Customer full life cycle management system and method based on multi-channel fusion

Abstract

The invention discloses a customer full life cycle management system and method based on multi-channel fusion, and relates to the technical field of customer management, and the system comprises a customer classification module which is used for collecting customer data through multiple channels for preprocessing fusion, and carrying out the classification of customer behaviors based on a clustering algorithm; the digital twinning module is used for creating a customer digital twinning body, defining a customer life cycle state and constructing a life cycle model to bind the digital twinning body; the behavior prediction module is used for performing customer behavior prediction based on the digital twin and performing processing; and the feedback optimization module is used for optimizing the life cycle model by using a reinforcement learning algorithm through the predicted customer life cycle state and customer data. According to the method, the initial clustering center is calculated to perform customer behavior clustering analysis, so that the accuracy and efficiency of customer behavior classification are improved, the accuracy of customer life cycle detection is effectively improved, and customer full life cycle management is facilitated.

Description

Technical Field

[0001] The present invention relates to the technical field of customer management, and in particular to a customer full life cycle management system and method based on multi-channel integration. Background Art

[0002] With the development of Internet technology, Customer Relationship Management (CRM) has gradually become an important tool for enterprises to improve their competitiveness and enhance customer satisfaction. In particular, with the diversified development of e-commerce, social media and offline channels, enterprises can obtain customer behavior data from multiple channels. Through these data, enterprises can more accurately identify customer needs and optimize customer service processes. However, traditional CRM systems mostly use a single-channel data collection method and cannot fully integrate customer behavior trajectories, resulting in limitations in customer behavior modeling and prediction. In addition, as the scale of data expands, enterprises have an increasingly urgent need for how to efficiently process, analyze and utilize customer data. Therefore, there is an urgent need for a multi-channel integrated data processing and customer management system to cope with the complex changes in the customer life cycle. Summary of the invention

[0003] In view of the problems existing in the above-mentioned existing customer life cycle management system and method based on multi-channel integration, the present invention is proposed.

[0004] Therefore, the problem to be solved by the present invention is that it is impossible to fully integrate the customer's behavior trajectory, resulting in limitations in the modeling and prediction of customer behavior, and it is difficult to adapt to the rapid changes in customer needs and diversified behavior patterns.

[0005] To solve the above technical problems, the present invention provides the following technical solutions: a customer life cycle management system based on multi-channel integration, comprising: a customer classification module, which is used to collect customer data through multiple channels for pre-processing and integration, and classify customer behavior based on a clustering algorithm; a digital twin module, which is used to create a customer digital twin, define the customer life cycle status, and build a life cycle model to bind the digital twin; a behavior prediction module, which is used to predict customer behavior based on the digital twin and perform processing; a feedback optimization module, which is used to optimize the life cycle model using a reinforcement learning algorithm through the predicted customer life cycle status and customer data.

[0006] As a preferred solution of the customer life cycle management system based on multi-channel integration described in the present invention, wherein: the pre-processing and integration of customer data collected through multiple channels refers to using a RESTful API interface to collect customer data from customer social networks, wherein the customer data includes customer identity data, browsing data, purchase data, and interaction data, unifying the timestamps of all customer data and performing cleaning and filtering, using feature engineering to extract customer data feature vectors, calculating the covariance between feature vectors based on the customer data feature vectors, and constructing a covariance matrix to calculate the Kalman gain, and using a Kalman filter algorithm to fuse the customer data to generate customer fusion features.

[0007] As a preferred solution of the customer life cycle management system based on multi-channel integration described in the present invention, wherein: the customer behavior classification based on the clustering algorithm refers to collecting customer data of different customers for pre-processing and fusion, calculating the Euclidean distance between the fused customer fusion features, and taking the negative value of the Euclidean distance as the similarity s(i, j) between customers;

[0008] Select the median of the similarities between all customers as the initial preference value p, initialize the availability values ​​of all customers to 0, and calculate the responsibility value r(i, j) of customer i choosing customer j:

[0009] r(i,j)=s(i,j)-max(a(i,k)+s(i,k)), k≠j;

[0010] Where r(i, j) is the responsibility value of customer i to choose customer j, a(i, k) is the availability value of the remaining customers, the initial value is 0, and s(i, k) is the similarity between customer i and the remaining customers;

[0011] Synchronously calculate the responsibility value r(j, j) selected by customer j:

[0012] r(j,j)=p-max(a(i,j)+s(i,j)), i≠j;

[0013] Where a(i, j) is the availability value of customer j;

[0014] Update the availability value of customer j based on the responsibility value r(i, j) selected by customer i and the responsibility value r(j, j) selected by customer j:

[0015]

[0016] where a ， (i, j) is the updated availability value of customer j, r(i ， , j) is customer i ， Select the responsibility value of customer j;

[0017] Substitute the updated availability value of customer j into r(i, j) and r(j, j) to obtain r ， (i, j) and r ， (j, j), and repeat the iterative calculation and update until convergence, and define the central customer selection condition as:

[0018] y=(a ， (i, j)+r ， (i, j));

[0019] Where y is the customer's selected value;

[0020] Calculate all customer selection values ​​and sort them, select the top n customers in the customer selection value sorting, and use the customer fusion features as the initial cluster center. For each remaining customer fusion feature, calculate the Euclidean distance between it and the initial cluster center, and assign it to the nearest initial cluster center to form a cluster;

[0021] Calculate the mean of all customer fusion features in each cluster, update the initial cluster center position, and repeat the iteration until the initial cluster center no longer changes. Then stop clustering and define the clustered customer behavior for the final clustering result.

[0022] As a preferred solution of the customer life cycle management system based on multi-channel integration described in the present invention, the creation of a customer digital twin refers to establishing a customer identifier based on customer identity data, recording customer data and customer fusion features in the customer identifier as a customer digital twin, and continuously collecting customer data, and using an incremental update algorithm to update the customer digital twin.

[0023] As a preferred solution of the customer lifecycle management method based on multi-channel integration described in the present invention, wherein: the definition of customer lifecycle status and the construction of lifecycle model binding digital twins include:

[0024] Based on clustered customer behavior, define customer hidden states L1, L2, L3 and L4, collect historical customer data within a specific time period for continuous clustering analysis, and count the changes in customer hidden states within a specific time period. Calculate the transition ratio between customer hidden states as the transition probability and construct the customer hidden state transition matrix A.

[0025] Based on the customer data feature vector, customer observation behaviors K1, K2, and K3 are defined, and hierarchical clustering analysis is performed based on customer data in a specific time period. The proportion of customer observation behaviors in each hidden state in the feature time period is statistically analyzed as the observation probability to construct the observation matrix B.

[0026] Based on the customer hidden state transfer matrix A and the observation matrix B, the Viterbi algorithm is used to calculate the future customer hidden state according to the current customer hidden state:

[0027] q t+1 (i) = max[q t (j)*A ji ]*B ki ;

[0028] where q t+1 (i) is the probability that the customer’s hidden state is Li at time t+1, q t (j) is the probability that the customer’s hidden state is Lj at time t, A ji is the probability that the customer changes from hidden state Lj to hidden state Li, B ki is the probability of the customer's current observed behavior Kk appearing in the hidden state Li;

[0029] Determine the customer's future hidden state based on the maximum hidden state probability, and bind the customer's future hidden state to the customer's digital twin.

[0030] As a preferred solution of the customer lifecycle management system based on multi-channel integration described in the present invention, wherein: the customer behavior prediction based on the digital twin and processing refers to constructing a convolutional neural network, including an input layer, a convolutional layer, a fully connected layer and an output layer, defining the input format of the input layer as the customer hidden state, and the output of the output layer as the customer behavior;

[0031] Use the training set to train the convolutional neural network, and use the cross entropy loss function and Adam optimizer to iteratively optimize the convolutional neural network parameters;

[0032] The future hidden state of the customer bound to the digital twin model is input into the convolutional neural network to obtain the predicted customer behavior, and the corresponding marketing strategy is executed based on the customer hidden state and customer behavior.

[0033] As a preferred solution of the customer life cycle management system based on multi-channel integration described in the present invention, wherein: the optimization of the life cycle model using the reinforcement learning algorithm through the predicted customer life cycle status and customer data refers to the use of the Q-learning algorithm based on the predicted customer hidden state and customer data to dynamically optimize the customer hidden state matrix and the observed behavior matrix, and continuously collect customer marketing feedback to dynamically adjust the customer marketing strategy.

[0034] Another object of the present invention is to provide a customer lifecycle management method based on multi-channel integration, which includes collecting multi-channel customer data for pre-processing and integration;

[0035] Perform cluster analysis based on customer data and define customer behavior based on the clustering results;

[0036] Build a customer digital twin, and build a customer hidden state transfer matrix and observation matrix based on the customer behavior obtained from cluster analysis to predict the customer hidden state, and finally predict the customer behavior based on the predicted customer hidden state;

[0037] Use reinforcement learning algorithms to optimize customer hidden state transition matrices and observation matrices and adjust customer marketing strategies based on feedback data.

[0038] A computer device includes: a memory and a processor; the memory stores a computer program, and the processor implements the steps of the above-mentioned customer life cycle management system based on multi-channel integration when executing the computer program.

[0039] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the above-mentioned customer life cycle management system based on multi-channel integration.

[0040] The beneficial effects of the present invention are as follows: the present invention collects customer data through multiple channels for processing and integration, and calculates the initial cluster center for customer behavior cluster analysis, thereby improving the accuracy and efficiency of customer behavior classification, constructing a customer hidden state transfer matrix and an observation matrix through the cluster analysis results to form a customer life cycle model, and predicting customer hidden states and customer behaviors for marketing, thereby effectively improving the accuracy of customer life cycle detection and facilitating customer full life cycle management. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0042] Figure 1 This is a flow chart of the customer life cycle management system based on multi-channel integration.

[0043] Figure 2 Schematic diagram of the process for building a customer digital twin.

[0044] Figure 3 This is a structural diagram of the customer life cycle management method based on multi-channel integration. DETAILED DESCRIPTION

[0045] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below in conjunction with the accompanying drawings.

[0046] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0047] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor is it a separate or selective embodiment that is mutually exclusive with other embodiments.

[0048] Example 1

[0049] Reference Figure 1 and Figure 2 , which is the first embodiment of the present invention, and which provides a customer lifecycle management system based on multi-channel integration. The customer lifecycle management system based on multi-channel integration includes:

[0050] S1, customer classification module, used to collect customer data through multiple channels for pre-processing and integration, and classify customer behavior based on clustering algorithm;

[0051] Specifically, collecting customer data through multiple channels for pre-processing and fusion refers to using a RESTful API interface to collect customer data from customer social networks, wherein the customer data includes customer identity data, browsing data, purchase data, and interaction data, unifying the timestamps of all customer data and performing cleaning and filtering, using feature engineering to extract customer data feature vectors, calculating the covariance between feature vectors based on the customer data feature vectors, constructing a covariance matrix to calculate the Kalman gain, and using a Kalman filter algorithm to fuse the customer data to generate customer fusion features.

[0052] Various types of customer data are collected from customer social networks through the RESTful API interface, including identity, browsing, purchase, interaction, etc. This multi-dimensional data collection enables the system to more comprehensively capture the customer's behavior trajectory and form a comprehensive portrait of the customer. Compared with the traditional single-channel data collection method, the multi-channel fusion of the present invention greatly improves the data coverage and accuracy, which helps to more accurately understand customer needs and behavior patterns. Before data fusion, the data is cleaned, filtered, and timestamped to ensure data consistency and accuracy. This process reduces the impact of noise data and improves data quality. The application of feature engineering further mines important information in the data and converts customer behavior data into feature vectors that are easy to model. This feature extraction can better capture customers in different channels. The key behavior patterns of customers lay the foundation for subsequent modeling and analysis. By calculating the covariance between customer feature vectors, the system can identify the correlation between various behavioral features. The construction of the covariance matrix provides data support for the calculation of the Kalman gain, and the optimization of the Kalman gain lies in the reasonable allocation of the fusion contribution of different data sources. This process can dynamically adjust the fusion results of customer characteristics according to the quality and weight of each channel data. The introduction of the Kalman filter algorithm significantly improves the efficiency and accuracy of data fusion. Through this algorithm, the system can dynamically update the customer's behavior data and correct the prediction error in real time. Compared with the traditional data fusion method, the Kalman filter algorithm can not only cope with the asynchrony of data from different channels, but also adaptively adjust the prediction model to generate more real-time customer characteristics.

[0053] Furthermore, customer behavior classification based on clustering algorithm refers to collecting customer data of different customers for pre-processing and fusion, calculating the Euclidean distance between the fused customer fusion features, and taking the negative value of the Euclidean distance as the similarity s(i, j) between customers;

[0054] Select the median of the similarities between all customers as the initial preference value p, initialize the availability values ​​of all customers to 0, and calculate the responsibility value r(i, j) of customer i choosing customer j:

[0055] r(i,j)=s(i,j)-max(a(i,k)+s(i,k)), k≠j;

[0056] Where r(i, j) is the responsibility value of customer i to choose customer j, a(i, k) is the availability value of the remaining customers, the initial value is 0, and s(i, k) is the similarity between customer i and the remaining customers;

[0057] Synchronously calculate the responsibility value r(j, j) selected by customer j:

[0058] r(j,j)=p-max(a(i,j)+s(i,j)), i≠j;

[0059] Where a(i, j) is the availability value of customer j;

[0060] Update the availability value of customer j based on the responsibility value r(i, j) selected by customer i and the responsibility value r(j, j) selected by customer j:

[0061]

[0062] where a ， (i, j) is the updated availability value of customer j, r(i ， , j) is customer i ， Select the responsibility value of customer j;

[0063] Substitute the updated availability value of customer j into r(i, j) and r(j, j) to obtain r ， (i, j) and r ， (j, j), and repeat the iterative calculation and update until convergence, and define the central customer selection condition as:

[0064] y=(a ， (i, j)+r ， (i, j));

[0065] Where y is the customer's selected value;

[0066] Calculate all customer selection values ​​and sort them, select the top n customers in the customer selection value sorting, and use the customer fusion features as the initial cluster center. For each remaining customer fusion feature, calculate the Euclidean distance between it and the initial cluster center, and assign it to the nearest initial cluster center to form a cluster;

[0067] Calculate the mean of all customer fusion features in each cluster, update the initial cluster center position, and repeat the iteration until the initial cluster center no longer changes. Then stop clustering and define the clustered customer behavior for the final clustering result.

[0068] By calculating the Euclidean distance between customers and taking its negative value as customer similarity, the system can accurately measure the behavioral similarity of different customers. Selecting the median of all customer similarities as the initial preference value p can avoid too many or too few cluster centers, making the clustering results more stable and reasonable. The beneficial effect of this step is that it enables the system to dynamically adjust the number of cluster centers according to actual data conditions, avoiding artificially setting an unreasonable number of cluster centers. The responsibility value and availability value are the core of the clustering algorithm. By iteratively calculating these two values, the center selection process between customers can be continuously optimized. The update of the responsibility value reflects the fitness of each customer as a central customer in the clustering process, and the availability value further reflects the degree to which each customer's central fitness is recognized by other customers. The beneficial effect of this step is that it ensures that the final selected cluster center can truly represent customer behavior by dynamically updating the responsibility value and availability value. The typical mode is to calculate the selection value of each customer and sort the selection values ​​of all customers, and select the first n customers with the highest selection values ​​as cluster centers. This step ensures that the cluster centers selected by the system are highly representative and can effectively reflect the behavioral characteristics of the customer group. The reasonable selection of the initial cluster center is crucial to the subsequent clustering quality, which can ensure that each customer can be assigned to the most appropriate group. By calculating the Euclidean distance between each customer and the initial cluster center, the system assigns each customer to the cluster center closest to it to complete the initial clustering. After that, the system will update the cluster center according to the mean of the customer fusion features in each cluster, and iterate until the cluster center no longer changes. The beneficial effect of this step is reflected in that it optimizes the customer grouping through continuous iteration, so that the final clustering result can accurately reflect the customer's behavior pattern, which is helpful for the subsequent marketing strategy formulation and customer behavior prediction.

[0069] S2, digital twin module, used to create customer digital twins, define customer lifecycle states, and build lifecycle model-bound digital twins;

[0070] Specifically, creating a customer digital twin means establishing a customer identifier based on customer identity data, recording customer data and customer fusion features in the customer identifier as the customer digital twin, continuously collecting customer data, and using an incremental update algorithm to update the customer digital twin.

[0071] By generating a unique customer identifier through customer identity data, it can ensure that the system can accurately identify and track the behavior trajectory of each customer. The customer identifier serves as the core index of the customer's digital twin. All customer behavior data and fusion features are recorded in the identifier. Through this identifier, the system can easily classify, identify and provide personalized services to different customers, providing a unified basis for subsequent data fusion and behavior analysis, avoiding data duplication and confusion, and ensuring the consistency and traceability of customer data. The uniqueness of the customer identifier ensures that the digital twin of each customer can exist independently and be synchronized with the actual customer's behavior in real time. The customer's behavior data changes dynamically. Over time, the customer's purchasing habits, preferences and needs will also change. The present invention continuously updates the customer's digital twin through an incremental update algorithm. In the incremental update algorithm, each time When a customer generates a new behavior, the system only needs to update the newly added behavior data without having to reprocess all the data. This algorithm can significantly improve data processing efficiency, ensure that the customer's digital twin can reflect the latest behavior status in real time, greatly improve the system's response speed, and reduce data processing resource consumption; secondly, it ensures that the customer's digital twin is always in the latest state, provides enterprises with real-time customer portraits, and helps enterprises make faster and more accurate marketing decisions. Through the continuously updated customer digital twin, the system can predict the customer's future behavior. Based on the customer's current behavior data and fusion features, combined with machine learning algorithms, the system can predict the customer's possible purchasing behavior or churn risk. For example, by analyzing the customer's browsing frequency, purchase frequency and other characteristics, the system can predict the customer's next purchase time and product type, thereby providing customers with more personalized recommendation services.

[0072] Furthermore, defining the customer lifecycle status and building a lifecycle model-bound digital twin includes:

[0073] Based on clustered customer behaviors, define customer hidden states L1, L2, L3, and L4. For example, customers with high purchase frequency and high browsing frequency are high-loyalty customers, customers with low purchase frequency but high interaction frequency are medium-loyalty customers, customers with low purchase frequency and interaction frequency but high browsing frequency are potential customers, and customers with low purchase frequency, browsing frequency, and interaction frequency are lost customers. Collect historical customer data within a specific time period for continuous clustering analysis, and count the changes in customer hidden states within a specific time period. Calculate the transition ratio between customer hidden states as the transition probability, and construct the customer hidden state transition matrix A.

[0074]

[0075] Where P(Li-Lj) is the transition probability from hidden state Li to hidden state Lj;

[0076] Based on the customer data feature vector, customer observation behaviors K1, K2, and K3 are defined, such as high-frequency purchases, high-frequency interactions, and low-frequency browsing. Hierarchical clustering analysis is performed based on customer data in a specific time period. The proportion of customer observation behaviors in each hidden state in the feature time period is counted as the observation probability to construct the observation matrix B.

[0077]

[0078] Where P(Ki|Lj) is the probability of the customer observing behavior Ki in the hidden state Lj;

[0079] Based on the customer hidden state transfer matrix A and the observation matrix B, the Viterbi algorithm is used to calculate the future customer hidden state according to the current customer hidden state:

[0080] q t+1 (i) = max[q t (j)*A ji ]*B ki ;

[0081] where q t+1 (i) is the probability that the customer’s hidden state is Li at time t+1, q t (j) is the probability that the customer’s hidden state is Lj at time t, A ji is the probability that the customer changes from hidden state Lj to hidden state Li, B ki is the probability of the customer's current observed behavior Kk appearing in the hidden state Li;

[0082] Determine the customer's future hidden state based on the maximum hidden state probability, and bind the customer's future hidden state to the customer's digital twin.

[0083] By clustering the customer's behavior data, the system divides the customer into four hidden states: high-loyalty customers, medium-loyalty customers, potential customers, and lost customers. This classification method helps companies to deeply understand the different life cycle stages of customers and carry out targeted customer maintenance and retention. For example, high-loyalty customers may need to maintain the existing service level or provide more personalized services, while potential customers need more incentives to facilitate purchase decisions. The system can automatically identify customer behavior patterns through cluster analysis and map them to specific states of the life cycle. Compared with traditional static life cycle management methods, clustering-based hidden state division provides a more dynamic and flexible customer life cycle management model, which helps companies respond to customer behavior changes in a timely manner. By constructing a transition matrix, companies can clearly understand the dynamic changes of customers between various stages of the life cycle. This predictive ability can help companies formulate response strategies in advance, such as providing special marketing activities or personalized services for customers with higher churn risks, thereby reducing customer churn rates. The construction of the observation matrix helps companies to achieve targeted customer churn in a limited manner. The actual status of customers can be inferred based on their behavioral data. For example, a customer may exhibit frequent purchasing behavior, which indicates that he or she may be in a high-loyalty state. The application of the observation matrix improves the company's ability to perceive the customer's status, enabling the company to respond quickly when certain behaviors are observed and provide targeted services and suggestions. Using the Viterbi algorithm, the system can calculate the customer's future hidden state based on the current customer's hidden state and observed behavior, combined with the transfer matrix and the observation matrix. For example, if a customer is currently in a medium-loyalty state and exhibits high-frequency interactive behavior, the system can predict that the customer may become a high-loyalty customer in the future period of time. On the contrary, if a customer's purchase frequency and interaction frequency both decrease, the system may predict that the customer is about to churn. Not only can the future status of customers be predicted based on historical data, but this prediction can also help companies formulate more accurate marketing and customer maintenance strategies. By binding the customer's future hidden state with the digital twin, the system can adjust the customer's personalized services and marketing activities in real time, thereby maximizing customer experience and reducing customer churn.

[0084] S3, behavior prediction module, used to predict customer behavior based on digital twins and perform processing;

[0085] Specifically, predicting customer behavior based on digital twins and processing means building a convolutional neural network, including an input layer, a convolutional layer, a fully connected layer, and an output layer. The input format of the input layer is defined as the customer's hidden state, and the output of the output layer is defined as the customer's behavior.

[0086] Use the training set to train the convolutional neural network, and use the cross entropy loss function and Adam optimizer to iteratively optimize the convolutional neural network parameters;

[0087] The future hidden state of the customer bound to the digital twin model is input into the convolutional neural network to obtain the predicted customer behavior, and the corresponding marketing strategy is executed according to the customer's hidden state and customer behavior. For example, more personalized offers are pushed to highly active customers, and awakening marketing is carried out for dormant customers.

[0088] By using convolutional neural networks for prediction, the system can better understand customer behavior patterns and provide more accurate personalized marketing strategies. For example, the system can predict customers' future purchasing behavior, browsing frequency or social interaction, helping companies to push relevant product recommendations or personalized offers to customers at the right time, thereby increasing customers' purchase rate and interaction rate. The use of cross-entropy loss function and Adam optimizer can effectively improve the training speed and accuracy of the model, ensuring that CNN can achieve stable performance on complex customer data. This enables the system to generate high-quality customer behavior prediction results in a short period of time and provide companies with real-time decision-making basis. Digital twins can reflect customer changes in real time, enabling CNN models to make predictions based on the latest customer data, improving the timeliness and accuracy of behavior predictions. By predicting customer behavior, companies can formulate personalized marketing strategies in advance to ensure the continuity and consistency of customer experience. The system can respond to customer needs and status changes in advance by predicting behavior, thereby improving customer satisfaction and loyalty. In addition, personalized marketing strategies for different customer life cycle states can effectively reduce customer churn and increase customer long-term value.

[0089] S4, a feedback optimization module, for optimizing the lifecycle model using a reinforcement learning algorithm through predicted customer lifecycle states and customer data;

[0090] Specifically, optimizing the lifecycle model using a reinforcement learning algorithm through predicted customer lifecycle status and customer data means dynamically optimizing the customer hidden state matrix and observed behavior matrix based on predicted customer hidden states and customer data using a Q-learning algorithm, and continuously collecting customer marketing feedback to dynamically adjust customer marketing strategies.

[0091] The Q-learning algorithm enables the system to be adaptive and automatically update model parameters according to changes in customer behavior, thereby improving the accuracy of predictions. Through continuous optimization, the system can promptly identify changes in the customer life cycle (such as the risk of converting from a highly loyal customer to a lost customer) and quickly adjust marketing strategies when customer behavior changes. The beneficial effect of Q-learning is that it can learn the optimal strategy for customer behavior without a clear model, can handle complex customer behavior patterns, and optimize the transition probability of customer life cycle status based on real-time feedback. Through the Q-learning algorithm, the system can more efficiently capture changes in customer life cycle status, thereby Ensuring more accurate and dynamic customer management, continuously collecting customer feedback and dynamically adjusting marketing strategies can improve the flexibility and real-time nature of marketing strategies. Companies can adjust marketing methods in a timely manner according to changes in customer behavior, push more personalized services to high-loyalty customers, and implement retention strategies for potential customers or lost customers. This dynamic adjustment not only improves customer satisfaction, but also enhances the company's market competitiveness. Through the optimization process driven by reinforcement learning, the system can adapt to behavioral changes of different customer groups, accurately adjust marketing strategies and improve the long-term value of customers. In addition, the self-learning ability of the reinforcement learning algorithm reduces the need for manual intervention. The system can automatically extract optimization strategies from the data, reducing the company's operating costs.

[0092] Example 2

[0093] Reference Figure 3 , which is the second embodiment of the present invention, and which is different from the previous embodiment, provides a customer lifecycle management method based on multi-channel integration, which includes:

[0094] Collect multi-channel customer data for pre-processing and integration;

[0095] Perform cluster analysis based on customer data and define customer behavior based on the clustering results;

[0096] Build a customer digital twin, and build a customer hidden state transfer matrix and observation matrix based on the customer behavior obtained from cluster analysis to predict the customer hidden state, and finally predict the customer behavior based on the predicted customer hidden state;

[0097] Use reinforcement learning algorithms to optimize customer hidden state transition matrices and observation matrices and adjust customer marketing strategies based on feedback data.

[0098] If the functions are implemented in the form of software functional 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, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program codes.

[0099] 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, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0100] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0101] 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, a plurality of 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 by hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used to implement: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0102] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A customer life cycle management system based on multi-channel integration, characterized by: include, The customer classification module is used to collect customer data through multiple channels for pre-processing and integration, and classify customer behavior based on clustering algorithms; Digital twin module, used to create customer digital twins, define customer lifecycle states, and build lifecycle model-bound digital twins; Behavior prediction module, used to predict customer behavior based on digital twins and perform processing; Feedback optimization module for optimizing the lifecycle model using reinforcement learning algorithm with predicted customer lifecycle states and customer data.

2. The customer lifecycle management system based on multi-channel integration as claimed in claim 1, characterized in that: The pre-processing and fusion of customer data collected through multiple channels refers to using a RESTful API interface to collect customer data from customer social networks, wherein the customer data includes customer identity data, browsing data, purchase data, and interaction data, unifying the timestamps of all customer data and performing cleaning and filtering, using feature engineering to extract customer data feature vectors, calculating the covariance between feature vectors based on the customer data feature vectors, constructing a covariance matrix to calculate the Kalman gain, and using a Kalman filter algorithm to fuse the customer data to generate customer fusion features.

3. The customer life cycle management system based on multi-channel integration as claimed in claim 2, characterized in that: The customer behavior classification based on the clustering algorithm refers to collecting customer data of different customers for pre-processing and fusion, calculating the Euclidean distance between the fused customer fusion features, and taking the negative value of the Euclidean distance as the similarity s(i, j) between customers; Select the median of the similarities between all customers as the initial preference value p, initialize the availability values ​​of all customers to 0, and calculate the responsibility value r(i, j) of customer i choosing customer j: r(i,j)=s(i,j)-max(a(i,k)+s(i,k)), k≠j; Where r(i, j) is the responsibility value of customer i to choose customer j, a(i, k) is the availability value of the remaining customers, the initial value is 0, and s(i, k) is the similarity between customer i and the remaining customers; Synchronously calculate the responsibility value r(j, j) selected by customer j: r(j,j)=p-max(a(i,j)+s(i,j)), i≠j; Where a(i, j) is the availability value of customer j; Update the availability value of customer j based on the responsibility value r(i, j) selected by customer i and the responsibility value r(j, j) selected by customer j: where a ， (i, j) is the updated availability value of customer j, r(i ， , j) is customer i ， Select the responsibility value of customer j; Substitute the updated availability value of customer j into r(i, j) and r(j, j) to obtain r ， (i, j) and r ， (j, j), and repeat the iterative calculation and update until convergence, and define the central customer selection condition as: y=(a ， (i,j)+r ， (i,j)); Where y is the customer's selected value; Calculate all customer selection values ​​and sort them, select the top n customers in the customer selection value sorting, and use the customer fusion features as the initial cluster center. For each remaining customer fusion feature, calculate the Euclidean distance between it and the initial cluster center, and assign it to the nearest initial cluster center to form a cluster; Calculate the mean of all customer fusion features in each cluster, update the initial cluster center position, and repeat the iteration until the initial cluster center no longer changes. Then stop clustering and define the clustered customer behavior for the final clustering result.

4. The customer lifecycle management system based on multi-channel integration as claimed in claim 3, characterized in that: The creation of a customer digital twin refers to establishing a customer identifier based on customer identity data, recording customer data and customer fusion features in the customer identifier as the customer digital twin, continuously collecting customer data, and updating the customer digital twin using an incremental update algorithm.

5. The customer lifecycle management system based on multi-channel integration as claimed in claim 4, characterized in that: Defining the customer lifecycle status and building the lifecycle model binding digital twin includes: Based on clustered customer behavior, define customer hidden states L1, L2, L3 and L4, collect historical customer data within a specific time period for continuous clustering analysis, and count the changes in customer hidden states within a specific time period. Calculate the transition ratio between customer hidden states as the transition probability and construct the customer hidden state transition matrix A. Based on the customer data feature vector, customer observation behaviors K1, K2, and K3 are defined, and hierarchical clustering analysis is performed based on customer data in a specific time period. The proportion of customer observation behaviors in each hidden state in the feature time period is statistically analyzed as the observation probability to construct the observation matrix B. Based on the customer hidden state transfer matrix A and the observation matrix B, the Viterbi algorithm is used to calculate the future customer hidden state according to the current customer hidden state: q t+1 (i)=max[q t (j)*A ji ]*B ki ; where q t+1 (i) is the probability that the customer’s hidden state is Li at time t+1, q t (j) is the probability that the customer’s hidden state is Lj at time t, A ji is the probability that the customer changes from hidden state Lj to hidden state Li, B ki is the probability of the customer's current observed behavior Kk appearing in the hidden state Li; Determine the customer's future hidden state based on the maximum hidden state probability, and bind the customer's future hidden state to the customer's digital twin.

6. The customer lifecycle management system based on multi-channel integration as claimed in claim 5, characterized in that: The customer behavior prediction based on the digital twin and processing refers to constructing a convolutional neural network, including an input layer, a convolutional layer, a fully connected layer and an output layer, defining the input format of the input layer as the customer hidden state, and the output of the output layer as the customer behavior; Use the training set to train the convolutional neural network, and use the cross entropy loss function and Adam optimizer to iteratively optimize the convolutional neural network parameters; The future hidden state of the customer bound to the digital twin model is input into the convolutional neural network to obtain the predicted customer behavior, and the corresponding marketing strategy is executed based on the customer hidden state and customer behavior.

7. The customer lifecycle management system based on multi-channel integration as claimed in claim 6, characterized in that: The method of optimizing the lifecycle model using a reinforcement learning algorithm through predicted customer lifecycle states and customer data refers to dynamically optimizing the customer hidden state matrix and the observed behavior matrix based on predicted customer hidden states and customer data using a Q-learning algorithm, and continuously collecting customer marketing feedback to dynamically adjust customer marketing strategies.

8. A customer lifecycle management method based on multi-channel integration according to any one of claims 1 to 7, characterized in that: include, Collect multi-channel customer data for pre-processing and integration; Perform cluster analysis based on customer data and define customer behavior based on the clustering results; Build a customer digital twin, and build a customer hidden state transfer matrix and observation matrix based on the customer behavior obtained from cluster analysis to predict the customer hidden state, and finally predict the customer behavior based on the predicted customer hidden state; Use reinforcement learning algorithms to optimize customer hidden state transition matrices and observation matrices and adjust customer marketing strategies based on feedback data.

9. A computer device comprising: Memory and processor; The memory stores a computer program, characterized in that when the processor executes the computer program, the steps of the customer life cycle management system based on multi-channel integration described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the customer lifecycle management system based on multi-channel integration described in any one of claims 1 to 7 are implemented.

Images