaigns, the integrated tree model and conversion prediction model are iteratively updated to generate the updated integrated tree model and conversion prediction model.

9. A marketing resource allocation device, characterized in that, The device includes: The core feature set acquisition unit is used to acquire the core feature set of users in historical marketing campaigns. The probability prediction unit is used to perform probability prediction on the core feature set, stored product basic attributes, marketing binary tags and purchase binary tags through a trained conversion prediction model, and generate basic purchase probability and marketing intervention purchase probability. The sensitivity level determination unit is used to determine the user's marketing sensitivity level based on the basic purchase probability and the marketing intervention purchase probability. The resource allocation unit is used to construct a marketing resource allocation scheme based on the user's marketing sensitivity level.

10. A computer-readable medium having a computer program stored thereon, characterized in that, When executed by a processor, the program implements the marketing resource allocation method as described in any one of claims 1 to 8.

11. A computer device comprising a memory and a processor, the memory for storing information including program instructions, and the processor for controlling the execution of the program instructions, characterized in that, When the program instructions are loaded and executed by the processor, they implement the marketing resource allocation method according to any one of claims 1 to 8.

12. A computer program product comprising a computer program / instructions, characterized in that, When the computer program / instructions are executed by the processor, they implement the marketing resource allocation method according to any one of claims 1 to 8.

Images