Target customer pushing method and system and medium

Abstract

The invention relates to a target customer pushing method and system and a medium, and relates to the technical field of big data, and the method comprises the steps: obtaining behavior information of a user, and constructing a user portrait according to the behavior information; clustering the user portraits according to a preset grouping rule to obtain at least one label group; and obtaining selection input of an operator based on the label group, and screening out user portraits in the label group based on the selection input to obtain target portrait information of the target customer. The method can be used for intelligently screening and pushing target customer information.

Description

Technical Field

[0001] The present invention relates to the field of big data technology, and in particular to a target customer push method, system and medium. Background Art

[0002] Hotel management system is a comprehensive software platform used to handle and optimize various operational and management tasks of the hotel. It covers functions such as room reservation, front desk reception, room management, account settlement, customer relationship management (CRM), inventory management and data analysis. Through automated and integrated operations, it improves the hotel's operational efficiency and service quality, and provides customers with a smoother and more satisfactory accommodation experience.

[0003] Existing hotel management systems are usually limited to basic customer information records and reservation management, lack in-depth analysis and application of customer data, and cannot fully utilize customers' historical behaviors, preferences, and feedback information. This system is difficult to achieve accurate insight and prediction of customer needs, resulting in significant limitations in providing personalized services, making it difficult to improve customer satisfaction and loyalty, and ultimately affecting the hotel's competitiveness and service level. Summary of the invention

[0004] In view of the above-mentioned shortcomings of the prior art, an object of the present invention is to provide a target customer push method, system and medium for intelligently screening and pushing target customer information.

[0005] To achieve the above purpose, the present invention adopts the following technical solution.

[0006] In the first aspect, the present invention provides a method for pushing messages to target customers, which adopts the following technical solution: Obtaining user behavior information and building a user profile based on the behavior information; Clustering the user portraits according to a preset grouping rule to obtain at least one label group; The selection input of the operator based on the label grouping is obtained, and the user portraits in the label grouping are filtered out based on the selection input to obtain the target portrait information of the target customer.

[0007] Furthermore, in the above target customer push method, the step of obtaining user behavior information and constructing a user profile based on the behavior information includes: Collecting user behavior information and cleaning and preprocessing the user behavior information to obtain a data warehouse; Extracting features from the data warehouse, the features including user basic information features, predetermined behavior features, consumption behavior features, feedback features, and preference features; The extracted features are used to construct a user profile, which includes basic information, behavioral characteristics, preference information, and value assessment.

[0008] Furthermore, in the above target customer push method, clustering the user portraits according to a preset grouping rule to obtain at least one label grouping includes: Creating a portrait data frame based on the user portrait, wherein the portrait data frame includes at least basic information parameters, behavior characteristic parameters, preference information parameters, and value assessment parameters; Using a preset coding model to encode each parameter in the image data frame into a high-dimensional vector; Performing weighted averaging on the high-dimensional vectors using a preset attention mechanism model, learning the attention weights of the parameters corresponding to the high-dimensional vectors, and fusing the high-dimensional vectors based on the attention weights to obtain a fused vector; The fusion vector is classified using a preset classifier to obtain the label grouping.

[0009] Furthermore, in the above-mentioned method for pushing messages to target customers, the method of using a preset coding model to encode each parameter in the portrait data frame into a high-dimensional vector includes: Processing the parameter x into the input format x′ of the coding model; The x′ is passed into the encoding model and forwarded to output a high-dimensional vector l i =f(x′1,x′2,…x′ n ), where i represents the length of the coding model.

[0010] Furthermore, in the above-mentioned target customer push method, the high-dimensional vectors are weighted averaged using a preset attention mechanism model, the attention weights of the parameters corresponding to the high-dimensional vectors are learned, and the high-dimensional vectors are fused based on the attention weights to obtain a fused vector, including: concatenating the high-dimensional vectors together to form a combined vector containing all inputs; Inputting the combined vector into a preset attention mechanism model; The attention mechanism model generates the attention weight by calculating the correlation between each vector in the combined vector and other vectors; Based on the attention weights, corresponding vectors in the combined vector are weighted summed to obtain a fusion vector representing the overall features of the user portrait.

[0011] Furthermore, in the above target customer push method, the method of using a preset classifier to classify the fusion vector to obtain the label grouping includes: Mapping the fusion vector to the hidden layer through the fully connected layer, and obtaining the hidden layer vector after being processed by the activation function; Passing the hidden layer vector to the output layer and applying a Softmax function after the output layer to convert the output into a probability distribution; After obtaining the probability of each category through the Softmax function, the category with the largest probability is selected as the final predicted label grouping.

[0012] Furthermore, in the above-mentioned method for pushing target customers, the acquisition operator selects input based on the tag grouping, and based on the selection input, filters out the user portraits in the tag grouping to obtain the target portrait information of the target customer, including: Output each tag group through a visual interface for operators to select; Monitor the order system, and based on the selection input of the operator, push the user portrait grouped by the label corresponding to the selection input to the visualization interface; The user portrait presented on the visualization interface is the target portrait information of the target customer.

[0013] Furthermore, the above target customer push method further includes: Based on the target portrait information and in accordance with preset generation rules, a personalized service list for the target portrait information is generated, and the personalized service list is a service plan for the target user.

[0014] Furthermore, the above target customer push method further includes: The personalized service list is pushed to an executor, and a completion input of the executor when completing the personalized service list is obtained.

[0015] In a second aspect, the present invention provides a target customer push system, which adopts the following technical solution: A customer portrait building module, at least used to obtain user behavior information and build a user portrait based on the behavior information; a customer portrait grouping module, at least used to cluster the user portrait according to a preset grouping rule to obtain at least one label group; The target customer screening module is at least used to cluster the user portraits according to a preset grouping rule to obtain at least one label grouping.

[0016] Furthermore, the target customer push system further includes: The service list creation module is at least used to generate a personalized service list for the target portrait information based on the target portrait information and according to preset generation rules, wherein the personalized service list is a service plan for the target user.

[0017] Furthermore, the target customer push system further includes: The response module is at least used to push the personalized service list to the executor and obtain the completion input of the executor when completing the personalized service list.

[0018] In a third aspect, the present invention provides a readable storage medium, which adopts the following technical solution: The readable storage medium stores computer instructions, and when the computer instructions are executed by the processor, the target customer push method described in any one of the first aspects above is implemented.

[0019] In summary, compared with the prior art, the present invention includes at least one of the following beneficial technical effects: The present invention constructs a user portrait by acquiring user behavior information, clusters the portrait according to preset grouping rules, obtains at least one label group, and then filters out the portrait information of the target customer according to the operator's selection input. Compared with the prior art, this method can more accurately mine and identify target customers, provide more detailed and personalized services, thereby improving customer satisfaction and loyalty, and at the same time improving the effectiveness and pertinence of marketing activities. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0021] Figure 1 It is a flowchart of a specific embodiment of a method for pushing messages to target customers according to the present invention.

[0022] Figure 2 It is a flowchart of another specific embodiment of a method for pushing messages to target customers according to the present invention.

[0023] Figure 3 It is a flowchart of another specific embodiment of a method for pushing messages to target customers according to the present invention.

[0024] Figure 4 It is a flowchart of another specific embodiment of a method for pushing messages to target customers according to the present invention.

[0025] Figure 5 It is a flow chart of a target customer push method of the present invention in an application scenario.

[0026] Figure 6 It is a structural diagram of a specific embodiment of a target customer push system of the present invention. DETAILED DESCRIPTION

[0027] The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present application. In addition, it should be understood that the specific implementation methods described herein are only used to illustrate and explain the present application, and are not used to limit the present application.

[0028] It should be noted that the description order of the following embodiments is not intended to limit the preferred order of the embodiments of the present application. In addition, in the following embodiments, the description of each embodiment has its own emphasis, and for parts not described in detail in one embodiment, reference can be made to the relevant description of other embodiments.

[0029] The method steps described in the embodiments of the present invention may be executed in the order described in the specific implementation manner, or the execution order of each step may be adjusted according to actual needs and on the premise that the technical problem can be solved, which will not be listed here one by one.

[0030] The following is combined with Figure 1-6 The present invention is described in further detail.

[0031] Reference Figure 1 , a target customer push method includes the following steps.

[0032] S1, obtain user behavior information and build a user profile based on the behavior information; user behavior information may include user browsing history, purchase history, click behavior, dwell time, etc. This information is collected through data collection and analysis tools (such as log analysis, data mining technology, etc.). Based on user behavior information, use machine learning or deep learning models (such as BERT, LSTM, etc.) to build a user profile and extract user interests, preferences, purchasing power and other characteristics.

[0033] S2, clustering the user portraits according to preset grouping rules to obtain at least one label group; clustering the user portraits using preset grouping rules (such as based on interests, purchasing behavior, demographic characteristics, etc.). The user portraits are grouped using a clustering algorithm (such as K-means, DBSCAN, hierarchical clustering, spectral clustering, etc.) to obtain multiple label groups. These groups can be dynamic and updated based on real-time data.

[0034] S3, obtain the selection input of the operator based on the label grouping, and filter out the user portraits in the label grouping based on the selection input to obtain the target portrait information of the target customer. The operator can view and select the user portraits in the label grouping through the user interface, and select a specific user group according to specific needs (such as marketing activities, personalized recommendations, etc.). Based on the selection input of the operator, the user portraits in the label grouping can be further filtered to finally obtain accurate target customer portrait information. This step can be combined with business rules, priority sorting, and weight allocation methods to optimize the screening results.

[0035] The method for pushing target customers based on steps S1 to S3 constructs a detailed user portrait by acquiring user behavior information, clusters according to preset grouping rules, and accurately filters the target customer portrait information in combination with the operator's selection input, thereby achieving accurate identification and push of target customers. Compared with the prior art, this method can more effectively analyze and apply user data, thereby providing highly personalized services, improving customer satisfaction and loyalty, and improving the pertinence and success rate of marketing activities.

[0036] Further, see Figure 2 As an implementation mode of the present invention, step S1, obtaining user behavior information and constructing a user portrait according to the behavior information, includes: S11, collecting user behavior information and cleaning and preprocessing the user behavior information to obtain a data warehouse; S12, extracting features from the data warehouse, wherein the features include user basic information features, predetermined behavior features, consumption behavior features, feedback features, and preference features; S13, constructing a user profile using the extracted features, wherein the user profile includes basic information, behavioral characteristics, preference information, and value assessment.

[0037] Specifically, first, in step S11, various types of user behavior data on the interactive platform are collected, such as browsing history, purchase history, click-through rate, search query, etc. Then, data cleaning is performed to remove incomplete, erroneous or irrelevant data, and the data is standardized, such as format unification, missing value processing, etc., so as to build a neat data warehouse for subsequent analysis. Then, key features are extracted from the data warehouse, which can be divided into several categories: user basic information features (such as age, gender, geographic location, etc.), reservation behavior features (such as reservation frequency, frequently reserved product types), consumption behavior features (such as average consumption amount, purchase preferences), feedback features (such as user ratings, comment content analysis) and preference features (such as page dwell time, preferred browsing time period). These features help to understand the user's behavior and preferences more deeply. Finally, the extracted features are used to build a detailed user portrait. This includes integrating and analyzing the above features to form a multi-dimensional user description, including basic information, behavior characteristics, preference information and value assessment. This user portrait can fully reflect the user's characteristics and potential needs, and provide a basis for subsequent personalized services and marketing decisions.

[0038] Based on this, useful information can be effectively extracted from the original user behavior data, and a user portrait that can represent the user's comprehensive attributes and behavioral tendencies can be constructed, which is of great value for achieving more accurate user segmentation and target market push.

[0039] Further, see Figure 3 As an implementation mode of the present invention, step S2, clustering the user portraits according to a preset grouping rule to obtain at least one label group, includes: S21, creating a portrait data frame based on the user portrait, wherein the portrait data frame includes at least basic information parameters, behavior characteristic parameters, preference information parameters, and value assessment parameters; S22, using a preset coding model to encode each parameter in the image data frame into a high-dimensional vector; S23, performing weighted averaging on each of the high-dimensional vectors using a preset attention mechanism model, learning the attention weights of the parameters corresponding to each of the high-dimensional vectors, and fusing each of the high-dimensional vectors based on the attention weights to obtain a fused vector; S24, using a preset classifier to classify the fusion vector to obtain the label grouping.

[0040] Specifically, first, various information about the user, including basic information, behavioral characteristics, preference information, and value assessment parameters, is integrated into a portrait data frame. This data frame serves as the basis for processing and analysis, ensuring the integrity and operability of the data. Next, a preset encoding model, such as the BERT model in deep learning, is used to convert the text and categorical data in the data frame into numerical high-dimensional vectors. These vectors can better represent the complex patterns and implicit relationships in the data, providing a suitable input format for subsequent machine learning processing. Then, a preset attention mechanism model, such as the multi-head attention mechanism in Transformer, is used to process the high-dimensional vector obtained in the previous step. This mechanism calculates the attention weights of different parameters, and weighted averages each vector according to the weights to reflect the relative importance of different features. After weighted averaging, the obtained weighted vectors are fused to form a comprehensive fusion vector that comprehensively considers all important user features. Finally, a preset classifier, such as a support vector machine (SVM), random forest, or neural network, is used to classify the fusion vector. The classification result is the user's label grouping, each of which contains a group of users with similar characteristics.

[0041] Through this series of detailed steps, step S2 can not only effectively integrate and analyze user data, but also ensure that each user group receives the service that best suits its characteristics through a sophisticated clustering process, thereby improving the efficiency and effectiveness of marketing.

[0042] Further, as an implementation mode of the present invention, step S22, using a preset coding model to encode each parameter in the portrait data frame into a high-dimensional vector, includes: processing the parameter x into the input format x′ of the coding model; passing the x′ into the coding model and forward passing it, and outputting the high-dimensional vector l i =f(x′1,x′2,…x′ n ), where i represents the length of the coding model.

[0043] Specifically, each parameter in the user portrait data frame (basic information, behavioral characteristics, preference information, and value assessment) is processed into the input format of the encoding model, including word segmentation of text data, standardization of digital data, and processing of date and category data into acceptable numerical input formats. This step is to ensure that the data can be correctly processed by the encoding model, usually including text vectorization processing such as TF-IDF, Word2Vec, or embedding using pre-trained models such as BERT.

[0044] Organize the processed data (denoted as x′) into a batch input format acceptable to the model, integrate all processed parameter values ​​into a unified data structure (such as an array or tensor), and ensure that they conform to the input dimension and data type requirements of the preset model.

[0045] The preprocessed input x′ is passed into the encoding model for forward propagation. In this step, the model converts the input data into an output high-dimensional vector l according to its trained parameters. i , l i Represents the feature vector obtained from the model, where i represents the dimension of the vector, which is usually a preset value, such as the number of hidden layer nodes or the dimension of the output feature determined when the model is designed.

[0046] The output high-dimensional vector l i Represents the encoding after model conversion, which captures the key information of the input data and expresses it in a high-dimensional space, so that these feature vectors can be used for subsequent machine learning tasks such as clustering and classification. Each high-dimensional vector l i =f(x′1,x′2,…x′ n ) are generated by the model based on the input x′, and these vectors will be directly used in the next step of clustering or classification analysis.

[0047] For example, in order to convert a specific parameter (such as "behavior feature") into a high-dimensional feature vector, a pre-trained language model such as BERT (Bidirectional Encoder Representations from Transformers) can be used. The BERT model is particularly suitable for processing natural language text and can extract complex text features. The following are detailed steps and code implementations to explain how to use Python and PyTorch libraries to convert the "behavior feature" parameter into a high-dimensional vector through the BERT model. The specific steps include: 1. Environment preparation: Install necessary libraries and ensure that transformers and torch are available; 2. Data preparation: Assume that "behavioral features" include the user's browsing sequence on the website, such as "viewed product A, added product B to cart, completed transaction"; 3. Text processing: Use BERT’s tokenizer to convert the “behavior feature” text into a format that the model can process, including tokenizing and converting it to token IDs; 4. Model forward pass: Input the processed data into the BERT model to obtain the output high-dimensional vector.

[0048] Through the above method, the user portrait (such as behavioral characteristics) can be converted into a high-dimensional vector containing rich information. This vector can be used for various machine learning tasks such as clustering and classification to achieve more refined user analysis and service optimization.

[0049] Furthermore, as an implementation mode of the present invention, step S23, weighted averaging is performed on each of the high-dimensional vectors using a preset attention mechanism model, the attention weights of the corresponding parameters of each high-dimensional vector are learned, and the high-dimensional vectors are fused based on the attention weights to obtain a fused vector, including: splicing the high-dimensional vectors together to form a combined vector containing all inputs; inputting the combined vector into a preset attention mechanism model; the attention mechanism model generates the attention weight by calculating the correlation between each vector in the combined vector and other vectors; and weighted summing the corresponding vectors in the combined vector based on the attention weights to obtain a fused vector representing the overall characteristics of the user portrait.

[0050] Exemplarily, the code for processing the features of four different dimensions of the user portrait (basic information, behavioral characteristics, preference information, and value assessment) in step S23 includes: 1. Import dependencies and define the attention mechanism class: 2. Forward pass 3. Model usage: feature_dim=128 attention_model = AttentionMechanism(feature_dim) / / Create an AttentionMechanism instance with feature dimension set to 128 #Create four 128-dimensional random vectors, representing different user features basic_info_vector=torch.rand(feature_dim) behavior_vector=torch.rand(feature_dim) preference_vector=torch.rand(feature_dim) value_vector = torch.rand(feature_dim) #Use torch.stack to stack the four feature vectors into a combined vector for processing together combined_vector=torch.stack([basic_info_vector,behavior_vector,preference_vector,value_vector]) #Process this combined vector through the attention model and output the final fusion vector, which integrates the information of each feature vector, and the weight is determined by the attention weight calculated by the model. fusion_vector=attention_model(combined_vector) The above code shows how to use the attention mechanism to dynamically weight and fuse data from different sources to capture the most important information in the data.

[0051] Further, as an implementation of the present invention, step S24, using a preset classifier to classify the fusion vector to obtain the label grouping, includes: Mapping the fusion vector to the hidden layer through the fully connected layer, and obtaining the hidden layer vector after being processed by the activation function; Passing the hidden layer vector to the output layer and applying a Softmax function after the output layer to convert the output into a probability distribution; After obtaining the probability of each category through the Softmax function, the category with the largest probability is selected as the final predicted label grouping.

[0052] Specifically, a preset classifier is used to classify the fused vector, including how to use the fully connected layer, activation function and Softmax function for classification, so as to obtain label grouping. This process can be implemented by using the deep learning framework PyTorch. Exemplarily, the specific code implementation of step S24 includes: 1. Define a simple neural network model, which includes a fully connected layer, an activation function, and an output layer: 2. Create a model instance, assuming that the input dimension, hidden layer dimension and number of categories are known. At the same time, simulate a fusion vector as input: #Parameter settings input_dim=128 #Assume that the dimension of the fused vector is 128 hidden_dim = 64 # Hidden layer dimension num_classes = 3 # Assume there are 3 categories #Create a model instance model=Classifier(input_dim,hidden_dim,num_classes) #Simulate a fusion vector fusion_vector = torch.rand(input_dim) # Randomly generate fusion vector 3. Process the fused vector through the model and output the category probability. Select the category with the highest probability as the final label group: #Use the model for classification output_probabilities=model(fusion_vector) #Get the predicted category predicted_label=torch.argmax(output_probabilities) print("Predicted Label:",predicted_label.item()) print("Class Probabilities:",output_probabilities) The above code shows how to apply the deep learning model to the classification problem based on the fusion vector, using the fully connected layer and the Softmax function to make effective category predictions.

[0053] Further, see Figure 4 As an implementation mode of the present invention, step S3, obtaining the selection input of the operator based on the tag grouping, and filtering out the user portraits in the tag grouping based on the selection input to obtain the target portrait information of the target customer, includes: S31, grouping each tag and outputting it through a visual interface for selection by an operator; S32, monitoring the order system, and based on the selection input of the operator, pushing the user portrait of the label group corresponding to the selection input to the visualization interface; S33, the user portrait presented on the visualization interface is the target portrait information of the target customer.

[0054] Specifically, first, the information of each tag grouping, such as the grouping statistics, feature overview, etc., is displayed through a user-friendly visual interface. Operators can evaluate and select different tag groups based on the displayed information, and select the most suitable customer group according to marketing strategies or other business needs; then, the operator's selection behavior on the visual interface is obtained. Once the operator determines a specific tag grouping, the user portrait data corresponding to the grouping is retrieved from the back-end order system or customer database, and the data is pushed to the visual interface in real time. Finally, after the operator makes a selection, the specific target user portraits of the selected tag grouping will be displayed on the visual interface. These target user portraits now represent the target customers, and they include detailed personal information, consumption behavior, preferences, etc., for further analysis and use by operators, such as directly used for target setting of marketing activities or for deeper customer relationship management.

[0055] Through this method, operators can intuitively select and locate specific target customer groups. This interactivity and intuitiveness greatly improves the accuracy and efficiency of marketing activities, while also optimizing customer relationship management and the implementation of personalized services.

[0056] Furthermore, as an implementation mode of the present invention, the target customer push method described in the embodiment of the present invention further includes: Based on the target portrait information and in accordance with preset generation rules, a personalized service list for the target portrait information is generated, and the personalized service list is a service plan for the target user.

[0057] Specifically, based on the target profile information obtained through screening, a personalized service list is customized using preset generation rules. This list specifically implements the service plan for the target user. Generation rules can be formulated based on the user's purchase history, preferences, behavior patterns, and other available data points, or they can be manually selected by the operator. For example, if the target profile information shows that the user prefers high-end products, the service list may include high-quality customer support options and high-end product recommendations. During implementation, the system parses the user profile through algorithms or business logic, automatically proposes the service items and products that best suit the user's characteristics, ensures the relevance and attractiveness of the service list, thereby enhancing user satisfaction and improving customer loyalty. This method can more effectively meet the needs of individual customers, improve the level of personalization of services, and thus optimize customer experience and improve business performance.

[0058] Furthermore, as an implementation mode of the present invention, the target customer push method described in the embodiment of the present invention further includes: The personalized service list is pushed to an executor, and a completion input of the executor when completing the personalized service list is obtained.

[0059] Specifically, the formulated personalized service list is pushed to the person in charge of execution to ensure that each service can be implemented as planned. After receiving the service list, the executor completes the corresponding service tasks according to the instructions on the list, such as marketing activities, customer support or other personalized interactions. After completing the service, the executor needs to enter the completion status in the system, which may include the execution time of the service, customer feedback, and records of any special circumstances. Such completion input helps companies monitor service quality, evaluate service effects, and adjust and optimize based on feedback, thereby improving the efficiency and effectiveness of service execution and ensuring that target customers receive the best personalized experience. This process not only enhances the transparency of service management, but also provides data support for continuous improvement and precision marketing.

[0060] Based on this, refer to Figure 5 , the implementation process of an embodiment of the target user push method of the present invention in an application scenario (such as a hotel scenario) includes: firstly, generating a user portrait tag group, and filtering the target user portrait according to the hotel's needs; the system monitors the order system in real time to check whether the hotel has enabled the customer portrait whitelist function; if enabled, the hotel can subscribe to the customer portrait tag; after the subscription is successful, the hotel generates a personalized service order for the customer according to the subscribed tag; after the service order is generated, it is pushed to the executor; the executor completes the task according to the service order and enters the completion status in the system; if the service order is not completed in time, it enters the timeout processing process. This process ensures the whole process management and tracking from customer portrait construction to personalized service push, and improves the accuracy of the service and customer satisfaction.

[0061] The embodiment of the invention also discloses a target customer push system.

[0062] Reference Figure 6 A target customer push system includes a customer portrait building module 1, a customer portrait grouping module 2, a target customer screening module 3, a service list creation module 4 and a response module 5.

[0063] The customer portrait building module 1 is at least used to obtain the user's behavior information and build a user portrait based on the behavior information. Specifically, the customer portrait building module 1 collects the user's behavior information from various channels (such as website browsing records, purchase history, click behavior, search queries, etc.); then, the behavior information is cleaned and preprocessed, including removing noise data, processing missing values ​​and standardizing data formats; next, the system extracts features from the preprocessed data to extract the user's basic information features, predetermined behavior features, consumption behavior features, feedback features and preference features; finally, using these extracted features, the system builds a detailed user portrait, covering the user's basic information, behavior features, preference information and value assessment, thereby providing a basis for subsequent user analysis and personalized services.

[0064] The customer portrait grouping module 2 is at least used to cluster the user portraits according to the preset grouping rules to obtain at least one label group. Specifically, the customer portrait grouping module 2 creates a portrait data frame containing basic information, behavioral characteristics, preference information and value assessment parameters based on the constructed user portrait; then, the preset encoding model is used to encode each parameter in the portrait data frame into a high-dimensional vector; then, the preset attention mechanism model is used to perform weighted averaging on these high-dimensional vectors to learn the importance of each parameter and generate a fusion vector; finally, the system classifies the fusion vector through a preset classifier and clusters the user portraits into multiple label groups, thereby achieving accurate classification and management of user groups.

[0065] The target customer screening module 3 is at least used to cluster the user portraits according to the preset grouping rules to obtain at least one label group. Specifically, the target customer screening module 3 displays the label groups generated by the customer portrait grouping module 2 through a visual interface for the operator to view and select; then, the system monitors the operator's selection input and identifies the selected label group; then, according to the operator's selection, the corresponding user portrait data is screened from the selected label group, and the data is pushed to the visual interface for the operator to further view and confirm; finally, the selected user portrait data is the target customer's target portrait information, which is used for subsequent personalized services and marketing strategy formulation.

[0066] The service list creation module 4 is at least used to generate a personalized service list for the target portrait information based on the target portrait information and according to the preset generation rules, and the personalized service list is a service plan for the target user. Specifically, the service list creation module 4 uses the preset generation rules to analyze the user's specific needs and preferences based on the screened target customer portrait information; then, the system selects the service items that best suit the target customer from the pre-defined service item library and generates a personalized service list; this list contains specific service plans for the target user, such as recommended products, specific offers, personalized messages, etc.; finally, the system ensures that the generated personalized service list is optimized and verified to ensure its accuracy and relevance, and prepares the list for push to relevant executives to ensure that customers can receive high-quality, personalized services.

[0067] The response module 5 is at least used to push the personalized service list to the executor and obtain the completion input of the executor when completing the personalized service list. Specifically, the response module 5 pushes the generated personalized service list to the person responsible for execution, and ensures that the executor receives the task in time through system notifications, text messages or emails; then, the executor completes the relevant tasks according to the instructions of the service list, such as providing specific services, sending personalized recommendations or discounts, etc.; during the execution process, the system monitors the progress of the task in real time to ensure that the service is carried out as planned; after the task is completed, the executor needs to enter the task completion status in the system, including the actual execution content, customer feedback, etc.; the system records these completion inputs for subsequent performance evaluation and service optimization, thereby closed-loop management of the entire service process and improving service quality and customer satisfaction.

[0068] The target customer push system achieves accurate user identification and personalized service push through the collaborative work of five modules: the customer portrait building module 1 collects and builds detailed user portraits; the customer portrait grouping module 2 clusters and classifies user portraits through high-dimensional vector encoding and attention mechanism models; the target customer screening module 3 uses the visual interface and operator selection to accurately screen target customers from the label grouping; the service list creation module 4 generates a personalized service list based on the target portrait information; the response module 5 ensures the efficient execution and feedback loop of the service list. The system significantly improves the accuracy and personalization of user services, optimizes customer experience, enhances customer satisfaction and loyalty, and at the same time improves the pertinence and success rate of marketing activities, achieving a dual improvement in enterprise service quality and operational efficiency.

[0069] The embodiment of the present invention also discloses a readable storage medium.

[0070] A readable storage medium stores a computer program, which, when executed by a processor, implements the steps of a method for pushing a target client described in any one of the above embodiments. The computer-readable storage medium may include: any entity or device capable of carrying a computer program, a recording medium, a USB flash drive, a mobile hard disk, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), and a software distribution medium, etc. The computer program includes a computer program code. The computer program code may be in source code form, object code form, an executable file, or some intermediate form, etc. The computer-readable storage medium may include: any entity or device capable of carrying a computer program code, a recording medium, a USB flash drive, a mobile hard disk, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), and a software distribution medium, etc.

[0071] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code that includes one or more executable instructions for implementing the steps of a specific logical function or process, and the scope of the preferred embodiments of the present invention includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by those skilled in the art to which the embodiments of the present invention belong.

[0072] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, apparatus or device (such as a computer-based system, a system including a processing module, or other system that can fetch instructions from an instruction execution system, apparatus or device and execute instructions), or used in combination with these instruction execution systems, apparatuses or devices.

[0073] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for pushing messages to target customers, characterized in that: include: Obtaining user behavior information and building a user profile based on the behavior information; Clustering the user portraits according to a preset grouping rule to obtain at least one label group; The selection input of the operator based on the label grouping is obtained, and the user portraits in the label grouping are filtered out based on the selection input to obtain the target portrait information of the target customer.

2. The method for pushing messages to target customers according to claim 1, characterized in that: The obtaining of user behavior information and constructing a user profile based on the behavior information includes: Collecting user behavior information and cleaning and preprocessing the user behavior information to obtain a data warehouse; Extracting features from the data warehouse, the features including user basic information features, predetermined behavior features, consumption behavior features, feedback features, and preference features; The extracted features are used to construct a user profile, which includes basic information, behavioral characteristics, preference information, and value assessment.

3. The method for pushing messages to target customers according to claim 2, characterized in that: The clustering of the user portraits according to a preset grouping rule to obtain at least one label grouping includes: Creating a portrait data frame based on the user portrait, wherein the portrait data frame includes at least basic information parameters, behavior characteristic parameters, preference information parameters, and value assessment parameters; Using a preset coding model to encode each parameter in the image data frame into a high-dimensional vector; Performing weighted averaging on the high-dimensional vectors using a preset attention mechanism model, learning the attention weights of the parameters corresponding to the high-dimensional vectors, and fusing the high-dimensional vectors based on the attention weights to obtain a fused vector; The fusion vector is classified using a preset classifier to obtain the label grouping.

4. The method for pushing messages to target customers according to claim 3, characterized in that: The method of using a preset coding model to encode each parameter in the image data frame into a high-dimensional vector includes: Process the parameter x into the input format x of the encoding model ′ ; The x′ is passed into the encoding model and forwarded to output a high-dimensional vector l i =f(x′1,x′2,…x′ n ), where i represents the length of the coding model.

5. The method for pushing messages to target customers according to claim 3, characterized in that: The high-dimensional vectors are weighted averaged using a preset attention mechanism model, the attention weights of the parameters corresponding to the high-dimensional vectors are learned, and the high-dimensional vectors are fused based on the attention weights to obtain a fused vector, including: concatenating the high-dimensional vectors together to form a combined vector containing all inputs; Inputting the combined vector into a preset attention mechanism model; The attention mechanism model generates the attention weight by calculating the correlation between each vector in the combined vector and other vectors; Based on the attention weights, corresponding vectors in the combined vector are weighted summed to obtain a fusion vector representing the overall features of the user portrait.

6. The method for pushing messages to target customers according to claim 3, characterized in that: The using a preset classifier to classify the fusion vector to obtain the label grouping includes: Mapping the fusion vector to the hidden layer through the fully connected layer, and obtaining the hidden layer vector after being processed by the activation function; Passing the hidden layer vector to the output layer and applying a Softmax function after the output layer to convert the output into a probability distribution; After obtaining the probability of each category through the Softmax function, the category with the largest probability is selected as the final predicted label grouping.

7. The method for pushing messages to target customers according to claim 1, characterized in that: The acquisition operator selects input based on the tag grouping, and based on the selection input, filters out the user portraits in the tag grouping to obtain the target portrait information of the target customer, including: Output each tag group through a visual interface for operators to select; Monitor the order system, and based on the selection input of the operator, push the user portrait grouped by the label corresponding to the selection input to the visualization interface; The user portrait presented on the visualization interface is the target portrait information of the target customer.

8. The method for pushing messages to target customers according to claim 1, characterized in that: The method further comprises: Based on the target portrait information, according to a preset generation rule, a personalized service list for the target portrait information is generated, wherein the personalized service list is a service plan for the target user; The personalized service list is pushed to an executor, and a completion input of the executor when completing the personalized service list is obtained.

9. A target customer push system, characterized in that: The system comprises: A customer profile building module, at least used to obtain user behavior information and build a user profile based on the behavior information; A customer portrait grouping module, at least used to cluster the user portraits according to a preset grouping rule to obtain at least one label group; The target customer screening module is at least used to cluster the user portraits according to a preset grouping rule to obtain at least one label grouping.

10. A readable storage medium, characterized in that: The readable storage medium stores computer instructions, and when the computer instructions are executed by a processor, the target client push method according to any one of claims 1 to 8 is implemented.

Images