An AI automation precision marketing recommendation method and system based on big data

Abstract

The application relates to the technical field of data processing, and discloses an AI automatic precision marketing recommendation method and system based on big data, which collects user basic data, behavior data, enterprise attribute data and marketing interaction data, carries out data storage and distributed calculation based on a Hadoop distributed framework, constructs a standardized data sample set after preprocessing, carries out preliminary customer grouping by adopting a K-means clustering algorithm, constructs a three-level label system, adopts dynamic feature fusion and an attention mechanism to realize real-time updating of customer portraits, constructs a demand prediction model to predict customer interest degrees, and outputs customer intention scores; in combination with the output result of the demand prediction model, a collaborative filtering algorithm is adopted to generate personalized marketing content, and an optimal push channel and push time are determined; personalized pushing is executed, customer interaction feedback data is collected, and a recommendation strategy is optimized based on the customer interaction feedback data; and the application improves marketing conversion rate and reduces customer acquisition cost.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, specifically to an AI-automated precision marketing recommendation method and system based on big data. Background Technology

[0002] With the rapid development of the digital economy, competition in the SaaS (Software as a Service) industry is becoming increasingly fierce, and rising customer acquisition costs have become a common pain point in the industry. Traditional SaaS marketing often adopts a broad-based approach, relying on manual lead screening and mass marketing content distribution. When faced with petabytes of user behavior data and enterprise attribute data, traditional databases are prone to storage crashes and low computing efficiency, making it impossible to achieve efficient integration and analysis of multi-source data. Existing methods are mostly based on static tags for customer classification, without considering the dynamic changes in user behavior and the correlation of multi-dimensional features, resulting in a mismatch between marketing content and customer needs, poor user experience, and low conversion rates. Marketing recommendation strategies are rigid and cannot be optimized in real time based on customer feedback, making it difficult to balance recommendation accuracy and diversity. This makes it difficult to accurately capture high-intent customers and uncover potential customer needs. Summary of the Invention

[0003] The purpose of this invention is to solve the above problems by designing an AI-automated precision marketing recommendation method and system based on big data.

[0004] The first aspect of this invention provides an AI-automated precision marketing recommendation method based on big data, the method comprising the following steps: Collect user basic data, behavioral data, enterprise attribute data and marketing interaction data, perform data storage and distributed computing based on the Hadoop distributed framework, and construct a standardized data sample set after preprocessing; The K-means clustering algorithm is used to initially segment customers and construct a three-level labeling system. Dynamic feature fusion and attention mechanisms are used to update customer profiles in real time. Construct a demand prediction model that integrates deep reinforcement learning and Transformer models to predict customer interest and output customer intention scores; Based on the results of the demand forecasting model, a collaborative filtering algorithm is used to generate personalized marketing content and determine the optimal push channel and push time. Perform personalized push notifications and collect customer interaction feedback data, then optimize recommendation strategies based on this data.

[0005] Optionally, in the first implementation of the first aspect of the present invention, the step of using the K-means clustering algorithm for initial customer segmentation, constructing a three-level labeling system, and employing dynamic feature fusion and attention mechanisms for real-time updating of customer profiles includes: Using a standardized data sample set as input, the initial parameters for customer segmentation and the range of the number of clusters are determined. The initial parameters for customer segmentation include the particle swarm size, the maximum number of iterations, the acceleration factor, and the inertia weight. The position of each particle is defined as a set of cluster centroids, and each particle corresponds to a potential customer segmentation scheme. The particle swarm optimization algorithm was used for iterative processing to obtain preliminary customer segmentation results, which divided customers into four categories: high-intent customers, nurturing customers, potential customers, and low-intent customers. Based on the initial customer segmentation results, a three-level tagging system consisting of basic tags, behavioral tags, and intention tags was established. Collect real-time and historical customer behavior data, perform standardized alignment processing, and filter out effective features related to customer needs and segmentation characteristics; Weights are assigned to the preprocessed real-time and historical behavioral features, and customer feature vectors are generated through weighted fusion. The customer feature vector is matched with the customer's current tagging system to update the customer's behavioral tags and intention tags. The feature fusion results and tag information are dynamically updated to update the customer profile in real time.

[0006] Optionally, in a second implementation of the first aspect of the present invention, the step of using a particle swarm optimization algorithm for iterative processing to obtain preliminary customer segmentation results divides customers into four categories: high-intent customers, nurturing customers, potential customers, and low-intent customers, including: The particle swarm optimization algorithm performs initial iterations on the initialized particle swarm to explore the solution space of the customer group; Particles are divided into two categories based on their fitness values: optimized and improved. Optimized particles are adjusted using the K-means algorithm. If an optimized particle gets stuck in a local convergence state, a Levy flight perturbation is applied to a single dimension randomly selected by it. After generating clones through cloning operations, the improved particles undergo Gaussian mutation. If empty clusters occur during the iteration process, an empty cluster allocation strategy is adopted to reassign the samples farthest from the centroid in the non-empty clusters to the empty clusters, update the cluster centroids, and continue until the iteration is completed to obtain the customer's preliminary clustering results.

[0007] Optionally, in a third implementation of the first aspect of the present invention, the construction of a demand prediction model that integrates deep reinforcement learning and a Transformer model to predict customer interest and output a customer intention score includes: The customer feature vector obtained after dynamic construction of customer profiles is used as input data, and the time-series behavioral data in the customer feature vector is input into the Transformer model; By leveraging the multi-head attention mechanism of the Transformer model, we can capture the correlation between customer behaviors across different time dimensions, extract deep temporal features of customer behavior sequences, and output a comprehensive customer feature vector. The customer's comprehensive feature vector is used as the state input of the DRL model. The action space of the DRL model is defined as the recommendation combination of different SaaS products and marketing content. Each action corresponds to a recommendation scheme. Positive interaction behavior of customers with the recommended content is set as positive reward, negative interaction behavior is set as negative reward, and no interaction behavior is set as zero reward. The DDPG algorithm is used for iterative training. During the training process, the DRL model outputs the corresponding recommended action based on the current input customer comprehensive feature vector. Combined with the customer interaction results fed back by the reward function, the action selection strategy is optimized. Input the comprehensive feature vector of the customer to be predicted into the trained demand prediction model, and output the customer’s interest in different SaaS products and marketing content to form a set of customer interest scores. Determine the multi-dimensional indicators for customer intention scoring, which include customer interest score, company size, industry matching degree, interaction frequency, and historical transaction similarity. Based on multi-dimensional indicators and weights, each indicator of the customer to be predicted is quantitatively scored, and the final intention score of the customer is obtained by weighted summation.

[0008] Optionally, in the fourth implementation of the first aspect of the present invention, the step of inputting the comprehensive feature vector of the customer to be predicted into the trained demand prediction model, and outputting the customer's interest in different SaaS products and marketing content to form a customer interest score set includes: Input the comprehensive feature vector of the customer to be predicted into the trained demand prediction model, and output the comprehensive feature vector of the customer through the Transformer model; The state of the prediction environment is initialized using the customer's comprehensive feature vector as the current state of the DRL model. Based on the current customer status and the trained policy network parameters, the Actor network outputs the probability distribution of actions for different SaaS products and marketing content, with each action corresponding to a specific recommendation strategy. The action probability distribution output by the Actor network is mapped to the value rating system of the corresponding SaaS products and marketing content, and the action probability is converted into the customer's specific interest rating for each product and content, forming a preliminary set of interest ratings. The initial interest score set is deduplicated and then categorized and aggregated according to SaaS product category and marketing content type to obtain the customer interest score set. By combining the intention tag information in the customer profile, the correlation of the aggregated customer interest score set is calibrated to generate the final customer interest score set.

[0009] Optionally, in a fifth implementation of the first aspect of the present invention, the step of combining the results output by the demand forecasting model, using a collaborative filtering algorithm to generate personalized marketing content, and determining the optimal push channel and push time includes: Based on the output of the demand forecasting model, we extract the types of SaaS products and marketing content categories that customers are interested in. At the same time, we combine the static and dynamic features in the customer profile to select feature dimensions that are highly relevant to customer needs. Construct a user and item rating matrix, where the rows of the matrix represent customers, the columns represent the types of SaaS products and marketing content that can be recommended, and the matrix elements are the corresponding interest scores output by the demand prediction model. Based on user and item rating matrices, Pearson correlation coefficients are calculated among customers to identify the nearest customer groups most similar to the interest characteristics of the customers to be predicted. Content similarity between various marketing content is also calculated to mine the relationships between content and form customer similarity sets and content similarity sets. By combining customer similarity sets and content similarity sets, a personalized marketing content candidate set containing different types and focuses is generated; The candidate set of personalized marketing content is matched with the optimal push channels and timing plans, and the marketing content is adjusted to form the final personalized marketing content and push plan.

[0010] Optionally, in a sixth implementation of the first aspect of the present invention, the step of combining the results output by the demand forecasting model, using a collaborative filtering algorithm to generate personalized marketing content, and determining the optimal push channel and push time further includes: Collect the attribute characteristics of push channels, including channel reach, response speed, and user acceptance. By combining the intent levels in the customer profile, channel selection priorities are set for high-intent customers, nurturing customers, and potential customers. The optimal push time is determined based on customer activity time data, and finally, the optimal push channel and time plan is generated.

[0011] A second aspect of this invention provides an AI-automated precision marketing recommendation system based on big data, the system comprising: The data processing module is used to collect user basic data, behavioral data, enterprise attribute data and marketing interaction data. It performs data storage and distributed computing based on the Hadoop distributed framework, and constructs a standardized data sample set after preprocessing. The real-time update module is used to perform initial customer segmentation using the K-means clustering algorithm, construct a three-level label system, and use dynamic feature fusion and attention mechanisms to update customer profiles in real time. The demand forecasting module is used to build a demand forecasting model that integrates deep reinforcement learning and Transformer models to predict customer interest and output customer intention scores. The recommendation generation module combines the results of the demand forecasting model with the collaborative filtering algorithm to generate personalized marketing content and determine the optimal push channels and push times. The push module is used to execute personalized push notifications and collect customer interaction feedback data, and optimize recommendation strategies based on the customer interaction feedback data.

[0012] A third aspect of the present invention provides an AI-automated precision marketing recommendation device based on big data, the AI-automated precision marketing recommendation device based on big data including a memory and at least one processor, the memory storing instructions; the at least one processor calling the instructions in the memory to cause the AI-automated precision marketing recommendation device based on big data to perform the various steps of the AI-automated precision marketing recommendation method based on big data as described in any of the preceding claims.

[0013] A fourth aspect of the present invention provides a computer-readable storage medium storing instructions that, when executed by a processor, implement the steps of the AI-automated precision marketing recommendation method based on big data as described in any of the preceding claims.

[0014] The technical solution provided by this invention collects basic user data, behavioral data, enterprise attribute data, and marketing interaction data. Data storage and distributed computing are performed based on the Hadoop distributed framework, and a standardized data sample set is constructed after preprocessing. K-means clustering algorithm is used for initial customer segmentation, constructing a three-level label system. Dynamic feature fusion and attention mechanisms are employed for real-time updates of customer profiles. A demand prediction model integrating deep reinforcement learning and Transformer models is constructed to predict customer interest and output customer intention scores. Based on the results of the demand prediction model, a collaborative filtering algorithm is used to generate personalized marketing content and determine the optimal push channel and push time. Personalized pushes are executed, and customer interaction feedback data is collected. Recommendation strategies are optimized based on this customer interaction feedback data. This invention solves the problems of low accuracy, low efficiency, and lack of dynamic optimization capabilities in traditional SaaS marketing, improving marketing conversion rates, reducing customer acquisition costs, breaking the limitations of traditional static profiles, accurately capturing dynamic changes in customer needs, and balancing recommendation accuracy and diversity. Attached Figure Description

[0015] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention.

[0016] Figure 1 A flowchart illustrating the AI-automated precision marketing recommendation method based on big data provided in this embodiment of the invention; Figure 2 A schematic diagram of the structure of an AI-automated precision marketing recommendation system based on big data, provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the structure of an AI-automated precision marketing recommendation device based on big data, provided in an embodiment of the present invention. Detailed Implementation

[0017] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a particular order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms “comprising” or “having,” and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, apparatus, product, or device that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0018] For ease of understanding, the specific process of the embodiments of the present invention is described below. Please refer to [link / reference]. Figure 1 The flowchart of the AI-automated precision marketing recommendation method based on big data provided in this embodiment of the invention includes the following steps: Step 101: Collect user basic data, behavioral data, enterprise attribute data and marketing interaction data, perform data storage and distributed computing based on the Hadoop distributed framework, and construct a standardized data sample set after preprocessing; Step 102: Use the K-means clustering algorithm to perform preliminary customer segmentation, construct a three-level labeling system, and use dynamic feature fusion and attention mechanisms to update customer profiles in real time. In this embodiment, a standardized data sample set is used as input to determine the initial parameters for customer segmentation and the range of the number of clusters. The initial parameters for customer segmentation include the particle swarm size, maximum number of iterations, acceleration factor, and inertia weight. The position of each particle is defined as a set of cluster centroids, and each particle corresponds to a potential customer segmentation scheme. The particle swarm algorithm is used for iterative processing to obtain the preliminary customer segmentation results, dividing customers into four categories: high-intent customers, nurturing customers, potential customers, and low-intent customers. Based on the preliminary customer segmentation results, a three-level tag system consisting of basic tags, behavioral tags, and intent tags is built. Real-time and historical customer behavioral data are collected and standardized and aligned to filter out effective features related to customer needs and segmentation characteristics. The preprocessed real-time and historical behavioral features are weighted and weighted to generate customer feature vectors. The customer feature vectors are matched with the current customer tag system to update the customer's behavioral and intent tags, dynamically updating the feature fusion results and tag information to achieve real-time updates of the customer profile.

[0019] In this embodiment, the particle swarm optimization algorithm performs initial iterations on the initialized particle swarm to explore the solution space of customer grouping. The particles are divided into two categories, optimized and improved, according to their fitness values. Optimized particles are adjusted using the K-means algorithm. If an optimized particle gets stuck in a local convergence state, a Lévy flight perturbation is applied to its randomly selected single dimension. Improved particles generate clones through cloning operations and then undergo Gaussian mutation. If empty clusters occur during the iteration process, an empty cluster allocation strategy is adopted to reassign the sample farthest from the centroid in the non-empty cluster to the empty cluster, update the cluster centroid, and continue until the iteration is completed to obtain the initial customer grouping results.

[0020] Step 103: Construct a demand prediction model that integrates deep reinforcement learning and Transformer model to predict customer interest and output customer intention score; In this embodiment, customer feature vectors obtained after dynamic construction of customer profiles are used as input data, and the temporal behavior data in the customer feature vectors are input into the Transformer model. Through the multi-head attention mechanism of the Transformer model, the correlation between customer behavior in different time dimensions is captured, deep temporal features of the customer behavior sequence are extracted, and a comprehensive customer feature vector is output. The comprehensive customer feature vector is used as the state input of the DRL model. The action space of the DRL model is defined as a recommendation combination of different SaaS products and marketing content. Each action corresponds to a recommendation scheme. Positive customer interactions with recommended content are set as positive rewards, negative interactions as negative rewards, and no interaction as zero rewards. The DDPG algorithm is used for iterative training. During training, the DRL model outputs corresponding recommended actions based on the current input customer comprehensive feature vector. Combined with the customer interaction results fed back by the reward function, the action selection strategy is optimized. The comprehensive feature vector of the customer to be predicted is input into the trained demand prediction model, which outputs the customer's interest in different SaaS products and marketing content, forming a customer interest score set. Multi-dimensional indicators for customer intention scoring are determined, including customer interest score, enterprise size, industry matching degree, interaction frequency, and historical transaction similarity. Based on the multi-dimensional indicators and weights, each indicator of the customer to be predicted is quantitatively scored, and a weighted summation method is used to obtain the customer's final intention score.

[0021] In this embodiment, the comprehensive feature vector of the customer to be predicted is input into the trained demand prediction model, and the Transformer model outputs the comprehensive feature vector of the customer. The comprehensive feature vector of the customer is used as the current state of the DRL model to initialize the state of the prediction environment. The Actor network outputs the action probability distribution for different SaaS products and marketing content based on the current customer state and the trained policy network parameters, with each action corresponding to a specific recommendation strategy. The action probability distribution output by the Actor network is mapped to the value rating system of the corresponding SaaS products and marketing content, converting the action probabilities into specific interest ratings for each product and content, forming a preliminary interest rating set. The preliminary interest rating set is deduplicated and categorized and aggregated according to SaaS product category and marketing content type to obtain a customer interest rating set. Combined with the intention tag information in the customer profile, the aggregated customer interest rating set is relevance-calibrated to generate the final customer interest rating set.

[0022] Step 104: Based on the results output by the demand forecasting model, use the collaborative filtering algorithm to generate personalized marketing content and determine the optimal push channel and push time. In this embodiment, based on the output of the demand prediction model, the types of SaaS products and marketing content categories that customers are interested in are extracted. Simultaneously, by combining static and dynamic features from the customer profile, feature dimensions highly relevant to customer needs are selected. A user and item rating matrix is ​​constructed, where rows represent customers, columns represent recommendable SaaS products and marketing content types, and matrix elements are the corresponding interest scores output by the demand prediction model. Based on the user and item rating matrix, the Pearson correlation coefficient between customers is calculated to identify the nearest customer group most similar to the customer's interest characteristics to be predicted. The content similarity between various marketing contents is also calculated to mine the relationships between content, forming a customer similarity set and a content similarity set. Combining the customer similarity set and the content similarity set, a personalized marketing content candidate set containing different types and focuses is generated. The personalized marketing content candidate set is then matched with the optimal push channel and timing scheme to adjust the marketing content, forming the final personalized marketing content and push plan.

[0023] In this embodiment, the attribute characteristics of the push channels are collected, including channel reach, response speed, and user acceptance. Combined with the intention level in the customer profile, channel selection priorities are set for high-intent customers, nurturing customers, and potential customers respectively. The optimal push time is determined based on customer active time data, and finally, the optimal push channel and time plan are generated.

[0024] Step 105: Execute personalized push notifications and collect customer interaction feedback data, and optimize the recommendation strategy based on the customer interaction feedback data.

[0025] Please see Figure 2 The present invention provides a schematic diagram of the structure of an AI-automated precision marketing recommendation system based on big data. The system includes: The data processing module is used to collect user basic data, behavioral data, enterprise attribute data and marketing interaction data. It performs data storage and distributed computing based on the Hadoop distributed framework, and constructs a standardized data sample set after preprocessing. The real-time update module is used to perform initial customer segmentation using the K-means clustering algorithm, construct a three-level label system, and use dynamic feature fusion and attention mechanisms to update customer profiles in real time. The demand forecasting module is used to build a demand forecasting model that integrates deep reinforcement learning and Transformer models to predict customer interest and output customer intention scores. The recommendation generation module combines the results of the demand forecasting model with the collaborative filtering algorithm to generate personalized marketing content and determine the optimal push channels and push times. The push module is used to execute personalized push notifications and collect customer interaction feedback data, and optimize recommendation strategies based on the customer interaction feedback data.

[0026] Figure 3 This is a schematic diagram of the structure of an AI-automated precision marketing recommendation device based on big data, according to an embodiment of the present invention. This AI-automated precision marketing recommendation device 300 based on big data can vary significantly due to different configurations or performance. It may include one or more central processing units (CPUs) 310 (e.g., one or more processors) and a memory 320, and one or more storage media 330 (e.g., one or more mass storage devices) for storing application programs 333 or data 332. The memory 320 and storage media 330 can be temporary or persistent storage. The program stored in the storage media 330 may include one or more modules (not shown in the diagram), each module including a series of instruction operations on the AI-automated precision marketing recommendation device 300 based on big data. Furthermore, the processor 310 may be configured to communicate with the storage media 330 and execute the series of instruction operations in the storage media 330 on the AI-automated precision marketing recommendation device 300 based on big data to implement the method provided in the above embodiment.

[0027] The AI-automated precision marketing recommendation device 300 based on big data may also include one or more power supplies 340, one or more wired or wireless network interfaces 350, one or more input / output interfaces 360, and / or one or more operating systems 331, such as Windows Server, Mac OS X, Unix, Linux, FreeBSD, etc. Those skilled in the art will understand that... Figure 3 The structure of the AI-automated precision marketing recommendation device based on big data shown does not constitute a limitation on the computer device provided by this invention. It may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0028] The present invention also provides a computer-readable storage medium, which can be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium, wherein the computer-readable storage medium stores instructions that, when the instructions are executed on a computer, cause the computer to perform the various steps of the AI-automated precision marketing recommendation method based on big data provided in the above embodiments.

[0029] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the above-described equipment or apparatus / unit can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0030] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0031] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely preferred examples and are not intended to limit the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended technical solutions and their equivalents.

Claims

1. A big data-based AI-automated precision marketing recommendation method, characterized in that, The method includes the following steps: Collect user basic data, behavioral data, enterprise attribute data and marketing interaction data, perform data storage and distributed computing based on the Hadoop distributed framework, and construct a standardized data sample set after preprocessing; The K-means clustering algorithm is used to initially segment customers and construct a three-level labeling system. Dynamic feature fusion and attention mechanisms are used to update customer profiles in real time. Construct a demand prediction model that integrates deep reinforcement learning and Transformer models to predict customer interest and output customer intention scores; Based on the results of the demand forecasting model, a collaborative filtering algorithm is used to generate personalized marketing content and determine the optimal push channel and push time. Perform personalized push notifications and collect customer interaction feedback data, then optimize recommendation strategies based on this data.

2. The AI-automated precision marketing recommendation method based on big data as described in claim 1, characterized in that, The process involves using K-means clustering to initially segment customers, constructing a three-level labeling system, and employing dynamic feature fusion and attention mechanisms to update customer profiles in real time. This includes: Using a standardized data sample set as input, the initial parameters for customer segmentation and the range of the number of clusters are determined. The initial parameters for customer segmentation include the particle swarm size, the maximum number of iterations, the acceleration factor, and the inertia weight. The position of each particle is defined as a set of cluster centroids, and each particle corresponds to a potential customer segmentation scheme. The particle swarm optimization algorithm was used for iterative processing to obtain preliminary customer segmentation results, which divided customers into four categories: high-intent customers, nurturing customers, potential customers, and low-intent customers. Based on the initial customer segmentation results, a three-level tagging system consisting of basic tags, behavioral tags, and intention tags was established. Collect real-time and historical customer behavior data, perform standardized alignment processing, and filter out effective features related to customer needs and segmentation characteristics; Weights are assigned to the preprocessed real-time and historical behavioral features, and customer feature vectors are generated through weighted fusion. The customer feature vector is matched with the customer's current tagging system to update the customer's behavioral tags and intention tags. The feature fusion results and tag information are dynamically updated to update the customer profile in real time.

3. The AI-automated precision marketing recommendation method based on big data as described in claim 2, characterized in that, The particle swarm optimization algorithm is used for iterative processing to obtain preliminary customer segmentation results, dividing customers into four categories: high-intent customers, nurturing customers, potential customers, and low-intent customers. The particle swarm optimization algorithm performs initial iterations on the initialized particle swarm to explore the solution space of the customer group; Particles are divided into two categories based on their fitness values: optimized and improved. Optimized particles are adjusted using the K-means algorithm. If an optimized particle gets stuck in a local convergence state, a Levy flight perturbation is applied to a single dimension randomly selected by it. After generating clones through cloning operations, the improved particles undergo Gaussian mutation. If empty clusters occur during the iteration process, an empty cluster allocation strategy is adopted to reassign the samples farthest from the centroid in the non-empty clusters to the empty clusters, update the cluster centroids, and continue until the iteration is completed to obtain the customer's preliminary clustering results.

4. The AI-automated precision marketing recommendation method based on big data as described in claim 1, characterized in that, The demand prediction model, which integrates deep reinforcement learning and Transformer models, predicts customer interest and outputs a customer intention score, including: The customer feature vector obtained after dynamic construction of customer profiles is used as input data, and the time-series behavioral data in the customer feature vector is input into the Transformer model; By leveraging the multi-head attention mechanism of the Transformer model, we can capture the correlation between customer behaviors across different time dimensions, extract deep temporal features of customer behavior sequences, and output a comprehensive customer feature vector. The customer's comprehensive feature vector is used as the state input of the DRL model. The action space of the DRL model is defined as the recommendation combination of different SaaS products and marketing content. Each action corresponds to a recommendation scheme. Positive interaction behavior of customers with the recommended content is set as positive reward, negative interaction behavior is set as negative reward, and no interaction behavior is set as zero reward. The DDPG algorithm is used for iterative training. During the training process, the DRL model outputs the corresponding recommended action based on the current input customer comprehensive feature vector. Combined with the customer interaction results fed back by the reward function, the action selection strategy is optimized. Input the comprehensive feature vector of the customer to be predicted into the trained demand prediction model, and output the customer’s interest in different SaaS products and marketing content to form a set of customer interest scores. Determine the multi-dimensional indicators for customer intention scoring, which include customer interest score, company size, industry matching degree, interaction frequency, and historical transaction similarity. Based on multi-dimensional indicators and weights, each indicator of the customer to be predicted is quantitatively scored, and the final intention score of the customer is obtained by weighted summation.

5. The AI-automated precision marketing recommendation method based on big data as described in claim 4, characterized in that, The process involves inputting the comprehensive feature vector of the customer to be predicted into the trained demand prediction model, and outputting the customer's interest in different SaaS products and marketing content to form a customer interest score set, including: Input the comprehensive feature vector of the customer to be predicted into the trained demand prediction model, and output the comprehensive feature vector of the customer through the Transformer model; The state of the prediction environment is initialized using the customer's comprehensive feature vector as the current state of the DRL model. Based on the current customer status and the trained policy network parameters, the Actor network outputs the probability distribution of actions for different SaaS products and marketing content, with each action corresponding to a specific recommendation strategy. The action probability distribution output by the Actor network is mapped to the value rating system of the corresponding SaaS products and marketing content, and the action probability is converted into the customer's specific interest rating for each product and content, forming a preliminary set of interest ratings. The initial interest score set is deduplicated and then categorized and aggregated according to SaaS product category and marketing content type to obtain the customer interest score set. By combining the intention tag information in the customer profile, the correlation of the aggregated customer interest score set is calibrated to generate the final customer interest score set.

6. The AI-automated precision marketing recommendation method based on big data as described in claim 1, characterized in that, The results from the demand forecasting model are used to generate personalized marketing content using a collaborative filtering algorithm, and the optimal push channels and push times are determined, including: Based on the output of the demand forecasting model, we extract the types of SaaS products and marketing content categories that customers are interested in. At the same time, we combine the static and dynamic features in the customer profile to select feature dimensions that are highly relevant to customer needs. Construct a user and item rating matrix, where the rows of the matrix represent customers, the columns represent the types of SaaS products and marketing content that can be recommended, and the matrix elements are the corresponding interest scores output by the demand prediction model. Based on user and item rating matrices, Pearson correlation coefficients are calculated among customers to identify the nearest customer groups most similar to the interest characteristics of the customers to be predicted. Content similarity between various marketing content is also calculated to mine the relationships between content and form customer similarity sets and content similarity sets. By combining customer similarity sets and content similarity sets, a personalized marketing content candidate set containing different types and focuses is generated; The candidate set of personalized marketing content is matched with the optimal push channels and timing plans, and the marketing content is adjusted to form the final personalized marketing content and push plan.

7. The AI-automated precision marketing recommendation method based on big data as described in claim 1, characterized in that, The process of combining the results from the demand forecasting model with a collaborative filtering algorithm to generate personalized marketing content and determine the optimal push channel and push time also includes: Collect the attribute characteristics of push channels, including channel reach, response speed, and user acceptance. By combining the intent levels in the customer profile, channel selection priorities are set for high-intent customers, nurturing customers, and potential customers. The optimal push time is determined based on customer activity time data, and finally, the optimal push channel and time plan is generated.

8. A big data-based AI-automated precision marketing recommendation system, characterized in that, The system includes: The data processing module is used to collect user basic data, behavioral data, enterprise attribute data and marketing interaction data. It performs data storage and distributed computing based on the Hadoop distributed framework, and constructs a standardized data sample set after preprocessing. The real-time update module is used to perform initial customer segmentation using the K-means clustering algorithm, construct a three-level label system, and use dynamic feature fusion and attention mechanisms to update customer profiles in real time. The demand forecasting module is used to build a demand forecasting model that integrates deep reinforcement learning and Transformer models to predict customer interest and output customer intention scores. The recommendation generation module combines the results of the demand forecasting model with the collaborative filtering algorithm to generate personalized marketing content and determine the optimal push channels and push times. The push module is used to execute personalized push notifications and collect customer interaction feedback data, and optimize recommendation strategies based on the customer interaction feedback data.

9. An AI-automated precision marketing recommendation device based on big data, characterized in that, The AI-automated precision marketing recommendation device based on big data includes a memory and at least one processor, wherein the memory stores instructions; the at least one processor invokes the instructions in the memory to cause the AI-automated precision marketing recommendation device based on big data to perform each step of the AI-automated precision marketing recommendation method based on big data as described in any one of claims 1-7.

10. A computer-readable storage medium storing instructions thereon, characterized in that, When the instructions are executed by the processor, they implement each step of the AI-automated precision marketing recommendation method based on big data as described in any one of claims 1-7.

Images