An intelligent customer stratification management system and method for operators

Abstract

The present invention discloses an intelligent customer hierarchical management system and method for operators, which relates to the technical field of customer management. It includes a customer relationship analysis module for deploying fog nodes to collect customer data and modeling the customer relationship through a GNN model; a hierarchical processing module for the fog nodes to calculate the node energy consumption based on the collected customer data to determine available nodes, and performing hierarchical processing on customers based on the available nodes and the GNN model; a cloud storage module for the fog nodes to upload the customer hierarchical result to the cloud through energy consumption selection of the optimal transmission path after the customer hierarchical processing is completed, and the cloud stores the customer hierarchical result and displays it to the operator. The present invention effectively improves the efficiency of customer data processing. By using available fog nodes to perform hierarchical processing on customer data and selecting the optimal transmission path to upload the hierarchical processing result to the cloud, it greatly improves the accuracy and security of customer stratification and ensures the transmission efficiency.

Description

Technical Field

[0001] The present invention relates to the technical field of customer management, and particularly to an intelligent customer stratification management system and method for operators. Background Art

[0002] With the continuous development of communication technologies and big data applications, operators' demand for intelligent customer management is increasing day by day. Traditional customer management and stratification technologies usually rely on data analysis in a single dimension, such as customers' consumption behaviors, service usage records, etc. However, these methods fail to fully utilize the interaction relationships and complex behavior patterns of customers in social networks, resulting in relatively low accuracy of customer stratification. In particular, it is unable to effectively identify high-value customer groups. In addition, the traditional centralized data processing mode relies on cloud computing. Although it has strong computing capabilities, there are significant data transmission delays and bandwidth pressures. Especially in large-scale data processing scenarios, the efficiency of this mode drops significantly. In recent years, fog computing has provided a new solution for distributed data processing. By sinking computing and storage resources to edge devices (such as gateways, routers, etc.) close to the data source, it can effectively alleviate data transmission delays and improve the real-time performance of data processing. However, there are still deficiencies in the existing technologies. During the deployment and use of fog nodes, available nodes cannot be selected for data analysis and processing based on energy consumption, and the transmission path selection cannot be fully optimized during data transmission, resulting in easy waste of bandwidth resources and low transmission efficiency during data transmission. Summary of the Invention

[0003] In view of the problems existing in the above-mentioned existing intelligent customer stratification management system and method for operators, the present invention is proposed.

[0004] Therefore, the problems to be solved by the present invention are that during the deployment and use of fog nodes, available nodes cannot be selected for data analysis and processing based on energy consumption, and the transmission path selection cannot be fully optimized during data transmission, resulting in easy waste of bandwidth resources and low transmission efficiency during data transmission.

[0005] To solve the above technical problems, the present invention provides the following technical solutions: An intelligent customer stratification management system for operators, which includes a customer relationship analysis module for deploying fog nodes to collect customer data and modeling customer relationships through a GNN model; a stratification processing module for fog nodes to calculate node energy consumption based on the collected customer data to determine available nodes, and performing customer stratification processing based on the available nodes and the GNN model; a cloud storage module for fog nodes to upload the customer stratification results to the cloud through energy consumption selection of the optimal transmission path after the customer stratification processing is completed, and the cloud stores the customer stratification results and displays them to the operator.

[0006] As a preferred solution of the intelligent customer hierarchical management system for operators described in the present invention, where: the deployment of fog nodes to collect customer data means deploying edge devices to collect customer data, taking each edge device as a fog node, combining all fog nodes to form a fog node network, and preprocessing the collected customer data through the fog nodes.

[0007] As a preferred solution of the intelligent customer hierarchical management system for operators described in the present invention, where: the modeling of customer relationships through the GNN model means that the fog node takes each customer as a separate node, extracts the basic customer data as the initial customer feature vector, inputs the initial customer feature vector into the customer node, forms node edges to connect the customer nodes through the customer social relationship data, and constructs the GNN model with the customer interaction frequency as the node edge weight;

[0008] Update the customer nodes through graph convolution operations of the GNN model. In each convolution layer update, normalize the neighbor nodes of the customer nodes and synchronously update the customer nodes. Use the backpropagation algorithm to optimize the weight matrix of the GNN model each time the customer nodes are updated. Set the cross-entropy loss function as the objective function. When the value of the cross-entropy loss function is lower than the set threshold, stop updating the customer nodes, obtain the customer relationship GNN model, and deploy the customer relationship GNN model to all fog nodes.

[0009] As a preferred solution of the intelligent customer hierarchical management method for operators described in the present invention, where: the fog node calculates the node energy consumption based on the collected customer data to determine available nodes means that the fog node obtains the size of the collected customer data through the edge device, sets the processing density based on the size of the customer data, and calculates the computing requirements of the customer data according to the size of the customer data and the processing density:

[0010]

[0011] Where D i is the computing requirement of the i-th customer data, O i is the size of the customer data, is the processing density;

[0012] The fog node obtains the computing frequency f j of the edge device j through the edge device hardware interface, and obtains the current load L j of the edge device j through the operating system interface;

[0013] The fog node calculates the processing energy consumption E i of the edge device through the customer data computing requirement D j , the edge device computing frequency f j and the current load L ij of the edge device:

[0014]

[0015] where K is the capacitance switching constant;

[0016] Define the energy consumption threshold E of the edge device j. If E ij ≤ E, then the node where the edge device j is located is used as an available node to directly perform hierarchical processing on the customer data. If E ij > E, then an available node is found through the fog node network, and the edge device j transmits the customer data hierarchical processing task to the available node for hierarchical processing.

[0017] As a preferred solution of the intelligent customer hierarchical management system for operators according to the present invention, wherein: the hierarchical processing of customers based on available nodes and the GNN model means collecting historical customer data for preprocessing, constructing an ARIMA model based on the available nodes, training to obtain ARIMA model parameters through the maximum gradient method, and predicting future customer data y based on the collected customer data through the ARIMA model t ;

[0018] Extract the feature vector h in the customer node of the customer relationship GNN model v and splice it with the predicted future customer data y t to form a combined feature vector x v ;

[0019] Use a three-layer neural network to construct a multi-layer perceptron, set the input of the multi-layer perceptron as the combined feature vector, the output as the customer hierarchical result, and use the cross-entropy loss function and gradient descent to train the parameters of the multi-layer perceptron;

[0020] Input the combined feature vector x v into the multi-layer perceptron to obtain the customer hierarchical result.

[0021] As a preferred solution of the intelligent customer hierarchical management system for operators according to the present invention, wherein: after the fog node finishes the hierarchical processing of customers, it uploads the customer hierarchical result to the cloud through energy consumption selection of the optimal transmission path, which means that the fog node obtains all the data transmission path information between the edge device and the cloud, and randomly selects any of them as the initial transmission path group, and calculates the upload and download times of the initial transmission path group based on the transmission path information:

[0022]

[0023] where O zi is the data size of the customer hierarchical result, T uij and T dij are the upload time and download time of the edge device j, Buijk and B dijk are the upload bandwidth and download bandwidth of the k-th transmission path, and P j is the transmission power of the edge device, and G jk is the channel gain of the k-th transmission path, and N jk is the noise of the k-th transmission path, and M is the number of transmission paths in the initial transmission path group;

[0024] Calculate the transmission energy consumption E of the initial transmission path group based on the upload and download times t :

[0025] E t = P j ×(T uij + T dij ) + E ij ;

[0026] Traverse and randomly combine all data transmission paths between the edge device and the cloud, calculate the transmission energy consumption of each combined transmission path group, and traverse all combinations to select the transmission path group with the lowest transmission energy consumption as the final transmission path;

[0027] The edge device transmits the customer stratification result to the cloud through the final transmission path, and monitors the transmission energy consumption in real time during the transmission. If the real-time transmission energy consumption is not higher than the calculated transmission energy consumption, use the current transmission path to transmit all customer stratification results. If the real-time transmission energy consumption is higher than the calculated transmission energy consumption, trigger the transmission path replacement operation and record the remaining data volume O of the customer stratification result si ;

[0028] The edge node obtains the transmission path replacement time T from the q-th transmission path group to the l-th transmission path group through the TCP protocol and the network interface wql , including the disconnection time of the original transmission path, the establishment time of the new transmission path, and the channel verification and handshake time, and calculate the transmission path replacement energy consumption E based on the transmission power of the edge node wql ;

[0029] Based on the transmission path replacement time T wql and the upload and download bandwidths of the l-th transmission path group at the current time t of the l-th transmission path group, use a sliding average window to calculate the upload and download bandwidths of the l-th transmission path group at time t + T wql and predict the channel gain of the l-th transmission path group at time t + T wql based on the LSTM model;

[0030] Calculate the remaining transmission energy consumption E si of the l-th transmission path group and the remaining transmission energy consumption E l of the q-th transmission path group through the remaining data volume O of the customer stratification resultq , and make a comparison:

[0031] If E l +E wql ≤E q , then the edge node will change the transmission path from the q-th transmission path group to the l-th transmission path group for transmitting the remaining data of customer stratification. If E l +E wql >E q , then the edge node still uses the transmission path in the q-th transmission path group to transmit the remaining data of customer stratification.

[0032] As a preferred solution of the intelligent customer stratification management system for operators according to the present invention, wherein: the cloud stores the customer stratification processing result and displays it to the operator, which means that after the cloud receives the customer stratification result transmitted by the edge node, it performs integrity verification, stores the verified customer stratification result in the database, and synchronously forms a visual chart of the customer stratification result to display to the operator.

[0033] Another object of the present invention is to provide an intelligent customer stratification management method for operators, which includes deploying edge devices and fog nodes to collect customer data and perform preprocessing;

[0034] Establish a customer pipeline GNN model according to the customer data and deploy the GNN model to the fog node;

[0035] Judge the available nodes for customer stratification processing by calculating the computing energy consumption of the fog node;

[0036] Select the optimal transmission path and monitor the transmission energy consumption, and transmit the customer stratification processing result to the cloud through the transmission path for storage and display.

[0037] A computer device includes: a memory and a processor; the memory stores a computer program, and when the processor executes the computer program, it implements the steps of the above-mentioned intelligent customer stratification management system for operators.

[0038] A computer-readable storage medium stores a computer program thereon, and when the computer program is executed by a processor, it implements the steps of the above-mentioned intelligent customer stratification management system for operators.

[0039] The beneficial effects of the present invention are as follows: The present invention collects customer data through fog nodes to establish a customer relationship GNN model, and judges the node availability based on the customer data by calculating the node energy consumption, effectively improving the efficiency of customer data processing. The customer data is stratified by available fog nodes, and the optimal transmission path is selected to upload the stratified processing result to the cloud, greatly improving the accuracy and security of customer stratification and ensuring the transmission efficiency. Brief Description of the Drawings

[0040] To more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0041] Figure 1 It is a schematic flowchart of an intelligent customer stratification management system for operators.

[0042] Figure 2 It is a schematic structural diagram for customer stratification processing to calculate energy consumption and judge node availability.

[0043] Figure 3 It is a schematic structural diagram of an intelligent customer stratification management method for operators. Detailed Embodiments

[0044] To make the above objects, features, and advantages of the present invention more obvious and understandable, the following will describe the detailed embodiments of the present invention in conjunction with the drawings of the specification.

[0045] In the following description, many specific details are set forth to fully understand the present invention. However, the present invention can also be implemented in other ways different from those described herein. Those skilled in the art can make similar generalizations without departing from the connotation of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.

[0046] Secondly, the so-called "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that can be included in at least one implementation manner of the present invention. The "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it an independent or selectively exclusive embodiment from other embodiments.

[0047] Embodiment 1

[0048] Referring to Figure 1 and Figure 2 , this is the first embodiment of the present invention. This embodiment provides an intelligent customer stratification management system for operators. The intelligent customer stratification management system for operators includes

[0049] S1. A customer relationship analysis module, which is used to deploy fog nodes to collect customer data and model the customer relationship through a GNN model;

[0050] Specifically, deploying fog nodes to collect customer data means deploying edge devices to collect customer data. Each edge device is regarded as a fog node, and all fog nodes are combined to form a fog node network. The fog nodes preprocess the collected customer data.

[0051] By deploying edge devices as fog nodes, the traditional way of collecting customer data is effectively extended. By deploying multiple fog nodes near the data source, real-time monitoring and collection of data such as customer behavior and consumption habits can be achieved. Compared with the traditional centralized data collection method that relies on the cloud, the deployment of fog nodes greatly reduces data transmission latency and improves the efficiency and real-time performance of data collection. This is significantly helpful for operators to quickly respond to customer needs and provide personalized services. At the same time, the deployment of edge devices has good scalability. Operators can adjust or add fog nodes at any time according to business needs to adapt to the dynamic changes in data volume and processing capacity. Regarding each edge device as an independent fog node can not only utilize the existing computing and storage resources of the device but also reduce the load on the central server through distributed processing. The independent operation of fog nodes enables each device to perform preliminary analysis and processing of data locally, thus reducing the dependence on the central cloud. Each fog node can perform intelligent processing of customer data according to its own resource situation and transmit the preprocessed data to the upper-level node or the cloud for further analysis. This design not only improves the flexibility of the system but also reduces operating costs and helps to cope with complex network environments and data transmission requirements. The formation of the fog node network is a key step in realizing distributed processing and data synchronization. By combining each scattered fog node into a network, multiple nodes can work together to form an overall effect. The fog node network can automatically balance the load of each node, ensure that each node can evenly distribute data processing tasks during high traffic, and avoid the situation of a single node being overloaded. At the same time, the construction of the network structure also enhances the fault tolerance and stability of the system. When a certain fog node fails, other nodes can immediately take over the corresponding work to ensure the continuity and stability of the entire system.

[0052] Furthermore, modeling customer relationships through the GNN model means that fog nodes regard each customer as a separate node, extract basic customer data (age, gender, consumption history, hobbies, etc.) as the initial customer feature vector, input the initial customer feature vector into the customer node, form node edges through customer social relationship data to connect customer nodes, and construct the GNN model by using the customer interaction frequency as the node edge weight;

[0053] Update the customer nodes through graph convolution operations of the GNN model. In each convolution layer update, normalize the neighbor nodes of the customer nodes and synchronously update the customer nodes. Use the backpropagation algorithm to optimize the weight matrix of the GNN model when updating the customer nodes each time. Set the cross-entropy loss function as the objective function. When the value of the cross-entropy loss function is lower than the set threshold, stop the update of the customer nodes, obtain the customer relationship GNN model, and deploy the customer relationship GNN model to all fog nodes.

[0054] Modeling customer relationships through a GNN model is one of the core steps of this invention. The fog nodes regard each customer as an independent node and form an initial feature vector through basic customer data (such as age, gender, consumption history, hobbies, etc.). This method can comprehensively reflect the personalized characteristics of each customer, providing a solid foundation for subsequent customer stratification. Compared with traditional static data models, this invention uses GNN to dynamically update customer features when processing customer social relationships, improving the flexibility and accuracy of the model. In addition, using GNN for modeling can automatically capture complex social relationship patterns among customers, especially the identification of implicit strong social interactions or potential high-value customers. When building a customer relationship graph, the interaction frequency between customers is set as the weight of the node edge, which greatly enhances the model's representation ability of customer social relationships. The interaction frequency not only reflects the interaction intensity between customers but also directly affects the scope and effect of information propagation in graph convolution operations through the weight method. This method ensures that information spreads more smoothly among customers with frequent social interactions, and the model can better capture the behavior patterns of these high-frequency interaction customers, thereby improving the identification accuracy of high-value customers. Graph convolution operation is the core step for updating customer node features. In each layer of convolution operation, the features of customer nodes are updated through the normalization of the features of their neighbor nodes, gradually integrating more levels of customer social relationship information. This mechanism ensures that each customer node not only depends on its own features but also can obtain information from other customer nodes through the social relationship network. The more layers of convolution operations, the more comprehensive the information obtained by the node. Compared with simple weighted averaging, the graph convolution operation in this invention can perform complex multi-layer feature fusion while maintaining the social relationship structure, making customer relationship modeling more accurate. When updating customer nodes each time, this invention uses the backpropagation algorithm to optimize the weight matrix of the GNN model. The backpropagation algorithm gradually adjusts the model's parameters by minimizing the value of the cross-entropy loss function, ensuring that the model can gradually improve the prediction accuracy when processing customer relationships. This weight optimization strategy enables the GNN model to continuously improve its performance according to the actual data of customer relationships and reach the best performance at the end of training. When the value of the cross-entropy loss function is lower than the set threshold, the model will stop updating customer nodes to ensure the convergence and stability of the model. By deploying the trained customer relationship GNN model to all fog nodes, this invention achieves the goal of distributed customer relationship modeling and processing. Each fog node can independently run the customer relationship model and process customer data locally, reducing the dependence on the central server. This distributed deployment not only improves the scalability of the system but also significantly reduces the data transmission delay and enhances the response ability to real-time customer stratification and precision marketing.

[0055] S2. Hierarchical processing module, which is used for the fog node to calculate the node energy consumption based on the collected customer data to judge available nodes, and hierarchically process customers based on the available nodes and the GNN model;

[0056] Specifically, the fog node calculates the node energy consumption based on the collected customer data to judge available nodes, which means that the fog node obtains the size of the collected customer data through the edge device, sets the processing density based on the customer data size, and calculates the computing requirements of the customer data according to the customer data size and the processing density:

[0057] D i =O i ×O Pi ;

[0058] Where D i is the computing requirement of the i-th customer data, O i is the customer data size, and O Pi is the processing density;

[0059] The fog node obtains the computing frequency f of the edge device j through the edge device hardware interface j , and obtains the current load L of the edge device j through the operating system interface j ;

[0060] The fog node calculates the processing energy consumption E of the edge device through the customer data computing requirement D i , the edge device computing frequency f j , and the current load L of the edge device j : ij :

[0061]

[0062] Where K is the capacitance switching constant;

[0063] Define the energy consumption threshold E of the edge device j. If E ij ≤E, then use the node where the edge device j is located as an available node to directly hierarchically process the customer data. If E ij >E, then find an available node through the fog node network, and the edge device j transmits the customer data hierarchical processing task to the available node for hierarchical processing.

[0064] The fog node obtains the size of customer data through edge devices and sets the processing density according to different types of data. The setting of the processing density enables the system to flexibly respond to the computing requirements of different customer data. For example, the computing requirements for processing video data and text data are obviously different, and the processing density of video data will be significantly higher than that of text data. Therefore, by combining the data size and processing density, the system can dynamically calculate the computing requirements of each customer data, ensuring the reasonable allocation of computing resources. This not only improves the efficiency of data processing but also provides an accurate basis for subsequent energy consumption management and task allocation. The computing frequency and current load of edge devices directly affect energy consumption. The higher the frequency and the heavier the load, the higher the energy consumption of the device. Through this energy consumption calculation, the system can monitor the working status of the device in real time, ensuring that the actual usage of the device is considered when calculating energy consumption, thereby optimizing resource allocation. This is particularly important for systems that need to process a large amount of customer data and can effectively reduce unnecessary energy waste. The introduction of an energy consumption threshold provides an efficient task management mechanism for the system. When the fog node calculates the processing energy consumption of the edge device, the system will compare it with the energy consumption threshold of the device. If the energy consumption of the device is lower than the preset energy consumption threshold, the node is considered available, and the system can directly complete the hierarchical processing of customer data on this device. However, if the energy consumption of the device exceeds the threshold, the system will search for other available nodes through the fog node network and transfer the processing task to a device with higher energy efficiency. When the energy consumption of a certain edge device exceeds the set threshold, the system will transmit the task to other available nodes through the fog node network for processing. This design ensures that in the face of a large amount of customer data, the system can automatically switch tasks to avoid overloading a single device. The distributed characteristics of the fog node network make data transmission and processing more flexible, capable of quickly adapting to changes in task load, not only improving the accuracy and stability of data processing but also enhancing the fault tolerance of the system. Even if a certain node fails, the system can quickly transfer the task to other nodes to ensure the continuity of data processing. Through optimized energy consumption management and task allocation strategies, the present invention can provide efficient and stable customer data processing services in a complex network environment.

[0065] Further, hierarchical processing of customers based on available nodes and the GNN model means collecting historical customer data for preprocessing, constructing an ARIMA model based on the available nodes, training to obtain ARIMA model parameters through the maximum gradient method, and predicting future customer data y based on the collected customer data through the ARIMA model. t ;

[0066] Extract the feature vector h in the customer nodes of the customer relationship GNN model. v And predict future customer data y. t Concatenate them to form a joint feature vector x. v ;

[0067] Construct a multi-layer perceptron using a three-layer neural network. Set the input of the multi-layer perceptron as the joint feature vector and the output as the customer stratification result, and use the cross-entropy loss function and gradient descent to train the parameters of the multi-layer perceptron;

[0068] Input the joint feature vector x v into the multi-layer perceptron to obtain the customer stratification result.

[0069] By building ARIMA models on available nodes, the system can predict the future behavior trends of customers based on their past behavior data. This process effectively improves the accuracy and real-time performance of customer data processing. Compared with traditional static data analysis methods, the ARIMA model can capture the dynamic changes in customer behavior. Especially when there are periodic fluctuations in customer behavior or emergencies occur, the ARIMA model can generate more accurate customer behavior prediction results by analyzing the trends in historical data. This technology is particularly crucial in customer stratification because accurate future behavior predictions can help operators adjust strategies in a timely manner and provide personalized services for different customer groups. The feature vectors of each customer node reflect the customer's influence in the social network and their interaction relationships with other customers. Extracting these feature vectors and splicing them with the future data predicted by the ARIMA model can generate a more comprehensive customer description. This feature vector not only contains the customer's historical behavior but also combines the hidden features in their social relationships, which can more comprehensively reflect the customer's potential value. In this way, the system can achieve more accurate customer stratification, helping to identify potential high-value customers or customer groups with special needs. By using the customer's historical behavior characteristics and future prediction data as inputs, the MLP can conduct in-depth analysis of customer behavior patterns. The three-layer MLP structure (input layer, hidden layer, and output layer) can capture the complex relationships in the feature vectors and generate accurate stratification results through multi-layer non-linear calculations. Compared with traditional simple linear classification methods, the MLP can more accurately identify the subtle differences between customers and provide more refined customer stratification results. By using the cross-entropy as the objective function, the multi-layer perceptron can continuously adjust its internal parameters during the training process, gradually improving the prediction accuracy of the model. At the same time, the gradient descent algorithm is used to optimize the weights and bias terms of the MLP, enabling the model to quickly converge to the optimal solution. Gradient descent makes the MLP effectively capture the relationship between customer characteristics and stratification results by gradually reducing the error of the model. By splicing the feature vectors in the customer relationship GNN model with the future customer data predicted by the ARIMA model, the present invention generates a joint feature vector containing historical and future behaviors. This joint feature vector not only integrates the customer's past behavior and future trends but also combines the customer's status and relationships in the social network. This multi-dimensional feature representation can more accurately capture the comprehensive value of customers, helping operators make more informed decisions when stratifying customers.

[0070] S3. Cloud storage module, which is used for the fog node to upload the customer stratification results to the cloud through the optimal transmission path selected according to energy consumption after the customer stratification process is completed. The cloud stores the customer stratification results and displays them to the operator;

[0071] Specifically, after the fog node finishes the hierarchical processing of customers, it uploads the customer hierarchical result to the cloud through energy consumption selection of the optimal transmission path, which means that the fog node obtains all the data transmission path information between the edge device and the cloud, and randomly selects any number of transmission paths as the initial transmission path group, and calculates the upload and download times of the initial transmission path group based on the transmission path information:

[0072]

[0073] Where O zi is the data size of the customer hierarchical result, T uij and T dij are the upload time and download time of the edge device j, B uijk and B dijk are the upload bandwidth and download bandwidth of the kth transmission path, P j is the transmission power of the edge device, G jk is the channel gain of the kth transmission path, N jk is the noise of the kth transmission path, and M is the number of transmission paths in the initial transmission path group;

[0074] Calculate the transmission energy consumption E t of the initial transmission path group based on the upload and download times:

[0075] E t = P j × (T uij + T dij ) + E ij ;

[0076] Traverse and randomly combine all the data transmission paths between the edge device and the cloud, calculate the transmission energy consumption of each combined transmission path group, traverse all combinations and select the transmission path group with the lowest transmission energy consumption as the final transmission path;

[0077] The edge device transmits the customer hierarchical result to the cloud through the final transmission path, and monitors the transmission energy consumption in real time during the transmission. If the real-time transmission energy consumption is not higher than the calculated transmission energy consumption, use the current transmission path to transmit all the customer hierarchical results. If the real-time transmission energy consumption is higher than the calculated transmission energy consumption, trigger the transmission path replacement operation and record the remaining data volume O si of the customer hierarchical result;

[0078] The edge node obtains the transmission path replacement time T wql from the qth transmission path group to the lth transmission path group through the TCP protocol and the network interface, including the disconnection time of the original transmission path, the establishment time of the new transmission path, and the channel verification and handshake time, and calculates the transmission path replacement energy consumption E wql based on the transmission power of the edge node;

[0079] Based on the transmission path replacement time T wql and the upload and download bandwidth usage sliding average window of the l-th transmission path group at the current time t of the l-th transmission path group to calculate the time t + T wql for the upload and download bandwidth of the l-th transmission path group, and predict the time t + T based on the LSTM model wql for the channel gain of the l-th transmission path group;

[0080] Through the remaining data volume O of the customer stratification result si calculate the remaining transmission energy consumption E of the l-th transmission path group l and the remaining transmission energy consumption E of the q-th transmission path group q , and make a comparison:

[0081] If E l + E wql ≤ E q , then the edge node replaces the transmission path from the q-th transmission path group to the l-th transmission path group for the remaining data transmission of customer stratification. If E l + E wql > E q , then the edge node still uses the transmission path in the q-th transmission path group for the remaining data transmission of customer stratification.

[0082] The fog node obtains all the transmission path information between the edge device and the cloud, and randomly selects several transmission paths to form an initial transmission path group. This random selection method provides diverse alternative solutions for subsequent path optimization, avoiding the problem of selecting local optimal paths. By calculating the upload and download times of these initial paths, the system can quickly screen out a more efficient transmission path group, improving the transmission efficiency. Through this step, the system can reduce the data transmission delay and, to a certain extent, improve the utilization rate of the transmission bandwidth. Especially when the customer hierarchical data is large, the optimization of the transmission path can significantly reduce the transmission time and lower the system energy consumption. Based on the upload and download times of the initial transmission path group, the system further calculates the transmission energy consumption of each transmission path group. By traversing and randomly combining all the transmission paths, it finally selects the transmission path group with the lowest energy consumption as the final transmission path. This optimization process ensures that the transmission of the customer hierarchical result not only has a high speed but also can maximize the energy conservation of the system. This energy consumption optimization mechanism can effectively extend the working time of the edge device. Especially for battery-powered edge devices, reducing the transmission energy consumption is crucial for the stability and persistence of the system. By selecting the optimal transmission path group, the system can reduce the energy consumption while achieving efficient transmission, providing a reliable basis for the upload of large-scale customer hierarchical data. During the transmission of customer hierarchical data, the fog node will monitor the energy consumption of the current transmission path in real time. If the actual energy consumption exceeds the calculated transmission energy consumption, the system will trigger a path replacement operation. This real-time monitoring mechanism ensures the dynamic optimization of the transmission path, avoiding problems such as low transmission efficiency or high energy consumption caused by the degradation of the path quality. The introduction of the real-time monitoring and path replacement mechanism enables the system to flexibly respond to changes in the network environment. By dynamically adjusting the transmission path, the system can switch to a better transmission path in a timely manner when there is network congestion or a decline in path performance, thus ensuring the continuity and efficiency of data transmission. Especially in a complex network environment, this path optimization mechanism can greatly improve the reliability of the system. When the system decides to replace the transmission path, it needs to consider the energy consumption during the replacement process, including the energy consumption during disconnecting the original path, establishing a new path, and channel verification and handshake. By calculating the replacement energy consumption of the transmission path based on the transmission power of the edge node, the system can evaluate the cost of the path replacement operation to ensure that the path replacement is overall the most energy-efficient. During the path replacement process, the system uses a sliding average window to calculate the bandwidth of the new transmission path and predicts the channel gain through the LSTM model. This way of combining short-term smoothing processing and long-term trend prediction enables the system to more accurately grasp the state changes of the transmission path. Before replacing the path, the system calculates the remaining transmission energy consumption of the current path group and the target path group based on the remaining data volume of the customer hierarchical result. If the remaining energy consumption of the new path group is lower than that of the current path group, the system will execute the path replacement; otherwise, it will continue to use the current path group.This decision-making mechanism effectively prevents unnecessary path switching and ensures the maximization of transmission efficiency. Through this energy consumption comparison and decision-making mechanism, the system can reduce unnecessary resource consumption while ensuring transmission efficiency, improving the stability and durability of data transmission. Especially during the transmission of large-scale customer data, energy consumption optimization plays a significant role in enhancing the overall performance of the system.

[0083] Furthermore, the cloud stores the customer stratification processing results and presents them to the operator, which means that after receiving the customer stratification results transmitted by the edge node, the cloud conducts integrity verification and stores the verified customer stratification results in the database. Meanwhile, the cloud forms a visual chart of the customer stratification results and presents it to the operator.

[0084] After the customer stratification results are transmitted from the edge node to the cloud, the cloud first conducts integrity verification on the data. Through hash values, checksums, or other common integrity verification technologies, the system can quickly detect whether data has been lost, tampered with, or in error during transmission. This verification process is a key step to ensure data accuracy and security. Especially when dealing with large-scale customer stratification data, any data loss or error may affect the accuracy of subsequent analysis and decision-making. The customer stratification results that pass the integrity verification will be stored in the cloud database. The database is not only used for storage but also provides efficient query and analysis functions to support subsequent data mining and analysis. This storage process ensures the security and durability of customer stratification data. The operator can call historical stratification data for analysis and comparison as needed. While storing the customer stratification results, the cloud synchronously presents the results to the operator in the form of a chart. This synchronous presentation mechanism enables the operator to obtain customer stratification information in real time, facilitating a quick understanding of changes in the customer stratification structure. Through visualization technology, complex data can be transformed into intuitive graphs and charts, helping the operator better understand the meaning behind the data. The presentation of customer stratification results is not only used for real-time monitoring but also serves as an important basis for subsequent business decisions. Through the customer stratification results stored in the cloud database, the operator can conduct long-term trend analysis, customer behavior prediction, and market strategy adjustment. Combining data analysis tools, the operator can further explore the business value behind the data, formulate precise marketing strategies for different customer groups, and improve customer satisfaction and loyalty.

[0085] Embodiment 2

[0086] Referring to Figure 3 , this is the second embodiment of the present invention. This embodiment is different from the previous one and provides an intelligent customer stratification management method for operators, which includes

[0087] Deploying edge devices and fog nodes to collect customer data and perform preprocessing;

[0088] Build a customer pipeline GNN model based on customer data and deploy the GNN model to the fog nodes;

[0089] Judge the available nodes by calculating the energy consumption of the fog nodes and perform customer stratification processing;

[0090] Select the optimal transmission path and monitor the transmission energy consumption, and transmit the customer stratification processing results to the cloud for storage and display through the transmission path.

[0091] If the above functions are implemented in the form of software function units and sold or used as independent products, they 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 a part of this technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical discs that can store program codes.

[0092] The logic and / or steps represented in the flowchart or described in other ways herein, for example, can be considered as a definite sequence list of executable instructions for implementing logical functions, and can be specifically implemented 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 processor, or other systems that can fetch instructions from the instruction execution system, apparatus, or device and execute the instructions), or in combination with these instruction execution systems, apparatuses, or devices. For the purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device.

[0093] More specific examples (non-exhaustive list) of computer-readable media include the following: an electrical connection (electronic device) having one or more wirings, a portable computer diskette (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disc read-only memory (CDROM). Additionally, the computer-readable media can even be paper or other suitable media on which the program can be printed, as the program can be obtained electronically, for example, by optically scanning the paper or other media, followed by editing, interpretation, or other suitable processing as necessary, and then stored in a computer memory.

[0094] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, any one or a combination of the following techniques known in the art can be used: discrete logic circuits having logic gate circuits for implementing logical functions on data signals, application specific integrated circuits having appropriate combinational logic gate circuits, programmable gate arrays (PGAs), field programmable gate arrays (FPGAs), etc.

[0095] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered by the scope of the claims of the present invention.

Claims

1. An intelligent customer stratification management system for operators, characterized in that: including, a customer relationship analysis module, configured to deploy fog nodes to collect customer data and model customer relationships through a GNN model; a hierarchical processing module, configured to have fog nodes calculate node energy consumption based on the collected customer data to determine available nodes, and perform hierarchical processing on customers based on the available nodes and the GNN model; a cloud storage module, configured to have fog nodes, after completing the hierarchical processing of customers, select an optimal transmission path based on energy consumption to upload the customer hierarchical results to the cloud, and the cloud stores the customer hierarchical results and displays them to the operator; Deploying fog nodes to collect customer data means deploying edge devices to collect customer data, taking each edge device as a fog node, combining all fog nodes to form a fog node network, and preprocessing the collected customer data through the fog nodes; The fog nodes calculating node energy consumption based on the collected customer data to determine available nodes means that the fog nodes obtain the size of the collected customer data through edge devices, set the processing density based on the size of the customer data, and calculate the computing requirements of the customer data according to the size of the customer data and the processing density: D i = O i × O Pi ; Among which D i is the computing requirement of the i-th customer data, O i is the customer data size, O Pi is the processing density; The fog node obtains the computing frequency f of edge device j through the edge device hardware interface j and obtains the current load L of edge device j through the operating system interface j ; The fog node calculates the demand D through customer data i , the edge device computing frequency f j and the current load L of the edge device j to calculate the processing energy consumption E of the edge device ij : where K is the capacitance switching constant; Define the energy consumption threshold E of the edge device j. If E ij ≤ E, then take the node where the edge device j is located as an available node and directly perform hierarchical processing on the customer data. If E ij > E, then find an available node through the fog node network, and the edge device j transmits the hierarchical processing task of the customer data to the available node for hierarchical processing.

2. The intelligent customer stratification management system for operators according to claim 1, characterized in that: Modeling customer relationships through the GNN model means that the fog nodes take each customer as a separate node, extract the basic customer data as the initial customer feature vector, input the initial customer feature vector into the customer node, form node edges through customer social relationship data to connect customer nodes, and construct a GNN model with the customer interaction frequency as the node edge weight; Updating customer nodes through graph convolution operations in the GNN model. In each convolutional layer update, normalize the neighbor nodes of the customer node and synchronously update the customer node. Use the backpropagation algorithm to optimize the weight matrix of the GNN model each time the customer node is updated. Set the cross-entropy loss function as the objective function. When the value of the cross-entropy loss function is lower than the set threshold, stop updating the customer node, obtain the customer relationship GNN model, and deploy the customer relationship GNN model to all fog nodes.

3. The intelligent customer stratification management system for operators according to claim 2, characterized in that: The hierarchical processing of customers based on available nodes and GNN models means collecting historical customer data for preprocessing, constructing an ARIMA model based on available nodes, training to obtain ARIMA model parameters through the maximum gradient method, and predicting future customer data y based on the collected customer data through the ARIMA model t ; Extract the feature vector h in the customer node of the customer relationship GNN model v And predict future customer data y t Concatenate them to form a joint feature vector x v ; Construct a multi-layer perceptron using a three-layer neural network, set the input of the multi-layer perceptron as the combined feature vector, the output as the customer hierarchical result, and use the cross-entropy loss function and gradient descent to train the parameters of the multi-layer perceptron; Input the combined feature vector x v into the multi-layer perceptron to obtain the customer stratification result.

4. The intelligent customer stratification management system for operators according to claim 3, characterized in that: After the fog nodes complete the hierarchical processing of customers, selecting an optimal transmission path based on energy consumption to upload the customer hierarchical results to the cloud means that the fog nodes obtain all the data transmission path information between the edge device and the cloud, and randomly select any number of transmission paths as the initial transmission path group, and calculate the upload and download times of the initial transmission path group based on the transmission path information: Among which O zi is the size of the customer stratification result data, T uij and T dij are the upload time and download time of the edge device j, B uijk and B dijk are the upload bandwidth and download bandwidth of the k-th transmission path, P j is the transmission power of the edge device, G jk is the channel gain of the k-th transmission path, N jk is the noise of the k-th transmission path, and M is the number of transmission paths in the initial transmission path group; Calculate the transmission energy consumption E of the initial transmission path group based on the upload and download times t : E t = P j × (T uij + T dij ) + E ij ; Randomly combine and traverse all the data transmission paths between the edge device and the cloud, calculate the transmission energy consumption of each combined transmission path group, and traverse all combinations to select the transmission path group with the lowest transmission energy consumption as the final transmission path; The edge device transmits the customer stratification results to the cloud through the final transmission path and monitors the transmission energy consumption in real time during the transmission. If the real-time transmission energy consumption is not higher than the calculated transmission energy consumption, all customer stratification results are transmitted using the current transmission path. If the real-time transmission energy consumption is higher than the calculated transmission energy consumption, a transmission path replacement operation is triggered and the remaining data volume of the customer stratification results is recorded. si ; The edge node obtains the transmission path replacement time T for changing from the q-th transmission path group to the l-th transmission path group through the TCP protocol and the network interface wql , including the disconnection time of the original transmission path, the establishment time of the new transmission path, and the channel verification and handshake time, and calculates the transmission path replacement energy consumption E based on the transmission power of the edge node wql ; Based on the transmission path replacement time T wql and the upload and download bandwidth of the l-th transmission path group at the current time t are calculated using a sliding average window at time t + T wql for the upload and download bandwidth of the l-th transmission path group, and the channel gain of the l-th transmission path group at time t + T is predicted based on the LSTM model wql ; Remaining data volume O through customer stratification results si Calculate the remaining transmission energy consumption E of the l-th transmission path group l and the remaining transmission energy consumption E of the q-th transmission path group q and make a comparison: If E l +E wql ≤E q , the edge node will change the transmission path from the q-th transmission path group to the l-th transmission path group for transmitting the remaining data of customer stratification. If E l +E wql >E q , the edge node will still use the transmission path in the q-th transmission path group to transmit the remaining data of customer stratification.

5. The intelligent customer stratification management system for operators according to claim 4, characterized in that: The cloud storing and displaying the customer hierarchical processing results to the operator means that after the cloud receives the customer hierarchical results transmitted by the edge node, it performs integrity verification, stores the verified customer hierarchical results in the database, and simultaneously forms a visualization chart of the customer hierarchical results to display to the operator.

6. An intelligent customer stratification management method for operators of the intelligent customer stratification management system for operators as described in any one of claims 1-5, characterized in that: including, Deploy edge devices and fog nodes to collect customer data and perform preprocessing; Establish a customer pipeline GNN model based on customer data and deploy the GNN model to fog nodes; Judge available nodes for customer stratification processing by calculating the computing energy consumption of fog nodes; Select the optimal transmission path and monitor the transmission energy consumption, and transmit the customer stratification processing results to the cloud through the transmission path for storage and display.

7. A computer device, comprising: A memory and a processor; The memory stores a computer program, characterized in that: when the processor executes the computer program, the steps of the intelligent customer stratification management system for operators described in any one of claims 1 to 5 are implemented.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, the steps of the intelligent customer stratification management system for operators described in any one of claims 1 to 5 are implemented.

Images