Enterprise-level customer marketing strategy management system and method

Abstract

The invention provides an enterprise-level customer marketing strategy management system, which comprises a data acquisition module, a label generation management module, a strategy generation management module, a strategy execution management module, an effect feedback module and a model optimization module which are connected in sequence, and is characterized in that the model optimization module is connected with the label generation management module; the data acquisition module is used for integrating, cleaning and storing multi-source data; the label generation management module is used for generating a native label and a derivative label; the strategy generation management module is used for dynamically generating a marketing strategy; the strategy execution management module is used for carrying out strategy accurate delivery through intelligent routing selection; the effect feedback module is used for constructing a multi-dimensional evaluation optimization system and performing quantitative feedback on a marketing strategy; and the model optimization module is used for reducing system resource consumption and improving reasoning precision through automatic training. According to the system, accurate customer positioning and continuous model optimization are realized through a hybrid architecture and multi-algorithm fusion modeling, and the marketing conversion rate and the resource utilization rate are improved.

Description

Technical Field

[0001] The present invention relates to the field of communication technology, and in particular to an enterprise-level customer marketing strategy management system and method. Background Art

[0002] In the communications industry, the biggest challenge facing all telecom operators is how to accurately target customers and leverage precision marketing processes and information-based tools to empower sales channels, effectively achieving targeted marketing and accelerating revenue growth. Therefore, in-depth market analysis, the use of data mining techniques to analyze customer tags, and the implementation of precision marketing have become increasingly crucial in an increasingly competitive market.

[0003] Current enterprise customer marketing has problems such as inaccurate customer positioning, unclear customer portraits, inaccurate strategy output, and unreasonable implementation methods. Therefore, it is very necessary to design an enterprise-level customer marketing strategy management system and method. Summary of the Invention

[0004] The purpose of this invention is to provide an enterprise-level customer marketing strategy management system and method, so as to achieve accurate customer positioning and continuous model optimization through hybrid architecture and multi-algorithm fusion modeling, and improve marketing conversion rate and resource utilization.

[0005] To achieve the above object, the present invention provides the following solutions:

[0006] An enterprise-level customer marketing strategy management system includes: a data acquisition module, a label generation management module, a strategy generation management module, a strategy execution management module, an effect feedback module, and a model optimization module; the data acquisition module, the label generation management module, the strategy generation management module, the strategy execution management module, the effect feedback module, and the model optimization module are connected in sequence, and the model optimization module is connected to the label generation management module;

[0007] The data acquisition module is used to integrate, clean and store multi-source data; the label generation management module is used to generate native labels and derived labels; the strategy generation management module is used to dynamically generate marketing strategies; the strategy execution management module is used to accurately deliver strategies through intelligent routing selection; the effect feedback module is used to build a multi-dimensional evaluation and optimization system and provide quantitative feedback on marketing strategies; the model optimization module is used to reduce system resource consumption and improve inference accuracy through automated training.

[0008] Optionally, the data collection module includes a multi-source integration layer, an intelligent cleaning layer, and a storage database connected in sequence, and the storage database is connected to the label generation management module; the storage database includes a customer information database, a business and usage information database, a location information database, an industry and business information database, an order acceptance information database, a business opportunity information database, a project information database, and an industrial and commercial information database, all of which are connected to the intelligent cleaning layer;

[0009] The multi-source integration layer collects multi-source data through the built-in distributed framework and stream processing framework; the intelligent cleaning layer cleans multi-source data through distributed pipelines and random forest algorithms.

[0010] Optionally, the label generation management module includes a label generation unit and a label management unit connected to each other; the label generation unit is connected to the data acquisition module, and the label management unit is connected to the policy generation management module; the label generation unit includes a native label generation subunit and a derived label generation subunit, and both the native label generation subunit and the derived label generation subunit are connected to the data acquisition module;

[0011] The native label generation sub-unit generates native labels by processing basic user information; the derived label generation sub-unit generates derived labels by characterizing user portraits through machine learning algorithms.

[0012] Optionally, the strategy generation management module includes: an intelligent recommendation layer, a strategy optimization layer, a simulation verification unit, and a dynamic pricing engine;

[0013] The intelligent recommendation layer generates fission marketing trigger rules based on native tags and derived tags; the strategy optimization layer evaluates and optimizes the fission marketing trigger rules through the built-in multi-objective particle swarm algorithm; the simulation verification unit predicts the business fluctuation risk of the optimized fission marketing trigger rules through AB testing; the dynamic pricing engine is used to generate marketing strategies based on business fluctuation risks.

[0014] Optionally, the policy execution management module includes: an omni-channel reach layer, an execution monitoring hub, and a dynamic tuning unit;

[0015] The omni-channel reach layer transmits marketing strategies to target enterprises through a built-in routing selection mechanism; the execution monitoring center is used to monitor data on routing transmission channels; and the dynamic tuning unit adjusts the weights of routing transmission channels through a built-in multi-armed bandit algorithm.

[0016] Optionally, the effect feedback module includes: an attribution analysis layer and a visualization insight layer;

[0017] The attribution analysis layer measures the incremental benefits of marketing strategies through the built-in causal forest model; the visualization insight layer calculates the quantitative feedback of incremental benefits through the built-in real-time feedback network.

[0018] Optionally, the model optimization module includes: an automated training layer, an incremental learning engine, and a federated learning layer;

[0019] The automated training layer is used to generate the optimal model structure through quantitative feedback; the incremental learning engine updates the model parameters of the optimal model structure through the built-in EWC algorithm; and the federated learning layer expands the feature dimensions of the updated optimal model structure through the built-in homomorphic encryption algorithm.

[0020] An enterprise-level customer marketing strategy management method, applied to the above-mentioned enterprise-level customer marketing strategy management system, comprises the following steps:

[0021] Through the constructed management system, we collect multi-source data from customers and integrate them to obtain the original data;

[0022] Based on the original data, native labels and derived labels are generated through machine learning algorithms;

[0023] Generate marketing strategies based on native and derived tags;

[0024] Execute marketing strategies for target companies and build a multi-dimensional evaluation and optimization system to evaluate and quantify feedback from target companies;

[0025] The management system model is optimized through feedback information to obtain an optimized model.

[0026] Optionally, based on the original data, native labels and derived labels are generated through machine learning algorithms, including:

[0027] Perform data cleaning on the original data to remove duplicate values, outliers and null values ​​in the original data;

[0028] Perform standard normalization operation on the cleaned raw data to obtain standard data;

[0029] The standard data is discretized by the WOE algorithm to obtain discrete features;

[0030] Perform feature selection on discrete features according to a preset threshold to obtain screening features;

[0031] The screening features are predicted based on the preset machine learning model to obtain native labels and derived labels.

[0032] Optionally, the machine learning model is any one of a support vector machine model, a k-nearest neighbor model, a random forest model or a gradient boosting decision tree model.

[0033] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects: an enterprise-level customer marketing strategy management system, comprising: a data acquisition module, a label generation management module, a strategy generation management module, a strategy execution management module, an effect feedback module and a model optimization module; the data acquisition module, the label generation management module, the strategy generation management module, the strategy execution management module, the effect feedback module and the model optimization module are connected in sequence, and the model optimization module is connected to the label generation management module. The data acquisition module is used to integrate, clean and store multi-source data; the label generation management module is used to generate native labels and derivative labels; the strategy generation management module is used to dynamically generate marketing strategies; the strategy execution management module is used to accurately deliver strategies through intelligent routing selection; the effect feedback module is used to build a multi-dimensional evaluation and optimization system and provide quantitative feedback on marketing strategies; the model optimization module is used to reduce system resource consumption and improve reasoning accuracy through automated training. The system achieves precise customer positioning and continuous model optimization through hybrid architecture and multi-algorithm fusion modeling, thereby improving marketing conversion rate and resource utilization. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0035] Figure 1 This is a structural diagram of an enterprise-level customer marketing strategy management system according to an embodiment of the present invention;

[0036] Figure 2 This is a flow chart of the enterprise-level customer marketing strategy management method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0037] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0038] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0039] like Figure 1As shown, the present invention provides an enterprise-level customer marketing strategy management system, comprising: a data acquisition module, a label generation management module, a strategy generation management module, a strategy execution management module, an effect feedback module, and a model optimization module; the data acquisition module, the label generation management module, the strategy generation management module, the strategy execution management module, the effect feedback module, and the model optimization module are connected in sequence, and the model optimization module is connected to the label generation management module;

[0040] The data acquisition module is used to integrate, clean and store multi-source data; the label generation management module is used to generate native labels and derived labels; the strategy generation management module is used to dynamically generate marketing strategies; the strategy execution management module is used to accurately deliver strategies through intelligent routing selection; the effect feedback module is used to build a multi-dimensional evaluation and optimization system and provide quantitative feedback on marketing strategies; the model optimization module is used to reduce system resource consumption and improve inference accuracy through automated training.

[0041] Specifically, the data collection module includes a multi-source integration layer, an intelligent cleaning layer, and a storage database, which are connected in sequence. The storage database is connected to the label generation and management module. The data collection module adopts a hybrid architecture. The multi-source integration layer uses Kafka and Flink technologies to capture high-concurrency streaming data such as 5G signaling and customer service logs in real time, processing out-of-order events in milliseconds. The built-in Scrapy-Redis distributed crawler combines Selenium dynamic rendering technology to achieve a breakthrough in anti-crawling mechanisms to collect competitive intelligence. It also uses a hybrid gateway combining RESTful APIs and GraphQL to significantly improve response efficiency. The built-in NB-IoT / LoRaWAN Internet of Things protocol enables the collection of data from physical spaces such as business hall heating and work badge tracks.

[0042] The intelligent cleaning layer builds a distributed ETL pipeline based on Spark SQL and uses the AC automaton and Levenshtein algorithms to correct data spelling errors. Its built-in Talend visualization platform handles missing values ​​using integrated KNN and MissForest algorithms, and uses a regular rule library and isolation forest model to detect anomalies, achieving an anomaly identification accuracy rate of up to 93%. A built-in integrated stream and batch quality monitoring system also significantly improves data availability. Finally, the cleaned data is stored in multiple databases, including a customer information database, a business and usage information database, a location information database, an industry and operations information database, an order acceptance information database, a business opportunity information database, a project information database, and a business information database.

[0043] Specifically, the label generation management module includes a label generation unit and a label management unit connected to each other; the label generation unit is connected to the data acquisition module, and the label management unit is connected to the policy generation management module; the label generation unit includes a native label generation subunit and a derived label generation subunit, and both the native label generation subunit and the derived label generation subunit are connected to the data acquisition module;

[0044] The native label generation sub-unit generates native labels by processing basic user information; the derived label generation sub-unit generates derived labels by characterizing user portraits through machine learning algorithms.

[0045] It should be noted that native tags are processed by extracting basic user information based on certain business experience rules, and are characterized by strong interpretability and low complexity. Derived tags are based on enterprise-level user data such as user attributes, user behavior, billing data, base station trajectories, and internet preferences. Multi-dimensional user features are extracted through data cleaning, data normalization, feature binning, and feature selection operations. The multi-dimensional tags are then generated by using machine learning data mining algorithms such as support vector machines, k-nearest neighbors, random forests, or gradient boosting decision trees to characterize user portraits. Information such as the tag name, category, corresponding code table, business caliber, and technical caliber are then entered into the tag management unit for management and to generate a tag log including historical tag modification records.

[0046] Specifically, the strategy generation management module includes: an intelligent recommendation layer, a strategy optimization layer, a simulation verification unit, and a dynamic pricing engine; the intelligent recommendation layer generates fission marketing trigger rules based on native tags and derived tags; the strategy optimization layer evaluates and optimizes the fission marketing trigger rules through the built-in multi-objective particle swarm algorithm; the simulation verification unit predicts the business fluctuation risk of the optimized fission marketing trigger rules through AB testing; the dynamic pricing engine is used to generate marketing strategies based on business fluctuation risks.

[0047] It should be noted that the strategy generation and management module integrates intelligent algorithms and simulation technology to achieve accurate output of marketing strategies. The intelligent recommendation layer uses the DDPG deep reinforcement learning framework to dynamically optimize resource allocation strategies, and combines the Wide & Deep model to generate personalized package combinations. At the same time, it analyzes social communication paths based on the GNN graph neural network to design exclusive marketing strategies for target companies. The strategy optimization layer maintains the dynamic balance between ARPU improvement and customer churn risk through the built-in multi-objective particle swarm algorithm, and then evaluates the cumulative effect of marketing strategies through causal inference models. The simulation verification unit is built with an Anylogic digital twin environment, conducts AB testing through historical customer behavior data streams, and combines Monte Carlo simulation to predict business volatility risks. The dynamic pricing engine integrates a game theory model, which can respond to market competition in real time, and combines the LSTM model to predict network load in the next 24 hours, while automatically generating elastic discount coefficients.

[0048] Furthermore, the Strategy Generation Management module uses AutoML automated modeling to shorten the strategy generation cycle to 4 hours, increasing marketing conversion rates while reducing customer complaints. This module also includes a new Strategy Unit and a Strategy Management Unit. The new Strategy Unit supports the system page entry of basic strategy information and the adaptation of target customer groups. Strategy basic information includes: strategy code, strategy name, strategy execution frequency, strategy type, execution time, strategy description, and strategy rules. It also supports manual entry and uploading of files in various formats. The Strategy Management Unit uses scheduled tasks to set the time and frequency of marketing strategy task issuance, supporting settings based on daily, weekly, monthly, quarterly, and annual schedules. The strategy's target customer group is adapted based on the business context. By cross-combining different tags and setting tag values ​​and tag ranges, target customers for the strategy are identified, and marketing strategies can be customized for these target customers. Customer information, task information, feedback information, details, and user details are visualized through dynamic SQL and page configuration, which supports SQL statement recognition. At the same time, the page configuration supports settings such as whether to list the field, whether to export the field, whether to query the field, whether to form the field, whether it is required, the query method, the display type, the field type, etc. It also supports the modification of the specified policy, such as basic information, scheduled tasks and dynamic SQL page configuration.

[0049] Specifically, the strategy execution management module includes: an omni-channel reach layer, an execution monitoring center, and a dynamic tuning unit; the omni-channel reach layer transmits marketing strategies to target enterprises through a built-in routing selection mechanism; the execution monitoring center is used to monitor data on routing transmission channels; and the dynamic tuning unit adjusts the weights of routing transmission channels through a built-in multi-armed bandit algorithm.

[0050] It should be noted that the policy execution management module has built an intelligent closed-loop execution system. The omnichannel reach layer integrates an intelligent routing engine, which uses reinforcement learning to dynamically select the optimal channel for SMS, WeChat, or outbound calls. Outbound calls are driven by ASR+TTS speech synthesis technology, which improves the accuracy of call emotion recognition. The execution monitoring center combines Flink and Prometheus to build a real-time feedback stream to track key indicators such as channel response rate and resource utilization. When key indicators are abnormal, the circuit breaker mechanism is automatically triggered; and the Seata framework is used for distributed transaction control to ensure cross-system operation consistency. The dynamic tuning unit uses a multi-armed bandit algorithm to adjust channel weights in real time, combines knowledge graph analysis to analyze execution bottlenecks to improve response speed, and uses digital watermarking technology to track policy leakage paths to ensure execution security.

[0051] Specifically, the effect feedback module includes: an attribution analysis layer and a visual insight layer; the attribution analysis layer measures the incremental benefits of marketing strategies through a built-in causal forest model; the visual insight layer calculates quantitative feedback of incremental benefits through a built-in real-time feedback network.

[0052] It should be noted that the attribution analysis layer uses the ShapleyValue algorithm to quantify omnichannel contributions, combined with the double difference method to separate organic growth from the actual effectiveness of strategies, and uses a built-in causal forest model to calculate the incremental benefits of marketing strategies. The visualization insight layer uses Tableau to build a dynamic war room, integrating and visualizing conversion funnel charts, customer LTV heat maps, and NPS sentiment analysis word clouds. This module also has a built-in real-time feedback network that calculates response rate indicators based on Flink window functions. When the indicators are abnormal, it triggers an alert and links the strategy execution module to perform a transmission cut-off operation. This module significantly reduces the measurement error of marketing strategies, shortening the strategy iteration cycle from 7 days to 48 hours, thereby improving iteration speed. It can also automatically generate eight types of effect analysis reports through AutoML, improving decision-making efficiency.

[0053] Specifically, the model optimization module includes: an automated training layer, an incremental learning engine, and a federated learning layer; the automated training layer is used to generate the optimal model structure through quantitative feedback; the incremental learning engine updates the model parameters of the optimal model structure through the built-in EWC algorithm; the federated learning layer expands the feature dimensions of the updated optimal model structure through the built-in homomorphic encryption algorithm.

[0054] It should be noted that the automated training layer uses NAS neural architecture search technology to automatically generate the optimal model structure, and then dynamically adjusts hyperparameters based on TPE Bayesian optimization, improving parameter adjustment efficiency. The incremental learning engine has a built-in TensorFlow Federated framework, which supports online sample streaming to update model parameters and combines the Elastic Weight Consolidation (EWC) algorithm to control catastrophic forgetting. The federated learning layer uses Paillier homomorphic encryption to collaboratively model cross-provincial company data, thereby constructing a vertical federated GBDT model, expanding the feature dimension to over 1500 while protecting privacy.

[0055] Furthermore, the model optimization module also includes a model compression unit, which compresses the model to 1 / 5 of its original volume through knowledge distillation technology, increasing the inference speed to 300% with an accuracy loss of less than 0.8%.

[0056] like Figure 2 As shown, the present invention also provides an enterprise-level customer marketing strategy management method, which is applied to the above-mentioned enterprise-level customer marketing strategy management system, including the following steps:

[0057] Step 100: Collect the customer's multi-source data through the constructed management system and integrate them to obtain the original data;

[0058] Step 200: Generate native labels and derived labels based on the original data through a machine learning algorithm;

[0059] Specifically, the native tags generated in this embodiment include: factory area, monthly video conferencing traffic growth rate, and industry, and the derived tags are customer level. Customer level tags include four categories: strategic customers, important customers, general customers, and small and micro enterprises. The steps for generating tags are as follows:

[0060] First, based on the original data such as customer attributes, consumption behavior, bill data, payment details, and the proportion of bad users, basic data features are extracted and trend data features such as standard deviation, growth rate, and dispersion coefficient are processed, and duplicate values, outliers, and null values ​​in the data features are removed. Then the Z-score normalization method is used to eliminate the data dimension so that the data features obey the normal distribution after processing to speed up the convergence of the algorithm. The WOE algorithm is then used to discretize the features and calculate the proportion of good and bad user responses in the feature segments. The WOE calculation result is multiplied by the weight of each feature to obtain the IV value of the feature. The IV value represents the contribution of the feature to target detection. This embodiment selects features with IV values ​​> 0.3 for modeling. Finally, native labels and derived labels are generated through a preset machine learning model.

[0061] Furthermore, the machine learning model is any one of a support vector machine model, a k-nearest neighbor model, a random forest model, or a gradient boosted decision tree model. This embodiment uses a random forest model to make predictions by integrating multiple decision trees, and then uses voting to synthesize the prediction results of multiple decision trees to generate native labels and derived labels.

[0062] Step 300: Generate a marketing strategy based on the original tag and the derived tag;

[0063] This example screens customers with a factory area greater than 500,000 square meters, a monthly video conferencing traffic growth rate greater than 15%, high-end manufacturing as an industry, and strategic customer status, and generates a targeted marketing strategy. The entered policy information includes: Policy Code: 23; Policy Name: 5G Private Network Precision Recommendation Strategy; Policy Execution Frequency: Monthly; Policy Type: Market Expansion; Execution Time: 8:00 AM on the 12th of each month; Policy Description: The account manager will conduct 5G private network marketing for customers with a factory area greater than 500,000 square meters, a monthly video conferencing traffic growth rate greater than 15%, high-end manufacturing as an industry, and strategic customer status. Policy Rules: Push to the account manager. If the customer is a list-based customer, the account manager is identified. If not, the grid calculates the customer's dual-line business revenue and pushes the task to the grid CEO with the highest revenue contribution. Scheduled Task: Set the 5G private network recommendation strategy task to execute at 8:00 AM on the 12th of each month. Customer information, task information, feedback information, details, and user details are all visualized on the page using dynamic SQL.

[0064] Step 400: Implement marketing strategies for target enterprises and build a multi-dimensional evaluation and optimization system to evaluate and quantify feedback from target enterprises;

[0065] In this embodiment, the marketing strategy is sent to the customer manager via SMS to implement the marketing strategy.

[0066] Step 500: Optimize the management system model through feedback information to obtain an optimized model.

[0067] The beneficial effects of the present invention are as follows:

[0068] 1) The system architecture is based on a scalable distributed and cloud computing architecture, constructed based on cloud-native technologies and a three-tier cloud architecture. It highlights the service integration capabilities based on microservices and service mesh, and achieves rapid response in industrial real-world scenarios through continuous integration and continuous release.

[0069] 2) The overall system adopts a front-end and back-end separation architecture, enabling rapid and frequent build, release, and deployment. Combined with cloud computing, it decouples the system from the underlying hardware and operating system, improving scalability, availability, portability, and cost-effectiveness.

[0070] 3) The system has a simple structure and supports sub-functions such as policy rule management, policy timeliness management, policy task management, task object management, and task feedback management. It enables the configuration of government and enterprise marketing strategies and can effectively support the issuance of personalized business rules for governments and enterprises.

[0071] 4) The system builds a closed-loop precision marketing work scenario system based on key scenarios such as strategy configuration, work order dispatching, and execution evaluation, thereby improving the work enthusiasm and quality of frontline account managers.

[0072] 5) Intelligent models and optimization algorithms such as machine learning improve the accuracy of label prediction, making strategy execution more accurate and data analysis more efficient.

[0073] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0074] The present invention uses specific examples to illustrate the principles and implementation methods of the present invention. The above examples are only intended to help understand the method and core concept of the present invention. At the same time, those skilled in the art will find that the specific implementation methods and application scopes may vary based on the concept of the present invention. In summary, the contents of this specification should not be construed as limiting the present invention.

Claims

1. An enterprise-level customer marketing strategy management system, characterized in that: include: A data acquisition module, a label generation management module, a policy generation management module, a policy execution management module, an effect feedback module, and a model optimization module; the data acquisition module, the label generation management module, the policy generation management module, the policy execution management module, the effect feedback module, and the model optimization module are connected in sequence, and the model optimization module is connected to the label generation management module; The data acquisition module is used to integrate, clean and store multi-source data; The tag generation management module is used to generate native tags and derived tags; the strategy generation management module is used to dynamically generate marketing strategies; The policy execution management module is used to accurately deliver policies through intelligent routing selection; The effect feedback module is used to build a multi-dimensional evaluation and optimization system and provide quantitative feedback on the marketing strategy; the model optimization module is used to reduce system resource consumption and improve reasoning accuracy through automated training.

2. The enterprise-level customer marketing strategy management system according to claim 1, characterized in that: The data acquisition module includes a multi-source integration layer, an intelligent cleaning layer, and a storage database connected in sequence, and the storage database is connected to the label generation management module; the storage database includes a customer information database, a business and usage information database, a location information database, an industry and business information database, an order acceptance information database, a business opportunity information database, a project information database, and an industrial and commercial information database, all of which are connected to the intelligent cleaning layer; The multi-source integration layer collects the multi-source data through a built-in distributed framework and stream processing framework; the intelligent cleaning layer cleans the multi-source data through a distributed pipeline and a random forest algorithm.

3. The enterprise-level customer marketing strategy management system according to claim 1, characterized in that: The label generation management module includes a label generation unit and a label management unit connected to each other; the label generation unit is connected to the data acquisition module, and the label management unit is connected to the policy generation management module; the label generation unit includes a native label generation subunit and a derived label generation subunit, and both the native label generation subunit and the derived label generation subunit are connected to the data acquisition module; The native tag generation subunit generates the native tag by processing the user basic information; The derived tag generation subunit generates the derived tags by characterizing the user portrait through a machine learning algorithm.

4. The enterprise-level customer marketing strategy management system according to claim 1, characterized in that: The strategy generation management module includes: an intelligent recommendation layer, a strategy optimization layer, a simulation verification unit and a dynamic pricing engine; The intelligent recommendation layer generates fission marketing trigger rules based on the native tags and the derived tags; the strategy optimization layer evaluates and optimizes the fission marketing trigger rules through the built-in multi-objective particle swarm algorithm; the simulation verification unit predicts the business fluctuation risk of the optimized fission marketing trigger rules through AB testing; the dynamic pricing engine is used to generate the marketing strategy based on the business fluctuation risk.

5. The enterprise-level customer marketing strategy management system according to claim 1, characterized in that: The strategy execution management module includes: an omni-channel reach layer, an execution monitoring hub, and a dynamic tuning unit; The omni-channel reach layer transmits the marketing strategy to the target enterprise through the built-in routing selection mechanism; the execution monitoring center is used to monitor the data of the routing transmission channel; the dynamic tuning unit adjusts the weight of the routing transmission channel through the built-in multi-armed bandit algorithm.

6. The enterprise-level customer marketing strategy management system according to claim 1, characterized in that: The effect feedback module includes: an attribution analysis layer and a visualization insight layer; The attribution analysis layer measures the incremental benefits of the marketing strategy through a built-in causal forest model; the visualization insight layer calculates the quantitative feedback of the incremental benefits through a built-in real-time feedback network.

7. The enterprise-level customer marketing strategy management system according to claim 1, characterized in that: The model optimization module includes: an automated training layer, an incremental learning engine, and a federated learning layer; The automated training layer is used to generate the optimal model structure through the quantized feedback; the incremental learning engine updates the model parameters of the optimal model structure through the built-in EWC algorithm; and the federated learning layer expands the feature dimensions of the updated optimal model structure through the built-in homomorphic encryption algorithm.

8. An enterprise-level customer marketing strategy management method, applied to the enterprise-level customer marketing strategy management system according to any one of claims 1 to 7, characterized in that: The steps include: Through the constructed management system, we collect multi-source data from customers and integrate them to obtain the original data; Based on the raw data, generating native labels and derived labels through a machine learning algorithm; generating a marketing strategy based on the native tag and the derived tag; Execute the marketing strategy on the target enterprise and build a multi-dimensional evaluation and optimization system to evaluate and quantify the feedback information of the target enterprise; The management system is model optimized using the feedback information to obtain an optimized model.

9. The enterprise-level customer marketing strategy management method according to claim 8, characterized in that: Based on the raw data, native labels and derived labels are generated through machine learning algorithms, including: Performing data cleaning on the raw data to remove duplicate values, abnormal values ​​and null values ​​in the raw data; Performing a standard normalization operation on the cleaned raw data to obtain standard data; Discretize the standard data using the WOE algorithm to obtain discrete features; Performing feature selection on the discrete features according to a preset threshold to obtain screening features; The screening features are predicted according to a preset machine learning model to obtain the native label and the derived label.

10. The enterprise-level customer marketing strategy management method according to claim 9, characterized in that: The machine learning model is any one of a support vector machine model, a k-nearest neighbor model, a random forest model or a gradient boosting decision tree model.

Images