Remote bank digital human service method and device based on comprehensive value classification and medium

Abstract

The invention discloses a remote bank digital human service method and device based on comprehensive value classification and a medium. According to the method, the comprehensive value of the customer is calculated and the customer feature tag is combined for classification, on the basis of hierarchical classification of the customer, an application mode of'regular model digital human + artificial billing 'is adopted, the digital human is mainly used for processing repeated and mechanical consultation and high-frequency regularized business in remote banking business, and the digital human is mainly used for processing the remote banking business. When an abnormal condition or a sensitive problem is encountered, a manual seat carries out the bottom, meanwhile, the acceptance degree of different customers to artificial intelligence technologies such as digital persons is fully considered, choices are given to the customers, optimal configuration and efficient utilization of bank service resources are promoted through digital person services, marketing clues of potential similar customer groups are mined, and the marketing quality of the customer groups is improved. The accuracy and effectiveness of remote bank services are comprehensively improved, and the innovative development of a bank digital service system is promoted.

Description

Technical Field

[0001] The present invention relates to the technical field of virtual digital human, and in particular to a remote banking digital human service method, device and medium based on comprehensive value classification. Background Art

[0002] Virtual digital human technology integrates cutting-edge technologies such as artificial intelligence, computer graphics, and speech synthesis, and can create highly realistic virtual images that can interact with humans, showing broad application prospects in multiple fields. However, the current virtual digital human application in the banking field still has problems such as low service accuracy and effectiveness, and inconsistent customer acceptance. The current large model cannot completely avoid the "hallucination" problem, which is easy to cause a series of problems such as public opinion and reputation risks for banks.

[0003] Currently, banks usually classify their customers based on feature similarities only, only considering the current customer value. This fails to express the comprehensive value and marketing potential of the customer base, resulting in inaccurate marketing. Summary of the invention

[0004] The purpose of the present invention is to provide a remote banking digital human service method, device and medium based on comprehensive value classification in view of the deficiencies in the prior art.

[0005] According to a first aspect of this specification, a remote banking digital human service method based on comprehensive value classification is provided, the method comprising:

[0006] Calculate the comprehensive value of customers and combine them with customer feature labels to perform customer stratification and classification modeling;

[0007] Remote banking digital human services are provided by combining rule models with manual backup. When a customer initiates a service request, the customer selects a service mode based on the customer stratification and classification principle. The service modes include: virtual digital human agent service, manual video agent service, and branch counter service.

[0008] When a customer chooses a virtual digital agent service, the digital agent will quickly judge and respond to the customer's consultation or business request based on built-in business rules;

[0009] When customers choose human video agent service, or encounter complex or sensitive issues that exceed the preset rules of digital humans, human video agents will provide service;

[0010] When customers choose bank counter services, or when remote banking services cannot solve the problem, the bank counter staff will handle it;

[0011] When the digital human processes the business rules and passes the video quality inspection, or when the service is provided by a manual video agent, dual video recording will be performed for evidence storage; after the service is completed, all service interactions will be recorded.

[0012] Further, the digital human seat service is mainly used to process repetitive and mechanical consultations and high-frequency, regularized, and low-risk services in remote banking, and the following business processes are preferably selected: product consultation, large transfer, password reset, information modification, account upgrade / downgrade, and dormant account activation.

[0013] Further, the customer group stratification and classification modeling includes the following steps:

[0014] (1) Calculate the comprehensive value of the customer, and the calculation formula is as follows:

[0015]

[0016] where tcv i is the comprehensive value of the i-th customer, MI is the monthly income of the customer, t is the number of periods, CR is the probability that the customer continues to stay after each period, f is the consumption frequency, and d is the discount rate;

[0017] (2) Perform dataset modeling on the customer information u ij ∈U, i ∈ [1, n], j ∈ [1, m], n is the total number of customers, m is the total number of labels, u ij is the j-th dimension label of the i-th customer, and U is the customer information set;

[0018] (3) Preprocess the information of each dimension of the customer to obtain the customer information data set The formula is as follows:

[0019]

[0020] where i ∈ [1, n], j ∈ [1, m], u qj is the j-th dimension label of the q-th customer, u ij is the j-th dimension label of the i-th customer, is the j-th dimension information of the i-th customer after preprocessing,

[0021] (4) Randomly select K customer information from the set as the center point set C = {c 1 ,..., c k ,..., c K} for classification, where c k is the k-th center point, k ∈ [1, K];

[0022] (5) Calculate the distance from each preprocessed customer information to each center point in C, and combine the comprehensive value of the customer for distance calculation, and classify the customer into the sample in C with the shortest distance to form the classified data set P = {P1 ,...,P k ,...,P K}, The distance calculation formula is:

[0023]

[0024] where \(i\in[1,n]\), \(j\in[1,m]\), \(k\in[1,K]\), \(tcv\) i is the comprehensive value of the \(i\)-th customer, \(tcv\) k is the comprehensive value of the \(k\)-th customer in \(C\), \(c\) kj is the \(j\)-th dimension label of the \(k\)-th customer in \(C\), \(P\) k is the set of customer group classifications closest to \(c\) k as the center point;

[0025] (6) Calculate the center point of each data set \(P\) k after classification and update it. The center point calculation formula is:

[0026]

[0027] where \(k\in[1,K]\), \(j\in[1,m]\), \(s\) is the total number of samples in the classified data set \(P\) k and \(p\) zj is the \(z\)-th element in \(P\) k , is the center point of the new \(j\)-th dimension label of \(P\) k ;

[0028] (7) Repeat steps (5) and (6) until the center point no longer changes significantly. The error threshold is set to \(\sigma\), that is where \(r\) refers to the \(r\)-th iteration.

[0029] Furthermore, the label system of customers is designed into seven categories: basic information, asset information, revenue contribution, product preference, consumption behavior, channel preference, and life cycle:

[0030] The basic information includes: gender, age, education level, work unit, native place;

[0031] The asset information includes: customer AUM, daily average value and point value of various products;

[0032] The revenue contribution includes: high-value customers, medium-high-value customers, low-value customers;

[0033] The product preference includes: regular holding, preference for life payment, preference for bank wealth management;

[0034] The consumption behavior includes: consumption geographical location, consumption time period, brand preference, advertising and marketing preference, and consumption hot spot preference;

[0035] The said channel preferences: counter, online banking, mobile banking, remote banking;

[0036] The said life cycle includes: acquisition period, promotion period, maturity period, decline period, and churn period.

[0037] Furthermore, when providing digital human services, preference settings and operation management are carried out according to the segmented and classified customer groups, and long-tail customers are regularly consulted to explore customer needs and conduct full life cycle management.

[0038] Furthermore, the digital human service driven by a rule model includes the following steps:

[0039] (1) Select high-frequency, regular, and low-risk business scenarios in the remote banking, establish a knowledge base, store it in the database, and configure corresponding answers;

[0040] (2) During the voice conversation, after the remote banking server transmits the voice signal to the middle platform, the middle platform processes and generates a response through ASR speech-to-text technology, NLP semantic understanding, and question-and-answer matching technology. After the server receives and parses the response, it makes corresponding feedback;

[0041] (3) For transaction call-type answers, the server will execute the corresponding transaction service; for display-type answers, the server will display the answers at the specified position on the interface; for interactive instructions, the server will generate voice from the text through the TTS service, and at the same time send the instruction text to the digital human. The driving engine controls the facial and limb movements, and the rendering engine renders and synthesizes lip movement and action elements, thereby generating video frames and pushing them to the digital human interface for playback.

[0042] Furthermore, digital human marketing is carried out based on the results of customer group segmentation and classification, including the following steps:

[0043] (1) For the classified customer groups P = {P 1 ,..., P k ,..., P K}, for each customer group P k dig out the product purchase set V of this customer group;

[0044] (2) For this customer group P k , for each customer, automated marketing is carried out using digital humans according to the complement of the purchased products and V;

[0045] (3) Repeat steps (1) and (2) to conduct precise marketing for all classified customer groups.

[0046] According to a second aspect of the present specification, there is provided an electronic device, including a memory and a processor, the memory being coupled to the processor; wherein, the memory is used for storing program data, and the processor is used for executing the program data to implement the remote banking digital human service method based on comprehensive value classification as described in the first aspect.

[0047] According to a third aspect of the present specification, there is provided a computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, it implements the remote banking digital human service method based on comprehensive value classification as described in the first aspect.

[0048] According to a fourth aspect of the present specification, there is provided a computer program product, including a computer program / instructions, and when the computer program / instructions are executed by a processor, it implements the remote banking digital human service method based on comprehensive value classification as described in the first aspect.

[0049] The beneficial effects of the present invention are as follows: The present invention classifies by calculating the comprehensive value of customers and combining customer feature tags. On the basis of customer stratification and classification, an application mode of "rule model digital human + manual backup" is adopted. The digital human mainly processes repetitive and mechanical consultations and high-frequency regularized services in remote banking. When encountering abnormal situations or sensitive issues, the manual seat will provide backup. At the same time, the acceptance degree of different customers for artificial intelligence technologies such as digital humans is fully considered, and customers are given the right to choose. Through digital human services, the optimal allocation and efficient utilization of bank service resources are promoted, potential marketing leads of similar customer groups are mined, the accuracy and effectiveness of remote banking services are comprehensively improved, and the innovative development of the bank's digital service system is promoted. Description of the Drawings

[0050] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required to be used in 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 be obtained based on these drawings.

[0051] Figure 1 It is a flowchart of digital human service based on "rule model + manual backup" from the perspective of customer stratification and classification shown in an exemplary embodiment;

[0052] Figure 2 It is a flowchart of a digital human solution driven by a rule model shown in an exemplary embodiment;

[0053] Figure 3 It is a schematic structural diagram of an electronic device shown in an exemplary embodiment. Detailed Embodiments

[0054] To better understand the technical solution of this application, the embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0055] It should be clear that the described embodiments are only a part of the embodiments of this application, rather than all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of this application.

[0056] The terms used in the embodiments of this application are only for the purpose of describing specific embodiments, and are not intended to limit this application. The singular forms of "a", "the" and "said" used in the embodiments of this application and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise.

[0057] The comprehensive value of bank customers is a concept that comprehensively evaluates the total contributions of bank customers to the bank. It not only considers the direct economic benefits brought by customers to the bank, but also includes factors in multiple dimensions such as the potential value, loyalty, and credit risk of customers. By evaluating the comprehensive value of bank customers, it can help the bank better understand the overall contributions of customers, provide a decision-making basis for customer management, and enable the bank to conduct more precise customer stratification management, product design, and marketing strategy formulation.

[0058] Digital humans have the ability to provide 7*24-hour services. In terms of customer service, they can not only meet customer needs, improve customer experience and service efficiency, but also improve business processing capabilities, reduce labor costs, accelerate data collection and analysis, and promote internal innovation and exploration in terms of internal management. Remote banks can use digital humans to make improvements in the following aspects:

[0059] One is to improve the service experience of remote banks. Digital humans have characteristics such as anthropomorphism, AI-driven, strong plasticity, and high ornamental value, and can perform generative learning based on historical data, "remember" the preferences of different customer groups, realize customer classification and hierarchical management, provide personalized suggestions and services, and conduct customer operations and precision marketing according to customer stratification and classification to enhance the customer experience.

[0060] The second is to improve quality and efficiency. Digital humans can provide services 7*24 hours, enabling round-the-clock service. At the same time, they can handle customer requests concurrently to reduce waiting time, thereby improving service quality and efficiency.

[0061] The third is to reduce costs through digital operation. Some repetitive, mechanical consultations and high-frequency regular business processes in remote banking services can be automatically processed by digital humans, reducing the need for seat personnel and achieving cost reduction.

[0062] The present invention provides a method for digital human services in remote banking based on comprehensive value classification. By means of a specific algorithm model, the comprehensive value of customers is calculated, and then classification is carried out in combination with customer feature tags. Through digital human services, the optimal allocation and efficient utilization of bank service resources are promoted, marketing leads of potential similar customer groups are mined, the accuracy and effectiveness of remote banking services are comprehensively improved, and the innovative development of the bank's digital service system is promoted.

[0063] In the customer service business handling scenario of the present invention, based on the stratification and classification of customer groups on channels such as VTM and Zhe e-office, an application mode of "rule model digital human + manual backup" is adopted. The digital human mainly processes some repetitive and mechanical consultations and high-frequency regularized services in remote banking. When encountering abnormal situations or sensitive issues, the manual seat will provide backup. In terms of scenario selection, business processes with relatively clear rules and low error probabilities are preferably selected. For example, scenarios such as product consultation, large amount transfer, password reset, information modification, account upgrade and downgrade, and dormant account activation are used for service by the digital human seat.

[0064] At the same time, fully considering the acceptance degree of different customers for artificial intelligence technologies such as digital humans, balancing the service experience, intelligent operation and precision marketing of different customer groups, a customer classification and grading strategy is adopted to divide the customer group according to characteristics, so as to provide customized service solutions according to the specific needs and values of customers. At the same time, according to the acceptance degree of customers, customers are given the option to choose whether to close the digital human service or switch to the manual service, and the preferences of customers are remembered to achieve customer classification and grading management.

[0065] In addition, because the current large model cannot completely avoid the "hallucination" problem, that is, the model may contain false information, misleading content or even harmful information during the content generation process, which is likely to cause customer complaints and pose risks to the bank's public opinion and reputation. In financial business, the consumer protection side requires that the information provided to customers must ensure accuracy and reliability, and this reason has become an important factor restricting the direct service to customers by the large model. Therefore, the digital human in the remote banking customer service scenario selects some high-frequency, regularized and low-risk business scenarios to ensure the accuracy and reliability of the service. At the same time, in order to further reduce risks, a method of combining the digital human with the rule model and the manual seat for backup is adopted to provide services, which can not only utilize the advantage of the digital human in efficiently processing routine tasks, but also handle complex or sensitive situations through the intervention of the manual seat to ensure the success rate of the business process and risk control.

[0066] (1) Customer group stratification and classification modeling method

[0067] The common customer group classification method only classifies according to feature similarity and cannot well express the comprehensive value and marketing potential of customer groups. For banks, the classification results of customer groups should not only have similarity in label dimensions but also in comprehensive value. Based on this, the present invention proposes a customer group classification method based on the comprehensive value of customers, where the comprehensive value considers not only the current value of customers but also their potential future value. Classifying customer groups based on this can better reflect the maximization of long-term value rather than only considering the current customer value.

[0068] Step 1: Calculate the comprehensive value of customers and obtain the set TCV of the comprehensive values of all customers. The calculation method for the comprehensive value of each customer is as follows:

[0069]

[0070] where, tcv i is the comprehensive value of the i-th customer, MI is the monthly income of the customer, t is the number of periods, CR is the probability that the customer continues to stay after each period, f is the consumption frequency, i.e., the number of times the customer purchases the enterprise's products or services within a certain period, and d is the discount rate used to discount future income to the current value.

[0071] Step 2: Model the customer information into a data set u ij ∈U, i ∈ [1, n], j ∈ [1, m], n is the total number of customers, m is the total number of labels, u ij is the j-th dimension label of the i-th customer, and U is the set of customer information.

[0072] Among them, the customer label system is designed into seven categories: basic information, asset information, income contribution, product preference, consumption behavior, channel preference, and life cycle:

[0073] Basic information includes: gender, age, education level, unit, native place, etc.;

[0074] Asset information includes: customer AUM (Asset Under Management), daily average value and point value of various products, etc.;

[0075] Income contribution includes: high-value customers, medium-high-value customers, low-value customers, etc.;

[0076] Product preference includes: regular holding, preference for life payment, preference for bank wealth management, etc.;

[0077] Consumption behavior includes: consumption geographical location, consumption time period, brand preference, advertising and marketing preference, and consumption hot spot preference, etc.;

[0078] Channel preference: counter, online banking, mobile banking, remote banking, etc.;

[0079] The life cycle includes: acquisition period, promotion period, maturity period, decline period, and churn period.

[0080] Step 3: Preprocess the information of each dimension of the customer to obtain a set of customer information data The preprocessing method is:

[0081]

[0082] where \(i\in[1,n]\), \(j\in[1,m]\), \(u\) qj is the \(j\)-th dimension label of the \(q\)-th customer, \(u\) ij the \(j\)-th dimension label of the \(i\)-th customer, is the \(j\)-th dimension information of the \(i\)-th customer after preprocessing,

[0083] Step 4: Randomly select \(K\) pieces of customer information from the set as the set of center points for classification \(C = \{c\) 1 ,..., \(c\) k ,..., \(c\) K \}, where \(c\) k is the \(k\)-th center point, \(k\in[1,K]\).

[0084] Step 5: Calculate the distance from each piece of preprocessed customer information to each center point in \(C\), and calculate the distance in combination with the comprehensive value of the customer, so as to reflect the classification of the comprehensive value, and classify the customer into the sample in \(C\) with the shortest distance, forming a classified data set \(P=\{P\) 1 ,..., \(P\) k ,..., \(P\) K \}, and the distance calculation method is:

[0085]

[0086] where \(i\in[1,n]\), \(j\in[1,m]\), \(k\in[1,K]\), \(tcv\) i is the comprehensive value of the \(i\)-th customer, \(tcv\) k is the comprehensive value of the \(k\)-th customer in \(C\), \(c\) kj is the \(j\)-th dimension label of the \(k\)-th customer in \(C\), \(P\) k is the customer group classification set with the shortest distance to \(c\) k as the center point.

[0087] Step 6: For each classified data set \(P\) k , calculate the center point of each data set and update the center point. The calculation method of the center point is:

[0088]

[0089] Among them, k∈[1,K], j∈[1,m] are customer label dimensions, and s is the classified dataset P k The total number of samples in, p zj P k The zth element in P k The center point of the new j-th dimension label.

[0090] Step 7: Repeat steps 5 and 6 until the center point no longer changes significantly. The error threshold is set to σ, that is, r refers to the rth iteration.

[0091] 2. Remote banking digital human service method based on customer stratification and classification

[0092] When providing digital human services, the customer base P is classified according to the stratification. 1 ,...,P k ,...,P K}, set preferences, conduct operations management based on stratified and classified customer groups, conduct regular consultations with long-tail customers, explore customer needs, and conduct full life cycle management.

[0093] Digital human service solutions based on "rule model + manual backup" from the perspective of customer stratification and classification Figure 1 As shown, the main process is as follows:

[0094] Step 1: When a customer initiates a service request through a mobile client such as VTM or Zhejiang e-office, based on the principle of customer stratification and classification, the customer selects a service mode, which can be freely selected: virtual digital human agent service, manual video agent service, or branch counter service;

[0095] Step 2: When a customer chooses the virtual digital human agent service, the digital human will quickly judge and respond to the customer's consultation or business request based on built-in business rules and algorithms, and handle multiple services including FAQs and standardized business processes;

[0096] When customers choose human video agent service, or encounter complex or sensitive issues that exceed the preset rules of digital humans, human video agents will provide service;

[0097] When customers choose bank counter services, or when remote banking services cannot solve their problems, bank counter staff will handle the problem to ensure that the problem is handled in a more in-depth and personalized manner;

[0098] Step 3: When the digital human processes the business rules and passes the video quality inspection, or when the service is provided by a human video agent, the video will be recorded and stored as evidence;

[0099] Step 4: Complete the service and record all service interactions for subsequent supervision and review, customer stratification and classification, and system upgrade to improve the performance and service quality of the digital human.

[0100] (3) Digital human method driven by rule models

[0101] Since the current large models cannot completely avoid the "hallucination" problem, that is, the models may contain false information, misleading content or even harmful information during the content generation process, which is likely to cause customer complaints and pose risks to the bank's public opinion and reputation. Therefore, a digital human solution driven by rule models is adopted, as Figure 2 shown.

[0102] Step 1: Select low-risk, high-frequency, and regular business scenarios in the remote bank, establish a knowledge base, store it in the database, and configure corresponding answers. For example: scenarios such as large amount transfer, dormant account activation, information modification, password reset, card opening, SMS signing, voucher password loss reporting, personal deposit account cancellation, mobile banking signing, account upgrade, account downgrade, balance query, etc.

[0103] Step 2: During the voice conversation, after the remote bank server transmits the voice signal to the AI middle platform, the AI middle platform processes and generates a response through technologies such as ASR speech-to-text technology, NLP semantic understanding, and question-answer matching. After the server receives and parses the response, it makes corresponding feedback.

[0104] Step 3: For transaction call-type answers, the server will execute the corresponding transaction service, such as balance query; for display-type answers, the server will display the answer at the specified position on the interface; for interactive instructions (such as action instructions, broadcast text), the server generates voice from the broadcast text through the TTS service, and at the same time sends the instruction text to the digital human. The driving engine controls the facial and limb movements, and the rendering engine renders and synthesizes elements such as lip shapes and actions, thereby generating video frames and pushing them to the digital human interface for playback.

[0105] (4) Digital human marketing method based on stratification and classification

[0106] For the stratified and classified customer groups P = {P 1 ,..., P k ,..., P K}, precise marketing is carried out for each layer of customer group P k . Because the customer groups after stratification are similar, they have similarities in product preferences and product purchases, making it easier to achieve successful marketing. The specific digital human service method for the remote bank is as follows:

[0107] Step 1: For the classified customer groups P = {P 1 ,..., P k ,..., PK}, for each customer group P k mine the product purchase set V of this customer group;

[0108] The second step: for this customer group P k , for each customer, use a digital human for automated marketing according to the complement of the purchased products and V;

[0109] The third step: repeat the first step and the second step to market all classified customer groups.

[0110] Correspondingly, the present application also provides an electronic device, including: one or more processors; a memory for storing one or more programs; when the one or more programs are executed by the one or more processors, the one or more processors implement the remote banking digital human service method based on comprehensive value classification as described above. As Figure 3 shown, it is a hardware structure diagram of any device with data processing capabilities where the remote banking digital human service method based on comprehensive value classification provided by the embodiment of the present invention is located. In addition to Figure 3 the processors, memory, and network interfaces shown, any device with data processing capabilities where the device in the embodiment is located usually also includes other hardware according to the actual functions of the any device with data processing capabilities, which will not be elaborated here.

[0111] Correspondingly, the present application also provides a computer-readable storage medium, on which computer instructions are stored. When the instructions are executed by a processor, the remote banking digital human service method based on comprehensive value classification as described above is implemented. The computer-readable storage medium may be an internal storage unit of any device with data processing capabilities described in any of the foregoing embodiments, such as a hard disk or memory. The computer-readable storage medium may also be an external storage device, such as a plug-in hard disk, a Smart Media Card (SMC), an SD card, a Flash Card, etc. equipped on the device. Further, the computer-readable storage medium may also include both an internal storage unit of any device with data processing capabilities and an external storage device. The computer-readable storage medium is used to store the computer program and other programs and data required by any device with data processing capabilities, and may also be used to temporarily store data that has been output or will be output.

[0112] Other embodiments of the present application will be readily contemplated by those skilled in the art upon consideration of the specification and practice of the disclosure herein. The present application is intended to cover any variations or adaptations of the present application, which follow the general principles of the present application and include well-known common general knowledge or conventional technical means in the technical field not disclosed in the present application. The specification and examples are only to be considered exemplary.

[0113] It should be understood that the present application is not limited to the exact structures described above and shown in the drawings, and various modifications and changes can be made without departing from its scope.

[0114] The above are only the preferred embodiments of the present invention. Although the present invention has been disclosed above with preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make many possible changes and modifications to the technical solution of the present invention, or modify it into equivalent embodiments with equivalent changes, without departing from the scope of the technical solution of the present invention. Therefore, any simple modification, equivalent change and modification made to the above embodiments according to the technical essence of the present invention without departing from the content of the technical solution of the present invention are still within the scope of the protection of the technical solution of the present invention.

Claims

1. A remote banking digital human service method based on comprehensive value classification, characterized in that: include: Calculate the comprehensive value of customers and combine them with customer feature labels to perform customer stratification and classification modeling; Remote banking digital human services are provided by combining rule models with manual backup. When a customer initiates a service request, the customer selects a service mode based on the customer stratification and classification principle. The service modes include: virtual digital human agent service, manual video agent service, and branch counter service. When a customer chooses a virtual digital agent service, the digital agent will quickly judge and respond to the customer's consultation or business request based on built-in business rules; When customers choose human video agent service, or encounter complex or sensitive issues that exceed the preset rules of digital humans, human video agents will provide service; When customers choose bank counter services, or when remote banking services cannot solve the problem, the bank counter staff will handle it; When the digital human processes the business rules and passes the video quality inspection, or when the service is provided by a manual video agent, dual video recording will be performed for evidence storage; after the service is completed, all service interactions will be recorded.

2. The remote banking digital human service method based on comprehensive value classification according to claim 1 is characterized in that: The digital human agent service is mainly used to handle repetitive, mechanical consultations and high-frequency, regularized, low-risk businesses in remote banking, with priority given to the following business processes: product consultation, large-value transfers, password reset, information modification, account upgrades and downgrades, and activation of inactive accounts.

3. The remote banking digital human service method based on comprehensive value classification according to claim 1 is characterized in that: The customer group stratification classification modeling includes the following steps: (1) Calculate the comprehensive value of the customer using the following formula: Among them, tcv i is the comprehensive value of the ith customer, MI is the customer's monthly income, t is the number of cycles, CR is the probability that the customer will continue to stay after each cycle, f is the consumption frequency, and d is the discount rate; (2) Model the customer information dataset ij ∈U, i∈[1,n], j∈[1,m], n is the total number of customers, m is the total number of tags, u ij is the j-th dimension label of the i-th customer, and U is the customer information set; (3) Preprocess the customer information in each dimension to obtain a customer information data set The formula is as follows: Among them, i∈[1,n],j∈[1,m],u qj is the j-th dimension label of the q-th customer, u ij The j-th dimension label of the i-th customer, is the j-th dimension information of the i-th customer after preprocessing, (4) From the collection Randomly select K customer information as the classification center point set C = {c1,...,c k ,...,c K }, where c k is the kth center point, k∈[1,K]; (5) For each preprocessed customer information, calculate the distance to each center point in C, and calculate the distance based on the customer's comprehensive value, and classify the customer into the sample in C with the shortest distance, forming a classified data set P = {P1, ..., P k ,...,P K }, the distance calculation formula is: Among them, i∈[1,n],j∈[1,m],k∈[1,K], tcv i is the comprehensive value of the i-th customer, tcv k is the comprehensive value of the kth customer in C, c kj is the j-th dimension label of the k-th customer in C, P k For c k The customer group classification set closest to the center point; (6) Calculate each data set P after classification k The center point is updated and the center point calculation formula is: Among them, k∈[1,K],j∈[1,m], s is the classified data set P k The total number of samples in, p zj P k The zth element in P k The center point of the new j-th dimension label; (7) Repeat steps (5) and (6) until the center point no longer changes significantly. The error threshold is set to σ, that is, r refers to the rth iteration.

4. The remote banking digital human service method based on comprehensive value classification according to claim 3 is characterized in that: The customer's labeling system is designed into seven categories: basic information, asset information, revenue contribution, product preference, consumer behavior, channel preference, and life cycle: The basic information includes: gender, age, education, unit, and place of origin; The asset information includes: customer AUM, daily average and point-in-time value of various products; The revenue contribution includes: high-value customers, medium-high-value customers, and low-value customers; The product preferences include: regular holding, living expense payment preference, and bank financial management preference; The consumption behavior includes: consumption location, consumption time, brand preference, advertising and marketing preference and consumption hot spot preference; Stated channel preferences: counter, online banking, mobile banking, remote banking; The life cycle includes: acquisition period, promotion period, maturity period, decline period and loss period.

5. The remote banking digital human service method based on comprehensive value classification according to claim 1 is characterized in that: When providing digital human services, preference settings and operational management are performed based on stratified and classified customer groups, and long-tail customers are consulted regularly to explore customer needs and conduct full life cycle management.

6. The remote banking digital human service method based on comprehensive value classification according to claim 1 is characterized in that: The digital human service driven by the rule model includes the following steps: (1) Select high-frequency, regularized, and low-risk business scenarios in remote banking, establish a knowledge base, store it in the database, and configure the corresponding answers; (2) During the voice conversation, the remote banking server transmits the voice signal to the middle station, which then processes and generates a response using ASR speech-to-text technology, NLP semantic understanding, and question-answer matching technology. The server receives and parses the response and then provides corresponding feedback. (3) For transaction-based answers, the server will execute the corresponding transaction service; for display-based answers, the server will display the answer at the specified location on the interface; for interactive commands, the server will generate speech from the broadcast text through the TTS service, and send the command text to the digital human at the same time. The driving engine will control the facial and body movements, and the rendering engine will render the synthesized lip movements and action elements, thereby generating video frames and pushing them to the digital human interface for playback.

7. The remote banking digital human service method based on comprehensive value classification according to claim 1 is characterized in that: Digital human marketing based on customer group stratification and classification results includes the following steps: (1) For the classified customer group P = {P1, ..., P k ,...,P K }, for each customer group P k Dig out the product purchase set V of this type of customer group; (2) For this customer group P k , for each customer, use digital people to conduct automated marketing according to the complement of purchased products and V; (3) Repeat steps (1) and (2) to conduct precision marketing for all customer groups.

8. An electronic device, comprising a memory and a processor, characterized in that: The memory is coupled to the processor; wherein, the memory is used to store program data, and the processor is used to execute the program data to implement the remote banking digital human service method as described in any one of claims 1-7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by the processor, it implements the remote banking digital human service method as described in any one of claims 1-7.

10. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instructions are executed by a processor, the remote banking digital human service method as described in any one of claims 1-7 is implemented.

Images