An intelligent marketing strategy optimization method based on big data

Abstract

The application relates to the technical field of big data processing, and discloses an intelligent marketing strategy optimization method based on big data, which comprises the following steps: step one, using a monitoring probe to collect multi-source heterogeneous marketing events in a full link, and performing cross-terminal space alignment on the collected marketing event sequences based on an anonymous consistency identifier to construct an original time sequence table; step two, calling a historical conversion sample library and counting the time difference distribution of user conversion behaviors, calculating a decay constant reflecting the conversion decay rate, and performing calibration processing on active samples corresponding to the original time sequence table based on the decay constant. The anonymous consistency identifier and the space alignment technology are used to achieve the logical correlation of cross-terminal marketing event sequences in a desensitization environment, realize the accurate recovery of the full link of user conversion, and solve the problems of attribution deviation and marketing resource mismatch caused by data fragmentation.

Description

Technical Field

[0001] This invention relates to the field of big data processing technology, specifically to a method for optimizing intelligent marketing strategies based on big data, an electronic device, and a storage medium. Background Technology

[0002] Due to the increasing complexity of the internet marketing ecosystem, businesses face a massive, multi-dimensional, and heterogeneous data environment when making traffic allocation and user growth decisions. In current digital marketing practices, users frequently switch between multiple devices and touchpoints, generating discrete and fragmented behavioral patterns. To improve marketing ROI, accurately assessing the value of each touchpoint and developing dynamic optimization strategies has become a core industry demand. However, in practical applications, existing marketing strategy optimization solutions still face the following significant challenges:

[0003] First, without privacy restrictions based on personally identifiable identifiers, it is difficult to effectively connect user behavior paths across different devices and marketing touchpoints, making it impossible for traditional attribution models to accurately reconstruct the entire user conversion journey. Due to the fragmentation of cross-device data, it is difficult to accurately assess the true contribution of each marketing touchpoint, resulting in the misallocation and waste of marketing resources.

[0004] Second, existing marketing analysis mainly relies on correlation analysis, which cannot distinguish the causal incremental relationship between marketing touchpoints and user conversion behavior. This results in a lack of scientific basis for the formulation of marketing strategies, and it cannot prove whether the conversion behavior is directly induced by specific marketing interventions or is a natural behavior of users, making the strategy optimization process lack logical support.

[0005] Third, there is a significant time gap between a user's exposure to marketing information and their actual purchase behavior. This feedback delay can cause strategy optimization algorithms to miss accurate signals in the short term. Due to the lack of modeling for the patterns of delayed conversions, frequent and misleading adjustments can be made, leading to drastic fluctuations in advertising budgets and strategy execution.

[0006] Fourth, in environments with strong privacy constraints, first-party data and data from third-party platforms cannot be correlated in plaintext, resulting in a lack of high-quality underlying feature support for model training. Because efficient cross-domain feature collaboration cannot be achieved while protecting data security, prediction accuracy significantly decreases during the cold start phase or in data-sparse scenarios.

[0007] To address the aforementioned issues, this invention proposes a big data-based intelligent marketing strategy optimization method. Through cross-platform anonymous alignment, causal incremental modeling, and differential privacy noise optimization, it achieves delayed conversion calibration and closed-loop dynamic adjustment, resolving shortcomings such as inaccurate attribution, unclear causal relationships, and strategy oscillations, thereby improving the scientific rigor and security of marketing decisions. Summary of the Invention

[0008] In view of the shortcomings of the prior art, the present invention provides a method for optimizing intelligent marketing strategies based on big data, an electronic device and a storage medium to solve one or more problems mentioned in the background art.

[0009] To achieve the above objectives, the present invention provides the following technical solution:

[0010] In a first aspect, the present invention provides a method for optimizing intelligent marketing strategies based on big data, comprising:

[0011] Step 1: Use monitoring probes to collect multi-source heterogeneous marketing events across the entire chain, and perform cross-terminal spatial alignment of the collected marketing event sequences based on anonymous consistent identifiers to construct the original time series table;

[0012] Step 2: Retrieve the historical conversion sample library and statistically analyze the time difference distribution of user conversion behavior, calculate the decay constant reflecting the conversion decay rate, and calibrate the active samples corresponding to the original time series table based on the decay constant.

[0013] Step 3: Using the attenuation constant as the input parameter of the time attenuation operator, the discrete events in the original time series table after calibration are mapped to construct a feature vector of touch point influence intensity containing spatiotemporal weight information.

[0014] Step 4: Input the feature vector of the touch point influence intensity into the pre-constructed causal inference network to predict the probability of the user being reached and the response level under different marketing actions, and calculate the net incremental contribution value corresponding to each marketing action.

[0015] Step 5: Using the net incremental contribution value as the objective function input, and taking privacy budget consumption and budget quota as the joint hard constraint space, solve to obtain the initial marketing strategy including the consumption fluctuation penalty term;

[0016] Step 6: Monitor the deviation signals between the observed indicators and target indicators at the marketing execution end in real time, calculate the strategy correction increment, and smoothly adjust the initial marketing strategy.

[0017] Step 7: Send the smoothed and adjusted marketing strategy to the campaign execution engine, and feed back the conversion data after execution to Step 1 to enter the next round of marketing strategy iteration.

[0018] Preferably, the joint hard constraint space in step six includes: setting an upper limit on the frequency of contact for a single user; when the historical contact frequency of a specific user reaches this upper limit, the initial marketing strategy forcibly modifies the recommendation action for that user to stop intervening.

[0019] Preferably, the operation of feeding back the conversion data after execution to step one includes: using the real conversion feedback generated by executing the initial marketing strategy as a new historical conversion sample to dynamically update the time difference distribution model in step two, thereby achieving adaptive iteration of the decay constant.

[0020] Preferably, step one further includes:

[0021] Sub-step By capturing multi-source heterogeneous raw data streams generated by user behavior through monitoring probes preset on different marketing channels, and using preset cleaning rules to perform outlier removal and format normalization on the multi-source heterogeneous raw data streams, a standardized marketing event stream containing original user identifiers, behavior trigger times, and touchpoint types is obtained.

[0022] Sub-step The asymmetric encryption algorithm is invoked to perform one-way desensitization processing on the original user identifiers in the standardized marketing event stream, generating a globally unique anonymous and consistent identifier. The anonymous consistency identifier The formula for generating it is:

[0023] ,

[0024] in, For the original user identifier, For any preset perturbation salt value, The XOR operator. For a hash function with collision resistance, the anonymous consistent identifier is used. By linking standardized marketing event streams that are discrete across different terminals, a preliminary aligned set of behaviors can be obtained;

[0025] Sub-step Based on the initial aligned set of behaviors, spatial alignment rules are used for the same anonymous consistent identifier. Perform time-series aggregation on cross-end behaviors to construct the original time series table. :

[0026] ,in, For behavior The corresponding contact type description value, For behavior The corresponding behavior trigger time;

[0027] Sub-step For the original time series table Perform integrity determination;

[0028] If the judgment condition is met ≤ The original time series table is determined to be a valid sequence.

[0029] If the judgment condition is met Perform truncation processing, where, This is the preset length of the attribution observation window.

[0030] Preferably, step two further includes:

[0031] Sub-step The system retrieves all historical samples from the historical conversion sample library, extracts the trigger time of marketing touchpoints and the time of final conversion behavior in each historical sample, and calculates the conversion time difference for each sample. The calculation formula is: ,in, The moment when the final conversion occurs. For marketing touchpoint trigger moments;

[0032] Sub-step Based on the conversion time difference of the single sample The sample set is fitted using the cumulative distribution function to determine the single-sample conversion time difference. median of the distribution And based on the median Calculate the decay constant that reflects the conversion decay rate. The calculation formula is: ,in, For the natural logarithm operator, through the decay constant Characterizes the exponential decay of the transformation probability over time;

[0033] Sub-step Based on the attenuation constant Set the conversion calibration function Used to calculate the current active sample over the observation period. The conversion probability compensation value, the conversion calibration function The formula is: ,in, The time span from the triggering of the marketing touchpoint to the current observation time for the currently active sample;

[0034] Sub-step A probability compensation criterion is introduced, and the active samples corresponding to the original time series table are input into the transformation calibration function. Perform calibration processing;

[0035] If the judgment condition is met The currently active sample is determined to be a highly uncertain sample, and according to the formula... Perform conversion probability compensation calibration;

[0036] If the judgment condition is met ≥ If the currently active sample is determined to be a stable sample, the observed transformation state of the currently active sample is directly recorded. The preset confidence threshold, The initial transformation value observed so far. This is the calibrated target conversion value.

[0037] Preferably, step three further includes:

[0038] Sub-step Retrieve the valid sequence as determined, and identify each behavior in the valid sequence. Corresponding trigger time Based on the current observation time Calculate the relative delay of each action. The calculation formula is: Through the relative time delay Measure the distance of the influence of each historical point on the current moment;

[0039] Sub-step Introducing the attenuation constant Construct a kernel function to quantify the effect of contact decay. The calculation formula is: , where the kernel function The output value is determined by the relative time delay. The influence of early touchpoints on current decisions decreases exponentially as they increase;

[0040] Sub-step Based on the kernel function For the same contact type description value The cumulative influence intensity values ​​of various contact points are calculated by nonlinearly superimposing multiple behaviors. The calculation formula is: ,in, The contact type in the valid sequence belongs to the category A set of behavioral indexes, For the preset corresponding behavior The business importance weighting coefficient, the cumulative influence strength value This comprehensively reflects the combined effect of contact trigger frequency and time decay;

[0041] Sub-step According to the preset contact category index order, the cumulative influence intensity values ​​of each contact category are calculated. By vectorizing the arrangement, the feature vector of the contact point influence intensity is constructed. And the contact point influence intensity feature vector As input features for causal inference networks, among which, This represents the total number of pre-defined marketing touchpoint categories.

[0042] Preferably, step four further includes:

[0043] Sub-step The constructed contact point influence intensity feature vector Input a pre-defined classifier model to calculate the specific marketing actions that will be performed on the user under the current environmental characteristics. Propensity score The calculation formula is: ,in, As a variable for marketing actions, To assign values ​​to specific marketing actions, To execute specific marketing actions The propensity score is used to characterize the selection bias of the sample;

[0044] Sub-step A result prediction model is constructed using a regression algorithm, based on the feature vector of the influence intensity of the contact point. Predict users' specific marketing actions First response expectation under the condition And the expected value of the second response under the condition of no marketing intervention. The calculation formula is:

[0045] , ,in, For the preset conversion indicator result variable, Refers to the expected conversion level under intervention conditions. Refers to the expected conversion level under control conditions;

[0046] Sub-step Based on the aforementioned tendency score and the expected value of the first response Second response expectation The net incremental contribution value corresponding to each marketing action is calculated using a dual-correction operator. The calculation formula is:

[0047] ,

[0048] in, For indicator functions, The net incremental contribution value is the score for the tendency to engage in no marketing intervention. Used to quantify specific marketing actions Compared to the absolute conversion improvement efficiency in the absence of intervention;

[0049] Sub-step Introducing significance criteria for the calculated net incremental contribution value Perform the screening; if the judgment criteria are met. and If the net incremental contribution value is determined to be a valid incremental signal, and if the determination condition is not met, the net incremental contribution value is determined to be a noise signal and a zeroing process is performed. The preset contribution threshold, This is the preset tendency score cutoff constant.

[0050] Preferably, step six further includes:

[0051] Sub-step The effective incremental signals are retrieved and combined with the updated causal inference network model to predict the target sample set and obtain the marketing actions. Corresponding expected net incremental contribution value Construct a global objective function The formula is:

[0052] ,in, The initial marketing strategy to be solved. A set of optional marketing actions. To consume the fluctuation penalty item, The preset penalty factor coefficient, This represents the total number of samples to be reached.

[0053] Sub-step Define the consumption fluctuation penalty term. The calculation formula is:

[0054] ,

[0055] in, In the initial marketing strategy Below users The expected costs of intervention, The average budget per user is preset, and the consumption fluctuation penalty item is applied. Constrain the smooth distribution of marketing resources in the sample space;

[0056] Sub-step A joint hard constraint space is defined, which includes budget quota constraints and cumulative privacy budget consumption constraints. The mathematical expression is:

[0057] ≤ ≤ ,in, Given a pre-set total marketing budget, the feasible domain of the initial marketing strategy is defined through the joint hard constraint space.

[0058] Sub-step An optimization criterion is introduced, and the Lagrange multiplier method is used to evaluate the global objective function. Perform the solution within the combined hard constraint space;

[0059] If the convergence criteria are met Determine the initial marketing strategy. The optimal numerical solution is obtained; if the convergence criterion is not met, the initial marketing strategy is corrected along the gradient ascent direction. The probability distribution parameters are then recalculated, where, Let be the gradient vector of the objective function. This is the preset convergence accuracy threshold.

[0060] Preferably, step seven further includes:

[0061] Sub-step Obtain the initial marketing strategy The optimal numerical solution is obtained and sent to the execution end, and the observation index vector of the marketing execution end is monitored in real time through monitoring probes. Calculate the observation index vector With the preset target indicator vector Deviation signal between The calculation formula is:

[0062] ,in, At the current monitoring sampling time, the deviation signal is used. It reflects in real time the degree of deviation between the current marketing strategy's execution effect and the expected goals;

[0063] Sub-step The deviation signal Input a proportional-integral-derivative (PID) controller to calculate the strategy correction increment used to eliminate the deviation signal. The calculation formula is:

[0064] ,

[0065] in, For the median variable of the integration time, This is a preset proportional coefficient. The preset integral coefficient, The preset differential coefficients are used to achieve a rapid response to execution fluctuations and the elimination of accumulated errors through a linear combination of the proportional term, integral term, and differential term.

[0066] Preferably, step seven further includes:

[0067] Sub-step Based on the strategy, the increment is corrected. Regarding the initial marketing strategy The optimal numerical solutions are superimposed online to calculate the adjusted marketing strategy. The calculation formula is:

[0068] ,in, The normalized exponential function is used to correct the increment through the aforementioned strategy. Implement the initial marketing strategy Smooth adjustment, and ensure the marketing strategy after adjustment. It conforms to probability distribution constraints;

[0069] Sub-step A stability criterion is introduced for the adjusted marketing strategy. Perform compliance verification; if the judgment conditions are met. ≤ If the current strategy adjustment is determined to be in a smooth state, the adjusted marketing strategy will be output. ;

[0070] If the judgment condition is met If the current strategy is determined to be highly volatile, step size truncation is performed: ,in, The preset maximum step size threshold for a single adjustment. For symbolic functions, This represents the policy state value from the previous moment.

[0071] Secondly, this application provides an electronic device, comprising:

[0072] A processor; and a memory storing program instructions that, when executed by the processor, cause the electronic device to implement the method disclosed in any of the embodiments of the first aspect above.

[0073] Thirdly, this application provides a storage medium, characterized in that the storage medium is a computer-readable storage medium storing computer-readable instructions thereon, which, when executed by one or more processors, implement the method disclosed in any embodiment of the first aspect above.

[0074] By collecting multi-source heterogeneous data across the entire chain and aligning it with anonymized, consistent identifiers in the spatial domain, we can break down the underlying data silos between different terminals while ensuring user privacy and security. This enables the complete association of discrete behavior sequences and provides a high-quality data foundation for reconstructing real user paths and accurate attribution.

[0075] By calculating the decay constant using the historical conversion time distribution pattern and performing probability compensation on active samples, the inherent lag problem of marketing feedback signals is effectively solved, ensuring that potential conversion trends can be captured in real time during the user decision-making cycle and avoiding evaluation bias caused by incomplete observation data.

[0076] By introducing a time decay operator, discrete events are mapped into feature vectors containing spatiotemporal weights. This scientifically quantifies the natural decay characteristics of the influence of different historical touchpoints on the current decision, significantly enhancing the fine-grainedness and timeliness of feature expression, enabling the model to sensitively capture the dynamic process of user interest evolution.

[0077] By constructing a causal inference network and combining it with a dual correction operator to calculate the net incremental contribution value, we can successfully isolate users' natural conversion tendencies, accurately lock in the absolute benefits triggered by specific marketing actions, realize the decision-making leap from superficial correlation to deep causal logic, and ensure the scientific nature of resource allocation.

[0078] Within the noise disturbance space, parameter updates are completed through sensitivity control and adaptive noise injection, ensuring that the model training process complies with strict differential privacy compliance requirements, preventing the leakage of single sample information, and improving the model's prediction accuracy while ensuring high security in the data processing process.

[0079] By combining net incremental contributions with privacy budgets and funding quotas to construct a joint hard constraint space, and solving the initial strategy containing consumption fluctuation penalties through Lagrange optimization, marketing resources are smoothly distributed while satisfying business boundary conditions, effectively preventing excessive concentration of resources in local spaces.

[0080] By monitoring the deviation signal at the execution end in real time and calculating the correction increment through a proportional-integral-derivative controller, closed-loop online compensation for external environmental fluctuations and execution errors is achieved, ensuring the robustness and smoothness of the strategy output and completely solving the risks of strategy oscillation and budget overrun that are prone to occur in traditional deployment systems.

[0081] By distributing the adjusted strategy and implementing a closed-loop feedback mechanism for conversion data, a continuously evolving adaptive learning mechanism can be built. This mechanism can continuously correct the decay model and prediction parameters based on the latest market feedback, ensuring the continued optimality of the intelligent marketing strategy in long-term operation.

[0082] The present invention has the following beneficial effects:

[0083] 1. This invention employs anonymous consistent identifiers and spatial alignment technology to achieve logical association of cross-platform marketing event sequences in an anonymized environment, enabling accurate restoration of the entire user conversion chain and solving the shortcomings of attribution bias and marketing resource mismatch caused by data fragmentation.

[0084] 2. This invention employs a causal inference network and a dual correction operator to achieve the technical effect of stripping away users' natural conversion tendencies and quantifying the real increase in marketing actions. It enables a leap from correlation to causation in decision-making, addressing the shortcomings of marketing strategy formulation that lack logical support and scientific basis.

[0085] 3. This invention employs conversion probability compensation calibration and closed-loop smoothing adjustment technology to achieve dynamic modeling of delayed conversion patterns and real-time correction of feedback errors, thereby enabling stable output of marketing strategies over long-term cycles and addressing the shortcomings of strategy oscillations and budget waste caused by delayed feedback signals.

[0086] 4. This invention employs a differential privacy injection and noise perturbation optimization scheme to achieve the technical effect of ensuring data security and feature collaboration under strict privacy constraints, thereby achieving a balance between high compliance and high prediction accuracy and solving the problem of significant performance degradation in cold start and sparse scenarios. Attached Figure Description

[0087] Figure 1 This is a flowchart illustrating an intelligent marketing strategy optimization method based on big data according to the present invention. Detailed Implementation

[0088] To enable those skilled in the art to understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort should fall within the scope of protection of the present invention.

[0089] The present invention will now be described in detail with reference to the accompanying drawings:

[0090] Example 1: Cross-platform high-value user causal incremental marketing scenario

[0091] In this embodiment, the present invention is applied to a cross-terminal promotional event on an e-commerce platform, aiming to verify the attribution accuracy and causal decision-making ability of the solution under the premise of privacy protection.

[0092] The implementation process is as follows:

[0093] First, monitoring probes deployed on mobile devices, web browsers, and tablets were used to capture the original user behavior flow, including clicking marketing links, browsing product details, and adding items to the shopping cart. Anonymous, consistent identifiers generated using asymmetric encryption were used to spatially align the discrete user behaviors across different devices, successfully reconstructing the complete user journey from receiving push notifications on mobile devices to placing orders on web browsers. Then, by retrieving conversion samples from historical promotional periods, a decay constant reflecting conversion latency characteristics was calculated, and conversion probability compensation was applied to active users who had not yet placed orders, identifying target groups with high conversion potential.

[0094] In the feature construction phase, a time decay operator is applied to the user touchpoint sequence to generate a touchpoint influence intensity vector containing spatiotemporal weighted features. Through a causal inference network, the model predicts user response levels under "discount coupon" intervention and without intervention, calculating the net incremental contribution of each marketing action. In the model update phase, adaptive noise compliant with differential privacy requirements is injected to ensure user information security. Finally, using the budget quota as a hard constraint, an initial marketing strategy balancing maximizing net incremental contribution and smoothing consumption is derived and then sent to the campaign execution engine after deviation correction by a proportional-integral-derivative controller.

[0095] Implementation results demonstrate that, through comparative experiments, this embodiment effectively eliminates naturally converting users who would have placed an order anyway, concentrating marketing budgets on incremental users who only convert through intervention. While ensuring user privacy compliance, the marketing return on investment is significantly improved, and the coverage of cross-platform attribution is greatly increased, verifying the feasibility of this invention in improving resource allocation efficiency in complex, multi-touchpoint environments.

[0096] Example 2: Optimization Scenario for Delayed Conversion of Products with Long Decision-Making Cycles

[0097] In this embodiment, the present invention is applied to high-value lead generation marketing for a certain automobile brand, aiming to verify the solution's ability to process delayed feedback signals and the stability of strategy execution.

[0098] The implementation process is as follows:

[0099] Because the decision-making cycle for car purchases is relatively long, monitoring probes continuously collect user lead retention and in-store appointment events over several weeks. An original time series table is constructed based on anonymous identifiers. Furthermore, considering the slow conversion rate in the automotive industry, a cumulative distribution function is used to fit an accurate decay constant. Conversion probabilities are then calibrated for active samples within the long decision-making cycle, avoiding the misconception of traditional models that classify unconverted users as invalid samples.

[0100] Next, a feature vector of the influence intensity of touchpoints is constructed as input to the causal inference network to quantify the absolute gain of actions such as test drive invitations and professional evaluation pushes on the final lead conversion. During model training, an upper bound on single-user sensitivity is set, and differential privacy noise is injected to achieve efficient cross-regional data collaboration. During the execution of the marketing strategy, the deviation signals between the observed and target metrics at the execution end are monitored in real time. A proportional-integral-derivative controller outputs the strategy correction increment, smoothly adjusting the initial delivery probability to prevent drastic fluctuations in budget consumption due to feedback delays. Post-execution conversion data is fed back to the data base in real time, driving the iterative updates of the decay model and prediction network for the next round.

[0101] Implementation results demonstrate that the proposed solution maintains extremely high stability in the face of feedback delays lasting several weeks, without exhibiting budget fluctuations. Through parameter updates under adaptive noise perturbation, prediction accuracy remains stable even during the cold start phase, proving the robustness and practicality of this invention in handling long-link feedback and strong privacy constraints.

[0102] Embodiments of the present invention have been presented and described. It will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to the embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

[0103] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0104] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0105] It is understood that the relevant features in the above methods and apparatus can be referenced interchangeably. Furthermore, the terms "first," "second," etc., in the above embodiments are used to distinguish between embodiments and do not represent the superiority or inferiority of any particular embodiment.

[0106] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0107] The algorithms and displays provided herein are not inherently related to any particular computer, virtual system, or other device. Various general-purpose systems can also be used in conjunction with the teachings herein. The required structure for constructing such systems is apparent from the above description. Furthermore, this invention is not directed to any particular programming language. It should be understood that the contents of the invention described herein can be implemented using various programming languages, and the above description of specific languages ​​is for the purpose of disclosing the best mode of implementation of the invention.

[0108] In addition, the memory may include non-permanent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0109] 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.

[0110] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. 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... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0111] 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.

[0112] 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.

[0113] In a typical configuration, an electronic device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0114] Memory may include non-persistent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, like read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0115] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, program modules, 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 magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by electronic devices. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0116] 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 process, method, article, or apparatus. Unless otherwise specified, 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 that element.

[0117] 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.

[0118] It should be noted that although the operations of the method of this application are described in a specific order in the accompanying drawings, this does not require or imply that these operations must be performed in that specific order, or that all the operations shown must be performed to achieve the desired result. On the contrary, the steps depicted in the flowchart can be performed in a different order. Additionally or alternatively, certain steps may be omitted, multiple steps may be combined into one step, and / or one step may be broken down into multiple steps.

[0119] It should be understood that when the terms "first," "second," "third," and "fourth," etc., are used in the claims, specification, and drawings of this application, they are used only to distinguish different objects and not to describe a specific order. The terms "comprising" and "including" as used in the specification and claims of this application indicate the presence of the described features, integrals, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components, and / or collections thereof.

[0120] It should also be understood that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the application. As used in this specification and claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in this specification and claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes such combinations.

[0121] Although the embodiments of this application are described above, the content is merely an example adopted for the purpose of facilitating understanding of this application and is not intended to limit the scope and application scenarios of this application. Any person skilled in the art described in this application may make any modifications and changes in the form and details of the implementation without departing from the spirit and scope disclosed in this application, but the scope of patent protection of this application shall still be determined by the scope defined in the appended claims.

Claims

1. A method for optimizing intelligent marketing strategies based on big data, characterized in that, include: Step 1: Use monitoring probes to collect multi-source heterogeneous marketing events across the entire chain, and perform cross-terminal spatial alignment of the collected marketing event sequences based on anonymous consistent identifiers to construct the original time series table; Step 2: Retrieve the historical conversion sample library and statistically analyze the time difference distribution of user conversion behavior, calculate the decay constant reflecting the conversion decay rate, and calibrate the active samples corresponding to the original time series table based on the decay constant. Step 3: Using the attenuation constant as the input parameter of the time attenuation operator, the discrete events in the original time series table after calibration are mapped to construct a feature vector of touch point influence intensity containing spatiotemporal weight information. Step 4: Input the feature vector of the touch point influence intensity into the pre-constructed causal inference network to predict the probability of the user being reached and the response level under different marketing actions, and calculate the net incremental contribution value corresponding to each marketing action. Step 5: Using the net incremental contribution value as the objective function input, and taking privacy budget consumption and budget quota as the joint hard constraint space, solve to obtain the initial marketing strategy including the consumption fluctuation penalty term; Step 6: Monitor the deviation signals between the observed indicators and target indicators at the marketing execution end in real time, calculate the strategy correction increment, and smoothly adjust the initial marketing strategy. Step 7: Send the smoothed and adjusted marketing strategy to the campaign execution engine, and feed back the conversion data after execution to Step 1 to enter the next round of marketing strategy iteration.

2. The intelligent marketing strategy optimization method based on big data according to claim 1, characterized in that, The joint hard constraint space in step six includes: setting an upper limit on the frequency of contact for a single user; when the historical contact frequency of a specific user reaches this upper limit, the initial marketing strategy forcibly modifies the recommendation action for that user to stop intervening.

3. The intelligent marketing strategy optimization method based on big data according to claim 1, characterized in that, The operation of feeding back the conversion data after execution to step one includes: using the real conversion feedback generated by executing the initial marketing strategy as a new historical conversion sample to dynamically update the time difference distribution model in step two, thereby achieving adaptive iteration of the decay constant.

4. The intelligent marketing strategy optimization method based on big data according to claim 1, characterized in that, Step one further includes: Sub-step By capturing multi-source heterogeneous raw data streams generated by user behavior through monitoring probes preset on different marketing channels, and using preset cleaning rules to perform outlier removal and format normalization on the multi-source heterogeneous raw data streams, a standardized marketing event stream containing original user identifiers, behavior trigger times, and touchpoint types is obtained. Sub-step The asymmetric encryption algorithm is invoked to perform one-way desensitization processing on the original user identifiers in the standardized marketing event stream, generating a globally unique anonymous and consistent identifier. The anonymous consistency identifier The formula for generating it is: ， in, For the original user identifier, For any preset perturbation salt value, The XOR operator. For a hash function with collision resistance, the anonymous consistent identifier is used. By linking standardized marketing event streams that are discrete across different terminals, a preliminary aligned set of behaviors can be obtained; Sub-step Based on the initial aligned set of behaviors, spatial alignment rules are used for the same anonymous consistent identifier. Perform time-series aggregation on cross-end behaviors to construct the original time series table. : ,in, For behavior The corresponding contact type description value, For behavior The corresponding behavior trigger time; Sub-step For the original time series table Perform integrity checks; If the judgment condition is met ≤ The original time series table is determined to be a valid sequence. If the judgment condition is met Perform truncation processing, where, This is the preset length of the attribution observation window.

5. The intelligent marketing strategy optimization method based on big data according to claim 1, characterized in that, Step two further includes: Sub-step The system retrieves all historical samples from the historical conversion sample library, extracts the trigger time of marketing touchpoints and the time of final conversion behavior in each historical sample, and calculates the conversion time difference for each sample. The calculation formula is: ,in, The moment when the final conversion occurs. For marketing touchpoint trigger moments; Sub-step Based on the conversion time difference of the single sample The sample set is fitted using the cumulative distribution function to determine the single-sample conversion time difference. median of the distribution And based on the median Calculate the decay constant that reflects the conversion decay rate. The calculation formula is: ,in, For the natural logarithm operator, through the decay constant Characterizes the exponential decay of the transformation probability over time; Sub-step Based on the attenuation constant Set the conversion calibration function Used to calculate the current active sample over the observation period. The conversion probability compensation value, the conversion calibration function The formula is: ,in, The time span from the triggering of the marketing touchpoint to the current observation time for the currently active sample; Sub-step A probability compensation criterion is introduced, and the active samples corresponding to the original time series table are input into the transformation calibration function. Perform calibration processing; If the judgment condition is met The currently active sample is determined to be a highly uncertain sample, and according to the formula... Perform conversion probability compensation calibration; If the judgment condition is met ≥ If the currently active sample is determined to be a stable sample, the observed transformation state of the currently active sample is directly recorded. The preset confidence threshold, The initial transformation value observed so far. This is the calibrated target conversion value.

6. The intelligent marketing strategy optimization method based on big data according to claim 5, characterized in that, Step three further includes: Sub-step Retrieve the valid sequence as determined, and identify each behavior in the valid sequence. Corresponding trigger time Based on the current observation time Calculate the relative delay of each action. The calculation formula is: Through the relative time delay Measure the distance of the influence of each historical point on the current moment; Sub-step Introducing the attenuation constant Construct a kernel function to quantify the effect of contact decay. The calculation formula is: , where the kernel function The output value is determined by the relative time delay. The influence of early touchpoints on current decisions decreases exponentially as they increase; Sub-step Based on the kernel function For the same contact type description value The cumulative influence intensity values ​​of various contact points are calculated by nonlinearly superimposing multiple behaviors. The calculation formula is: ,in, The contact type in the valid sequence belongs to the category A set of behavioral indexes, For the preset corresponding behavior The business importance weighting coefficient, the cumulative influence strength value This comprehensively reflects the combined effect of contact trigger frequency and time decay; Sub-step According to the preset contact category index order, the cumulative influence intensity values ​​of each contact category are calculated. By vectorizing the arrangement, the feature vector of the contact point influence intensity is constructed. And the contact point influence intensity feature vector As input features for causal inference networks, among which, This represents the total number of pre-defined marketing touchpoint categories.

7. The intelligent marketing strategy optimization method based on big data according to claim 6, characterized in that, Step four further includes: Sub-step The constructed contact point influence intensity feature vector Input a pre-defined classifier model to calculate the specific marketing actions that will be performed on the user under the current environmental characteristics. Propensity score The calculation formula is: ,in, As a variable for marketing actions, To assign values ​​to specific marketing actions, To execute specific marketing actions The propensity score is used to characterize the selection bias of the sample; Sub-step A result prediction model is constructed using a regression algorithm, based on the feature vector of the influence intensity of the contact point. Predict users' specific marketing actions First response expectation under the condition And the expected value of the second response under the condition of no marketing intervention. The calculation formula is: , ,in, For the preset conversion indicator result variable, Refers to the expected conversion level under intervention conditions. Refers to the expected conversion level under control conditions; Sub-step Based on the aforementioned tendency score and the expected value of the first response Second response expectation The net incremental contribution value corresponding to each marketing action is calculated using a dual-correction operator. The calculation formula is: ， in, For indicator functions, The net incremental contribution value is the score for the tendency to engage in no marketing intervention. Used to quantify specific marketing actions Compared to the absolute conversion improvement efficiency in the absence of intervention; Sub-step Introducing significance criteria for the calculated net incremental contribution value Perform the screening; if the judgment criteria are met. and If the net incremental contribution value is determined to be a valid incremental signal, and if the determination condition is not met, the net incremental contribution value is determined to be a noise signal and a zeroing process is performed. The preset contribution threshold, This is the preset tendency score cutoff constant.

8. The intelligent marketing strategy optimization method based on big data according to claim 1, characterized in that, Step six further includes: Sub-step The effective incremental signals are retrieved and combined with the updated causal inference network model to predict the target sample set and obtain the marketing actions. Corresponding expected net incremental contribution value Construct a global objective function The formula is: ,in, The initial marketing strategy to be solved. A set of optional marketing actions. To consume the fluctuation penalty item, The preset penalty factor coefficient, This represents the total number of samples to be reached. Sub-step Define the consumption fluctuation penalty term. The calculation formula is: ， in, In the initial marketing strategy Below users The expected costs of intervention, The average budget per user is preset, and the consumption fluctuation penalty item is applied. Constrain the smooth distribution of marketing resources in the sample space; Sub-step A joint hard constraint space is defined, which includes budget quota constraints and cumulative privacy budget consumption constraints. The mathematical expression is: ≤ ≤ ,in, Given a pre-set total marketing budget, the feasible domain of the initial marketing strategy is defined through the joint hard constraint space. Sub-step An optimization criterion is introduced, and the Lagrange multiplier method is used to evaluate the global objective function. Perform the solution within the combined hard constraint space; If the convergence criteria are met Determine the initial marketing strategy. The optimal numerical solution is obtained; if the convergence criterion is not met, the initial marketing strategy is corrected along the gradient ascent direction. The probability distribution parameters are then recalculated, where, Let be the gradient vector of the objective function. This is the preset convergence accuracy threshold.

9. The intelligent marketing strategy optimization method based on big data according to claim 1, characterized in that, Step seven further includes: Sub-step Obtain the initial marketing strategy The optimal numerical solution is obtained and sent to the execution end, and the observation index vector of the marketing execution end is monitored in real time through monitoring probes. Calculate the observation index vector With the preset target indicator vector Deviation signal between The calculation formula is: ,in, At the current monitoring sampling time, the deviation signal is used. It reflects in real time the degree of deviation between the current marketing strategy's execution effect and the expected goals; Sub-step The deviation signal Input a proportional-integral-derivative (PID) controller to calculate the strategy correction increment used to eliminate the deviation signal. The calculation formula is: ， in, For the median variable of the integration time, This is a preset proportional coefficient. The preset integral coefficient, The preset differential coefficients are used to achieve a rapid response to execution fluctuations and the elimination of accumulated errors through a linear combination of the proportional term, integral term, and differential term.

10. The intelligent marketing strategy optimization method based on big data according to claim 9, characterized in that, Step seven further includes: Sub-step Based on the strategy, the increment is corrected. Regarding the initial marketing strategy The optimal numerical solutions are superimposed online to calculate the adjusted marketing strategy. The calculation formula is: ,in, The normalized exponential function is used to correct the increment through the aforementioned strategy. Implement the initial marketing strategy Smooth adjustment, and ensure the marketing strategy after adjustment. It conforms to probability distribution constraints; Sub-step A stability criterion is introduced for the adjusted marketing strategy. Perform compliance verification; if the judgment conditions are met. ≤ If the current strategy adjustment is determined to be in a smooth state, the adjusted marketing strategy will be output. ; If the judgment condition is met If the current strategy is determined to be highly volatile, step size truncation is performed: ,in, The preset maximum step size threshold for a single adjustment. For symbolic functions, This represents the policy state value from the previous moment.

Images