Task processing method and device of website and computer program product

Abstract

The invention discloses a task processing method and device of a website and a computer program product. Relates to the field of artificial intelligence, and the method comprises the steps that a website collaboration graph of M websites of a financial institution is determined, M is a positive integer, the website collaboration graph is constructed by related data of the M websites, the website collaboration graph comprises various types of nodes and edges, and the nodes comprise at least one of website nodes, problem nodes and operation and maintenance nodes; under the condition that a website task is received, a target feature vector associated with the website task is extracted from the website collaboration map, and the website task comprises at least one of website queuing number prediction, problem occurrence probability prediction and solution recommendation; and inputting the target feature vector into the target model to obtain a processing result of the network task. Through the method and the device, the problem of low task processing efficiency of websites in related technologies is solved.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence, and more specifically, to a task processing method, apparatus, and computer program product for a network of outlets. Background Technology

[0002] Financial institutions' physical branch networks are facing increasingly severe collaboration challenges, particularly in terms of efficiency, agility, and intelligence. In daily operations, the valuable knowledge assets accumulated by each branch—such as successful service cases for resolving complex customer complaints, risk management plans for identifying and preventing new transaction risks, and successful promotional strategies targeting specific customer groups—are primarily shared through manual compilation, email transmission, or static knowledge base archiving. This model suffers from significant lag; new branches or branches encountering similar problems cannot obtain relevant solutions in real time and accurately, resulting in low reuse rates of best practices, high costs for repeatedly handling similar problems, and slow improvement in overall service levels and risk control capabilities. Facing unpredictable business fluctuations, such as sudden customer traffic surges, large-scale events attracting customers, or localized failures (e.g., a branch's core system outage or centralized maintenance of ATMs), communication relies heavily on methods like telephone and email, requiring a hierarchical reporting, approval, and coordination mechanism from branch to sub-branch to head office. This chain-response mechanism is not only time-consuming but also prone to information distortion during transmission. Customers' valuable time while waiting for service is wasted on lengthy processes, leading to a sharp deterioration in customer experience, excessively long queues, and even customer abandonment, which seriously damages the reputation and business opportunities of financial institutions.

[0003] Among related technologies, tools widely used within financial institutions, such as internal forums, systems, email groups, and static knowledge bases, while intended to promote information sharing, are essentially "passive" information dissemination platforms. The core flaw of these tools lies in their inability to perceive the dynamic operational status of the entire branch network in real time and automatically. They struggle to characterize and quantify the complex and multi-dimensional business relationships between branches, resulting in existing management mechanisms failing to fully leverage the potential synergistic advantages between branches and failing to meet the high-efficiency and intelligent operational needs of financial institutions.

[0004] There is currently no effective solution to the problem of low task processing efficiency at network points in related technologies. Summary of the Invention

[0005] The main objective of this application is to provide a task processing method, apparatus, and computer program product for service outlets, in order to solve the problem of low task processing efficiency in related technologies.

[0006] To achieve the above objectives, according to one aspect of this application, a branch task processing method is provided. The method includes: determining a branch collaboration graph of M branches of a financial institution, where M is a positive integer; the branch collaboration graph is constructed from relevant data of the M branches; the branch collaboration graph includes multiple types of nodes and edges; the nodes include at least one of the following: branch nodes, problem nodes, and maintenance nodes; upon receiving a branch task, extracting a target feature vector associated with the branch task from the branch collaboration graph; wherein the branch task includes at least one of the following: branch queue length prediction, problem occurrence probability prediction, and solution recommendation; inputting the target feature vector into a target model to obtain the processing result of the branch task; wherein the target model is trained from multiple sets of training samples, each set of training samples including historical feature vectors and historical processing results.

[0007] Optionally, the network point collaboration graph is constructed as follows: For each network point, network point attribute data, problem data, and operation and maintenance data are extracted from relevant data. Network point nodes are constructed based on network point attribute data, problem nodes are constructed based on problem data, and operation and maintenance nodes are constructed based on operation and maintenance data. First-type edges are connected between all network point nodes. Second-type edges are connected between network point nodes and problem nodes based on the network point's historical problems. Third-type edges are connected between operation and maintenance nodes and problem nodes based on the problems solved by the historical business processing cases of operation and maintenance nodes. The weight of each type of edge is determined, and the network point collaboration graph is composed of network point nodes, problem nodes, operation and maintenance nodes, first-type edges, second-type edges, third-type edges, and the weight of each type of edge.

[0008] Optionally, extracting branch attribute data, problem data, and operation and maintenance data from relevant data includes: identifying at least one of the following as branch attribute data: geographical location, customer base, and queuing status; identifying at least one of the following as problem data: problem type, solution, and solution effectiveness evaluation value, where problem types include customer complaints, business system problems, and business risk problems; and identifying at least one of the following as operation and maintenance data: employee ID, business processing type, and historical business processing cases. Determining the weight of each type of edge includes: determining the first weight of the first type of edge based on the geographical distance between branches and the similarity of customer base; determining the second weight of the second type of edge based on the customer base characteristics of branch nodes and the solution characteristics of problem nodes; and determining the third weight of the third type of edge based on the solution effectiveness evaluation value and the matching degree between business processing type and problem type.

[0009] Optionally, determining the first weight of the first type of edge based on the geographical distance between outlets and the similarity of business customer groups includes: determining the first outlet node and the second outlet node connected by the first type of edge; determining the first business customer group and the first geographical location information of the first outlet node; determining the second business customer group and the second geographical location information of the second outlet node; extracting the first feature of the first business customer group and the second feature of the second business customer group, wherein the first feature or the second feature includes at least one of the following: customer group type, amount of funds, and quantity distribution of customer group type, and the customer group type includes at least one of the following: individuals, enterprises, and related departments; calculating the geographical distance based on the first geographical location information and the second geographical location information; calculating the cosine similarity between the first feature and the second feature; and calculating the product of the cosine similarity and the geographical distance to obtain the first weight.

[0010] Optionally, determining the second weight of the second type of edge based on the business customer group characteristics of the network node and the solution characteristics of the problem node includes: determining the third network node and the first problem node connected by the second type of edge, determining the third business customer group of the third network node and the first solution of the first problem node; extracting the third feature of the third business customer group and the solution text of the first solution; converting the third feature into a demand vector and the solution text into a semantic vector through a pre-trained language model; concatenating the demand vector and the semantic vector to obtain a concatenated vector, and inputting the concatenated vector into an activation function to obtain the second weight.

[0011] Optionally, determining the third weight of the third type of edge based on the matching degree between the solution effect evaluation value and the business processing type and the problem type includes: determining the first maintenance node and the second problem node connected by the third type of edge; determining the target business processing type of the first maintenance node and the target solution effect evaluation value and the target problem type of the second problem node; determining whether the target business processing type and the target problem type match; if the target business processing type and the target problem type match, determining the matching degree as 1; if the target business processing type and the target problem type do not match, determining the matching degree as 0; calculating the product of the target solution effect evaluation value and the matching degree to obtain the third weight.

[0012] Optionally, extracting target feature vectors associated with network tasks from the network collaboration graph includes: when the network task is to predict the number of people queuing at the network, converting the queuing status of each network node in the network collaboration graph into a target feature vector; when the network task is to predict the probability of a problem occurring, extracting the network feature vector of each network node and the problem feature vector of each problem node from the network collaboration graph, and determining the network feature vector and the problem feature vector as the target feature vector; when the network task is to recommend a solution, extracting the expert feature vector of each operation and maintenance node and the problem feature vector of each problem node from the network collaboration graph, and determining the expert feature vector and the problem feature vector as the target feature vector.

[0013] Optionally, inputting the target feature vector into the target model to obtain the processing results of the network task includes: when the network task is to predict the number of people queuing at the network, inputting the target feature vector into the target model to obtain the predicted number of people queuing at each network node within the target period, determining a passenger flow allocation scheme based on the predicted number of people queuing at each network node, and determining the passenger flow allocation scheme as the processing result; when the network task is to predict the probability of a problem occurring, inputting the target feature vector into the target model to obtain the predicted probability of each network node for each type of problem, pushing a solution for the corresponding type of problem to the network node when the predicted probability of occurrence is greater than or equal to the probability threshold, and determining the predicted probability of occurrence and the solution as the processing result; when the network task is to recommend a solution, inputting the target feature vector into the target model to obtain the probability that the problem type corresponding to each problem node will be solved by the expert corresponding to each operation and maintenance node, and determining the expert solution probability sequence for each problem node as the processing result, wherein the expert solution probability sequence is sorted from largest to smallest according to the probability of the expert solving the problem type corresponding to the problem node.

[0014] To achieve the above objectives, according to another aspect of this application, a branch task processing apparatus is provided. The apparatus includes: a first determining unit, configured to determine a branch collaboration graph of M branches of a financial institution, where M is a positive integer, the branch collaboration graph being constructed from relevant data of the M branches, the branch collaboration graph including various types of nodes and edges, the nodes including at least one of the following: branch nodes, problem nodes, and maintenance nodes; an extraction unit, configured to extract a target feature vector associated with the branch task from the branch collaboration graph upon receiving a branch task, wherein the branch task includes at least one of the following: branch queue length prediction, problem occurrence probability prediction, and solution recommendation; and an input unit, configured to input the target feature vector into a target model to obtain the processing result of the branch task, wherein the target model is trained from multiple sets of training samples, each set of training samples including historical feature vectors and historical processing results.

[0015] In this embodiment, a branch collaboration graph is constructed using M branches of a financial institution, where M is a positive integer. The branch collaboration graph is built from relevant data of the M branches and includes various types of nodes and edges. Nodes include at least one of the following: branch nodes, problem nodes, and maintenance nodes. Upon receiving a branch task, a target feature vector associated with the branch task is extracted from the branch collaboration graph. The branch task includes at least one of the following: branch queue length prediction, problem occurrence probability prediction, and solution recommendation. The target feature vector is input into a target model to obtain the processing result of the branch task. The target model is trained using multiple sets of training samples, each set including historical feature vectors and historical processing results. By integrating multiple data sources from various branches to construct the branch collaboration graph, the matching between problem nodes and maintenance nodes ensures that when a branch encounters a specific problem, it can quickly obtain the most suitable expert or solution to solve the problem. This accelerates the reuse of excellent service cases, risk management plans, and strategies, and improves the overall service level and risk control capabilities of the branches. Through a network collaboration map, intelligent matching of resources and capabilities across networks can be achieved based on factors such as customer profile similarity, complementary service capabilities, and the spatiotemporal characteristics of business traffic. For example, geographically adjacent networks can share promotional strategies or allocate human resources when one network experiences peak customer traffic; networks with expertise can provide advice or support in specific business areas to other networks, achieving efficient task processing and synergistic effects between networks. This improves the technical efficiency of network task processing and solves the technical problem of low network task processing efficiency. Attached Figure Description

[0016] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings:

[0017] Figure 1 A hardware structure block diagram of a computer terminal (or mobile device) for implementing a task processing method for network points is shown.

[0018] Figure 2 This is a flowchart of a task processing method for service outlets provided according to an embodiment of this application;

[0019] Figure 3 This is a flowchart illustrating the construction of training samples for the target model provided in the embodiments of this application;

[0020] Figure 4 This is a schematic diagram of a task processing device for a service point provided according to an embodiment of this application;

[0021] Figure 5This is a structural block diagram of an electronic device according to an embodiment of this application. Detailed Implementation

[0022] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0023] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0024] It should be noted that the information collected in this application (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for display, data used for analysis, etc.) are information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, storage, use, processing, transmission, provision, disclosure, and application of this data all comply with relevant laws, regulations, and standards, necessary confidentiality measures have been taken, and they do not violate public order and good morals. Corresponding access points are provided for users to choose to authorize or refuse. For example, interfaces are set up between this system and relevant users or organizations, providing users with corresponding access points to choose to agree to or refuse automated decision-making results; if the user chooses to refuse, the process proceeds to the expert decision-making stage.

[0025] Example 1

[0026] According to an embodiment of this application, a method embodiment for processing tasks at a branch office is also provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0027] The method embodiment provided in Embodiment 1 of this application can be executed on a mobile terminal, computer terminal, or similar computing device. Figure 1 A hardware block diagram of a computer terminal (or mobile device) for implementing a task processing method for branch offices is shown. Figure 1 As shown, the computer terminal 10 (or mobile device) may include one or more processors 102 (shown as 102a, 102b, ..., 102n in the figure) 102 (processor 102 may include, but is not limited to, processing devices such as MCU (Microcontroller Unit) or FPGA (Field-Programmable Gate Array), memory 104 for storing data, and transmission device 106 for communication functions. In addition, it may also include: a display, input / output interfaces (I / O interfaces), a USB (Universal Serial Bus) port (which may be included as one of the ports of a BUS (Business Bus), a network interface, a power supply, and / or a camera. Those skilled in the art will understand that... Figure 1 The structure shown is for illustrative purposes only and does not limit the structure of the aforementioned electronic device. For example, computer terminal 10 may also include... Figure 1 The more or fewer components shown, or having the same Figure 1 The different configurations shown.

[0028] It should be noted that the aforementioned one or more processors 102 and / or other data processing circuits are generally referred to herein as "data processing circuits". These data processing circuits may be embodied, in whole or in part, in software, hardware, firmware, or any other combination thereof. Furthermore, the data processing circuits may be a single, independent processing module, or may be integrated, in whole or in part, into any other element within the computer terminal 10 (or mobile device). As involved in the embodiments of this application, the data processing circuits serve as a processor control mechanism (e.g., selection of a variable resistor termination path connected to an interface).

[0029] The memory 104 can be used to store software programs and modules of application software, such as the program instructions / data storage device corresponding to the task processing method for the network outlet in this embodiment. The processor 102 executes various functional applications and data processing by running the software programs and modules stored in the memory 104, thereby realizing the aforementioned task processing method for the network outlet. The memory 104 may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include memory remotely located relative to the processor 102, and these remote memories can be connected to the computer terminal 10 via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0030] The transmission device 106 is used to receive or send data via a network. Specific examples of the network described above may include a wireless network provided by the communication provider of the computer terminal 10. In one example, the transmission device 106 includes a Network Interface Controller (NIC), which can connect to other network devices via a base station to communicate with the Internet. In another example, the transmission device 106 may be a Radio Frequency (RF) module, used for wireless communication with the Internet.

[0031] The display may be, for example, a touchscreen LCD display that allows the user to interact with the user interface of the computer terminal 10 (or mobile device).

[0032] Under the above operating environment, this application provides a task processing method for network points. Figure 2 This is a flowchart of a task processing method for service outlets provided according to an embodiment of this application, such as... Figure 2 As shown, the method includes:

[0033] Step S201: Determine the branch collaboration graph of M branches of the financial institution, where M is a positive integer. The branch collaboration graph is constructed from the relevant data of the M branches. The branch collaboration graph includes multiple types of nodes and edges. The nodes include at least one of the following: branch nodes, problem nodes, and operation and maintenance nodes.

[0034] In step S201, the node types in the network collaboration graph are defined, including but not limited to network nodes, problem nodes, and maintenance nodes. These nodes carry corresponding attribute information. For example, the attributes of network nodes may include geographical location, business scale, customer group tags, real-time status, etc.; problem nodes may include problem type, solution text, effect rating, etc.; maintenance nodes may be associated with employee ID, skill tags, historical solution case ID, etc. The edges in the graph are used to connect different types of nodes, representing the dynamic relationship between different networks, the correlation between problems and networks, and the matching degree between maintenance personnel and problems.

[0035] Step S202: Upon receiving a branch task, extract the target feature vector associated with the branch task from the branch collaboration graph, wherein the branch task includes at least one of the following: branch queue number prediction, problem occurrence probability prediction, and solution recommendation.

[0036] In step S202, the received branch task is identified, and the task type is determined, such as branch queue length prediction, problem occurrence probability prediction, or solution recommendation. Based on the task type, the nodes related to the task in the graph are located. For example, for the branch queue length prediction task, the branch node and its related real-time status and historical business data are located; for the problem occurrence probability prediction task, the problem node and its association edge with the branch are located; for the solution recommendation task, the relationship between the problem node and the operation and maintenance node is located.

[0037] Feature vectors are extracted from the located nodes. For predicting the number of people queuing at a service point, attribute feature vectors of the service point nodes are extracted, such as geographical location, business scale, and real-time queuing status. For predicting the probability of a problem occurring, attribute feature vectors of the problem node are extracted, such as problem type and historical solution effectiveness ratings. For solution recommendation, attribute feature vectors of the operation and maintenance node are extracted, such as employee skill tags and historical solution cases. Furthermore, based on the characteristics of graph neural networks, feature vectors from other nodes related to the target service point need to be fused, such as by using edge weight transfer mechanisms to integrate information from neighboring service points, historical cases, and employee experience. During the feature vector extraction process, dynamic edge weight calculation methods (such as coupled calculation of geographical distance and business similarity) are used to adjust the strength of the association between service points, ensuring that the extracted feature vectors accurately reflect the actual business collaboration relationships between service points and the most relevant information in the current task context. Combining the above steps, a target feature vector is generated. This vector contains all the necessary information related to the service point task and is used as input to the model to predict the number of people queuing at the service point, the probability of a problem occurring, or to recommend solutions.

[0038] Step S203: Input the target feature vector into the target model to obtain the processing result of the network task. The target model is trained by multiple sets of training samples, and each set of training samples includes historical feature vectors and historical processing results.

[0039] In step S203, the target model is trained based on historical data. It can predict future queue lengths, problem probabilities, or recommended solutions based on the input feature vector. The model can be built based on a relational graph convolutional network or other types of graph neural networks to ensure it can handle network collaboration graphs and learn complex relationships between networks. The extracted target feature vector is fed into the target model as input. The target model processes the input feature vector through internal computational mechanisms (such as message passing, node feature updates, and weighted aggregation), ultimately outputting the processing result of the network task, for example:

[0040] For predicting queue length at service outlets, the model outputs the predicted queue length at a future point in time for each outlet, along with possible confidence intervals or uncertainty measures and recommended alternative outlets. For predicting the probability of problems occurring, the model provides predictions of the future occurrence probability of specific problem types at each outlet, such as customer complaints, system failures, and transaction risks. For recommending solutions, the model can recommend the most suitable solutions or experts based on the current problem and the characteristics of the outlet, enabling a quick and effective response to the problem.

[0041] For example, Figure 3 This is a flowchart illustrating the construction of training samples for the target model provided in the embodiments of this application, such as... Figure 3 As shown, a dataset of relevant data from M network points is obtained. The original data first undergoes feature enhancement to supplement temporal features, such as adding peak and off-peak time periods and holiday markers, to enhance the model's ability to capture business cycle. Spatial features, such as regional economic indices (e.g., population density and GDP), are introduced to enable the model to distinguish business differences among network points with different economic levels. Data noise is cleaned; for continuous variables such as queue length, extreme values ​​outside the 5% to 95th percentile are removed to avoid abnormal data interfering with model training. Missing fields are filled in using the mean or mode to ensure data integrity. Unstructured text is encoded into machine-processable vectors, preserving the deep semantic information of the text. Graph construction includes node creation: defining three types of nodes, each with structured attributes; calculating edge weights based on business logic; and sorting the data chronologically, with the first 80% used as the training set and the last 20% as the test set to ensure that the model evaluation reflects the time dependencies of real business scenarios.

[0042] The branch task processing method provided in this application embodiment determines a branch collaboration graph of M branches of a financial institution, where M is a positive integer. The branch collaboration graph is constructed from relevant data of the M branches and includes various types of nodes and edges. Nodes include at least one of the following: branch nodes, problem nodes, and maintenance nodes. Upon receiving a branch task, a target feature vector associated with the branch task is extracted from the branch collaboration graph. The branch task includes at least one of the following: branch queue length prediction, problem occurrence probability prediction, and solution recommendation. The target feature vector is input into a target model to obtain the processing result of the branch task. The target model is trained by multiple sets of training samples, each set of training samples including historical feature vectors and historical processing results. By integrating multiple data sources from various branches to construct the branch collaboration graph, the matching between problem nodes and maintenance nodes ensures that when a branch encounters a specific problem, it can quickly obtain the most suitable expert or solution to solve the problem. This accelerates the reuse of excellent service cases, risk management plans, and strategies, and improves the overall service level and risk control capabilities of the branches. Through a network collaboration map, intelligent matching of resources and capabilities across networks can be achieved based on factors such as customer profile similarity, complementary service capabilities, and the spatiotemporal characteristics of business traffic. For example, geographically adjacent networks can share promotional strategies or allocate human resources when one network experiences peak customer traffic; networks with expertise can provide advice or support in specific business areas to other networks, achieving efficient task processing and synergistic effects between networks. This improves the technical efficiency of network task processing and solves the technical problem of low network task processing efficiency.

[0043] To learn the collaborative relationships between different network points, a network point collaboration graph is constructed based on relevant network point data. Optionally, in the network point task processing method provided in this application embodiment, the network point collaboration graph is constructed in the following way: For each network point, network point attribute data, problem data, and operation and maintenance data are extracted from relevant data; network point nodes are constructed based on network point attribute data, problem nodes are constructed based on problem data, and operation and maintenance nodes are constructed based on operation and maintenance data; first-type edges are connected between all network point nodes; second-type edges are connected between network point nodes and problem nodes based on historical problems of the network point; third-type edges are connected between operation and maintenance nodes and problem nodes based on problems solved by historical business processing cases of operation and maintenance nodes; the weight of each type of edge is determined, and the network point collaboration graph is composed of network point nodes, problem nodes, operation and maintenance nodes, first-type edges, second-type edges, third-type edges, and the weight of each type of edge.

[0044] In some embodiments, the definition of a branch network collaboration graph node may include: Branch node: attributes = {geographical location, business scale, customer group tags, real-time status}, where the customer group is divided into enterprises and individuals; the real-time status includes the number of people in queue and ATM malfunction indicators; Problem node: attributes = {problem type, solution text, effectiveness rating}, where the problem type is divided into customer complaints / system issues / risk control issues, and the effectiveness rating can be defined as customer satisfaction and resolution time; Operation and maintenance node: attributes = {employee ID, skill tags, historical resolution case ID}. The weight of the first type of edge = geographical distance × business similarity, where business similarity can be calculated using cosine similarity to determine customer group distribution; the second type of edge points to historical cases that occurred at the branch; the third type of edge points to the employee who resolved the case and their solution rating.

[0045] This embodiment constructs a branch network collaboration graph to optimize resource scheduling among branches, accurately match experience and knowledge, and coordinate emergency response. The system integrates geospatial information, real-time business data, historical case libraries, and employee skill tags. It utilizes the message passing mechanism of the target model to uncover implicit connections between branches, resolving the inefficiencies, delays, and information silos inherent in manual coordination. It is suitable for scenarios involving multi-branch management at the branch level in financial institutions, regional service collaboration, and joint risk prevention and control.

[0046] Optionally, in the task processing method for service outlets provided in this application embodiment, extracting service outlet attribute data, problem data, and operation and maintenance data from relevant data includes: determining at least one of the following as service outlet attribute data: geographical location, business customer group, and queuing status; determining at least one of the following as problem data: problem type, solution, and solution effectiveness evaluation value, where problem types include customer complaint issues, business system issues, and business risk issues; and determining at least one of the following as operation and maintenance data: employee number, business processing type, and historical business processing cases. Determining the weight of each type of edge includes: determining a first weight of a first type of edge based on the geographical distance between service outlets and the similarity of business customer groups; determining a second weight of a second type of edge based on the business customer group characteristics of service outlet nodes and the solution characteristics of problem nodes; and determining a third weight of a third type of edge based on the solution effectiveness evaluation value and the matching degree between business processing type and problem type.

[0047] In some embodiments, the following parameters are considered: Geographic Location: Obtaining GPS coordinates for each branch. Customer Base: Analyzing the main customer types served by the branches, including but not limited to individuals, micro-enterprises, and large enterprises, as well as segmented customer characteristics such as age, occupation, and consumption habits. Queue Status: Collecting real-time and historical information such as queue length and waiting time. Problem Type: Categorizing various problems encountered by the branches, such as customer complaints, business system problems, and business risk issues. Solutions: Recording effective solutions for various problems, including processing procedures, required resources, and time. Solution Effectiveness Evaluation: Evaluating the effectiveness of solutions based on customer satisfaction surveys, problem resolution speed, and follow-up event tracking. Employee ID: A unique ID identifying each employee. Business Processing Type: Categorizing employees' business processing capabilities, covering different service areas such as deposits, loans, and foreign exchange. Historical Business Processing Cases: Recording past problem cases solved by employees, including problem type, processing time, and customer feedback.

[0048] The first weight of the first type of edge (branch-branch edge): Geographical distance: The distance between branches is calculated using algorithms such as Euclidean distance or Manhattan distance. Similarity is quantified by calculating the cosine similarity of the customer distribution of each branch, reflecting potential collaborative relationships between branches. The second weight of the second type of edge (branch-problem edge): Matching degree between business customer characteristics and solution characteristics: Deep learning methods are used to encode the solution text, generating semantic feature vectors. These vectors are then processed with the branch customer feature vectors using the Sigmoid function to measure the likelihood of a branch encountering a specific problem. The third weight of the third type of edge (Operations-Problem edge): Quantifying the actual effectiveness of the solution, such as customer satisfaction ratings and problem resolution speed. The degree of matching between employee processing capabilities and problem requirements is calculated and quantified through logical operations such as the ratio of intersection and union.

[0049] This embodiment, through refined data extraction and node / edge definition, combined with a reasonable weight calculation method, can construct a dynamic collaborative graph that can reflect the real-time operational status of branches, problem prediction capabilities, and employee problem-solving efficiency. This provides financial institutions with a powerful knowledge management and resource scheduling tool, greatly improving operational efficiency and service quality.

[0050] The first weight is determined by the geographical distance between outlets and the similarity of business customer groups. Optionally, in the task processing method for outlets provided in this application embodiment, determining the first weight of the first type of edge based on the geographical distance between outlets and the similarity of business customer groups includes: determining the first outlet node and the second outlet node connected by the first type of edge; determining the first business customer group and the first geographical location information of the first outlet node; determining the second business customer group and the second geographical location information of the second outlet node; extracting the first feature of the first business customer group and the second feature of the second business customer group, wherein the first feature or the second feature includes at least one of the following: customer group type, amount of funds, and quantity distribution of customer group type, and the customer group type includes at least one of the following: individuals, enterprises, and related departments; calculating the geographical distance based on the first geographical location information and the second geographical location information; calculating the cosine similarity between the first feature and the second feature; and calculating the product of the cosine similarity and the geographical distance to obtain the first weight.

[0051] In some embodiments, the weighting of outlets is determined by a coupling of geographical location and customer group tags. Customer group similarity can be calculated using cosine similarity, and vector di is the customer group distribution vector. The formula for calculating the first weight is as follows:

[0052] ;

[0053] Where i is the first network node, j is the second network node, and w ij As the first weight, loc i It is the primary geographic location information, loc j This is the second geographic location information, γ is the geographic location attenuation parameter, and d i It is the first feature, d j This is the second characteristic.

[0054] This embodiment calculates the first weight based on the geographical distance between branches and the similarity of business customer groups. This can accurately quantify the collaborative relationships between branches, provide data-driven decision support for the intelligent management of branches, improve service efficiency, reduce customer waiting time, optimize resource allocation, and enhance the adaptability and collaborative capabilities of the bank's branch network.

[0055] The second weight can be calculated from the business customer group characteristics and solution characteristics. Optionally, in the task processing method for network nodes provided in this application embodiment, determining the second weight of the second type of edge based on the business customer group characteristics of the network node and the solution characteristics of the problem node includes: determining the third network node and the first problem node connected by the second type of edge, determining the third business customer group of the third network node and the first solution of the first problem node; extracting the third feature of the third business customer group and the solution text of the first solution; converting the third feature into a demand vector and the solution text into a semantic vector through a pre-trained language model; concatenating the demand vector and the semantic vector to obtain a concatenated vector, and inputting the concatenated vector into an activation function to obtain the second weight.

[0056] In some embodiments, the solution text for the first problem node k is encoded into a semantic vector s using a pre-trained language model. k Customer characteristics d of the third network node i i Mapped to demand vector q i The formula for calculating the second weight is as follows:

[0057] ;

[0058] Among them, w ik The second weight is denoted by sigmoid, which is the activation function, and ⊕ represents vector concatenation.

[0059] This embodiment calculates a second weight based on business customer characteristics and solution characteristics, which can effectively assess the strength of the correlation between solutions between branch nodes and problem nodes, provide key information for intelligent decision-making in the branch collaboration graph, promote the intelligent upgrade of branch operation and management, and enhance customer service experience and risk control capabilities.

[0060] The third weight can be determined by the matching degree between the business processing type and the problem type. Optionally, in the task processing method for network points provided in the embodiments of this application, determining the third weight of the third type of edge based on the scheme effect evaluation value and the matching degree between the business processing type and the problem type includes: determining the first maintenance node and the second problem node connected by the third type of edge; determining the target business processing type of the first maintenance node and the target scheme effect evaluation value and the target problem type of the second problem node; determining whether the target business processing type and the target problem type match; if the target business processing type and the target problem type match, determining the matching degree as 1; if the target business processing type and the target problem type do not match, determining the matching degree as 0; calculating the product of the target scheme effect evaluation value and the matching degree to obtain the third weight.

[0061] In some embodiments, the third weight can be calculated using the following formula:

[0062] ;

[0063] Among them, w kp As the third weight, k is the second problem node, p is the first maintenance node, the score k is the target solution effect evaluation value, and II() represents the matching function. If the target business processing type contains the processing flow of the target problem type, then the target business processing type is determined to match the target problem type. II(skill) p ∈ type k If the target business processing type does not contain the processing flow for the target problem type, then the target business processing type and the target problem type are determined to be mismatched. (II (Skill)) p ∈ type k The value is 0.

[0064] This embodiment calculates a third weight based on the matching degree between the solution effect evaluation value and the business processing type and the problem type. This can quantify the matching relationship between operation and maintenance experts and problems, making the intelligent decision-making mechanism in the collaborative graph more accurate and improving the emergency response capability and overall operational efficiency of bank branches.

[0065] For different site tasks, different features are extracted from the site collaboration graph to construct target feature vectors. Optionally, in the site task processing method provided in this application embodiment, extracting the target feature vector associated with the site task from the site collaboration graph includes: when the site task is to predict the number of people queuing at the site, the queuing status of each site node in the site collaboration graph is converted into a target feature vector; when the site task is to predict the probability of a problem occurring, the site feature vector of each site node and the problem feature vector of each problem node are extracted from the site collaboration graph, and the site feature vector and the problem feature vector are determined as the target feature vector; when the site task is to recommend a solution, the expert feature vector of each operation and maintenance node and the problem feature vector of each problem node are extracted from the site collaboration graph, and the expert feature vector and the problem feature vector are determined as the target feature vector.

[0066] In some embodiments, if the task at a service point is to predict the number of people queuing at the service point, queuing status data for each service point node is obtained from the service point collaboration graph, including real-time queuing numbers and historical queuing patterns. The queuing status is transformed into a feature vector, where each element represents information such as the average number of people queuing, the trend of queuing number changes, and queuing time over a specific time period. The transformed queuing status feature vector is considered as the target feature vector and is used to predict the number of people queuing in future time periods.

[0067] If the task at a service point is to predict the probability of a problem occurring, the attributes of each service point node are obtained from the service point collaboration graph, including geographical location, customer distribution, and real-time operational status. These attributes are then transformed into feature vectors using encoding algorithms (such as one-hot encoding and numerical encoding) to reflect the service capabilities and potential risks of each service point. Information for each problem node is collected, including problem type, historical solutions, and solution effectiveness ratings. A deep learning model is used to encode the problem description and solution text, generating semantic feature vectors. This helps quantify the complexity of the problem and the relevance of the solution. The service point feature vectors and problem feature vectors are combined to form the target feature vector. This combination can be achieved through concatenation, dot product, or other mathematical operations.

[0068] If the task at the network point is solution recommendation, information about employees or experts is extracted from each operations and maintenance node, including professional skill tags, types of cases handled, and case handling results. Expert attributes are vectorized using encoding or deep learning models to reflect their problem-solving abilities and experience. The expert feature vectors and problem feature vectors are combined to form the target feature vector.

[0069] This embodiment extracts high-quality target feature vectors from the branch collaboration graph for different branch tasks, providing rich and intuitive data input for prediction and recommendation algorithms based on graph neural networks. This ensures that the intelligent system can accurately and efficiently respond to the actual business needs of bank branches, thereby improving overall operational efficiency and service quality.

[0070] Each type of network task is processed by the corresponding task module in the target model, and the corresponding processing result is output. Optionally, in the network task processing method provided in this application embodiment, inputting the target feature vector into the target model to obtain the processing result of the network task includes: when the network task is to predict the number of people queuing at the network, inputting the target feature vector into the target model to obtain the predicted number of people queuing at each network node within the target period, determining the passenger flow allocation scheme based on the predicted number of people queuing at each network node, and determining the passenger flow allocation scheme as the processing result; when the network task is to predict the probability of a problem occurring, inputting the target feature vector into the target model to obtain the predicted probability of each type of problem at each network node, pushing the solution of the corresponding type of problem to the network node when the predicted probability of occurrence is greater than or equal to the probability threshold, and determining the predicted probability of occurrence and the solution as the processing result; when the network task is to recommend a solution, inputting the target feature vector into the target model to obtain the probability that the problem type corresponding to each problem node will be solved by the expert corresponding to each operation and maintenance node, and determining the expert solution probability sequence of each problem node as the processing result, wherein the expert solution probability sequence is sorted from largest to smallest according to the probability of the expert solving the problem type corresponding to the problem node.

[0071] In some embodiments, for the task of predicting queue length at service outlets, the target feature vector (queue state feature vector) transformed from each outlet node is input into a trained prediction model. The model outputs the predicted queue length for each outlet within a target period, which helps outlets prepare for customer flow management in advance. Based on the predicted queue length for each outlet, the imbalance in customer flow distribution between outlets is analyzed. A customer flow allocation plan is generated, which may include measures such as opening more windows, adjusting service staff configuration, and guiding customers to outlets with lower customer flow to optimize the overall queuing situation.

[0072] For the problem occurrence probability prediction task, the combined network node feature vector and problem feature vector are used as the target feature vector and input into the corresponding problem prediction model. The model outputs the predicted occurrence probability of each network node for each type of problem. A probability threshold is set; when the predicted occurrence probability is greater than or equal to this threshold, the system considers that the network node to have a risk of experiencing this type of problem. For high-risk problems, the system automatically pushes preset solutions and coping strategies to the corresponding network node.

[0073] For the solution recommendation task, a target feature vector, combining expert feature vectors and problem feature vectors, is input into the solution recommendation model. The model generates a probability sequence of expert problem-solving outcomes for each operation and maintenance node and for each problem node. For each problem node, the system generates an expert solution probability sequence, where each expert's position in the sequence reflects their probability of solving the problem. The sequence is sorted from high to low according to the experts' probability of solving the problem. The sorted expert solution probability sequence is output as the processing result, recommending the most likely professional to solve the problem for the network node encountering the problem.

[0074] For example, when using a graph convolutional network to handle the heterogeneity of a bank branch collaboration graph, the feature update of node i in the l-th layer is as follows:

[0075] ;

[0076] ;

[0077] Among them, h i （l） W0 is the feature vector of node i in the l-th layer. （l） It is a self-updating weight matrix (used to update the features of the nodes themselves), N i r Under relation r, the set of neighbors of node i, {net2net, net2prob, prob2exp}, represents: node collaboration, problem association, and expert resolution relationship. r （l） It is the weight matrix of relation r, used to transform the features of neighboring nodes.

[0078] The task of predicting queue length at service outlets focuses on estimating customer queue lengths in real time for each outlet. The model extracts feature vectors from each outlet node, performs a non-linear transformation in the first layer, and then directly outputs the predicted queue length in the second layer. This prediction helps outlet managers anticipate customer flow pressure and promptly open more service windows or allocate more manpower when queue lengths may exceed a threshold, effectively improving customer service experience. The loss function used in the training process of the queue length prediction task is as follows:

[0079] ;

[0080] Among them, W state It is the regression layer weight matrix, b state This is the bias term, where N is the sample size, yi is the actual number of people queuing, and y is the bias term. i ^ and y i ^state It predicts the number of people in the queue.

[0081] The problem occurrence probability prediction task is responsible for filtering the most likely types of business problems for each branch office. The model generates branch office feature vectors and problem feature vectors, and by calculating the matching degree between the branch office features and all problem features, it outputs the occurrence probability for each problem type. The system automatically sorts these based on probability values ​​and pushes targeted problem solutions to branches with different characteristics. The loss function during the training process of the problem occurrence probability prediction task is as follows:

[0082] ;

[0083] Among them, W rec It is a bilinear transformation matrix, p ik E represents the probability that node i will have problem k, E is the network node collaboration graph, and k' represents other problem nodes besides problem node k.

[0084] The solution recommendation task addresses the need for rapid response to new problems. The model outputs problem feature vectors and expert feature vectors, calculates the professional fit between the problem and the experts, and generates a ranked list of expert solutions based on probability. The matching process comprehensively considers expert qualifications and current work status. The loss function during the training process of the solution recommendation task is as follows:

[0085] ;

[0086] Where Wexp is the compatibility matrix, o kp This represents the matching probability between the maintenance node p and the problem node k. The joint training of the three tasks uses a weighted loss function, designed as follows:

[0087] ;

[0088] Among them, λ1, λ2, and λ3 are weight parameters that can be customized and modified.

[0089] This embodiment generates accurate predictions or recommendations for different types of branch tasks, thereby providing strong support for dynamic resource scheduling, risk prevention, and efficient problem-solving at the branch level, and further improving the bank's operational efficiency and customer satisfaction.

[0090] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0091] Example 2

[0092] This application also provides a task processing device for service outlets. It should be noted that the task processing device for service outlets in this application can be used to execute the task processing method for service outlets provided in this application. The task processing device for service outlets provided in this application will be described below.

[0093] According to an embodiment of this application, an apparatus for implementing the above-described task processing method for branch offices is also provided. Figure 4 This is a schematic diagram of a task processing device for a network point provided according to an embodiment of this application, such as... Figure 4 As shown, the device includes:

[0094] The first determining unit 401 is used to determine the branch collaboration graph of M branches of a financial institution, where M is a positive integer. The branch collaboration graph is constructed from the relevant data of the M branches. The branch collaboration graph includes multiple types of nodes and edges. The nodes include at least one of the following: branch nodes, problem nodes, and operation and maintenance nodes.

[0095] Extraction unit 402 is used to extract target feature vectors associated with branch tasks from the branch collaboration graph when a branch task is received, wherein the branch task includes at least one of the following: branch queue number prediction, problem occurrence probability prediction, and solution recommendation.

[0096] Input unit 403 is used to input the target feature vector into the target model to obtain the processing result of the network task. The target model is trained by multiple sets of training samples, and each set of training samples includes historical feature vectors and historical processing results.

[0097] The branch task processing device provided in this application embodiment determines the branch collaboration map of M branches of a financial institution through a first determining unit 401, where M is a positive integer. The branch collaboration map is constructed from the relevant data of the M branches and includes various types of nodes and edges. The nodes include at least one of the following: branch nodes, problem nodes, and operation and maintenance nodes. The extraction unit 402 extracts the target feature vector associated with the branch task from the branch collaboration map when a branch task is received, where the branch task includes at least one of the following: branch queue number prediction, problem occurrence probability prediction, and solution recommendation. The input unit 403 inputs the target feature vector into the target model to obtain the processing result of the branch task, where the target model is trained by multiple sets of training samples. Each set of training samples includes historical feature vectors and historical processing results. By integrating multiple data sources of each branch to construct the branch collaboration map, the matching between problem nodes and operation and maintenance nodes can ensure that when a branch encounters a specific problem, it can quickly obtain the most suitable expert or solution to solve the problem. This accelerates the reuse of best service cases, risk management plans, and strategies, improving the overall service level and risk control capabilities of the network. Through the network collaboration map, intelligent matching of resources and capabilities across networks can be achieved based on factors such as customer profile similarity, complementary service capabilities, and the spatiotemporal characteristics of business traffic. For example, geographically adjacent networks can share promotional strategies or allocate human resources when one network experiences peak customer traffic; networks with expertise can provide advice or support in specific business areas to other networks, achieving efficient task processing and synergistic effects between networks. This improves the technical efficiency of network task processing and solves the technical problem of low network task processing efficiency.

[0098] Optionally, in the task processing device for network points provided in this application embodiment, the network point collaboration graph is constructed in the following manner: For each network point, network point attribute data, problem data, and operation and maintenance data are extracted from relevant data; network point nodes are constructed based on network point attribute data, problem nodes are constructed based on problem data, and operation and maintenance nodes are constructed based on operation and maintenance data; first-type edges are connected between all network point nodes; second-type edges are connected between network point nodes and problem nodes based on historical problems of the network point; third-type edges are connected between operation and maintenance nodes and problem nodes based on problems solved by historical business processing cases of operation and maintenance nodes; the weight of each type of edge is determined, and the network point collaboration graph is composed of network point nodes, problem nodes, operation and maintenance nodes, first-type edges, second-type edges, third-type edges, and the weight of each type of edge.

[0099] Optionally, in the task processing device for network points provided in this application embodiment, extracting network point attribute data, problem data, and operation and maintenance data from relevant data includes: determining at least one of the following as network point attribute data: geographical location, business customer group, and queuing status; determining at least one of the following as problem data: problem type, solution, and solution effectiveness evaluation value, where problem types include customer complaint issues, business system issues, and business risk issues; and determining at least one of the following as operation and maintenance data: employee number, business processing type, and historical business processing cases. Determining the weight of each type of edge includes: determining a first weight of a first type of edge based on the geographical distance between network points and the similarity of business customer groups; determining a second weight of a second type of edge based on the business customer group characteristics of network point nodes and the solution characteristics of problem nodes; and determining a third weight of a third type of edge based on the solution effectiveness evaluation value and the matching degree between business processing type and problem type.

[0100] Optionally, in the task processing device for network outlets provided in this application embodiment, determining the first weight of the first type of edge based on the geographical distance between network outlets and the similarity of business customer groups includes: determining the first network outlet node and the second network outlet node connected by the first type of edge; determining the first business customer group and the first geographical location information of the first network outlet node; determining the second business customer group and the second geographical location information of the second network outlet node; extracting the first feature of the first business customer group and the second feature of the second business customer group, wherein the first feature or the second feature includes at least one of the following: customer group type, amount of funds, and quantity distribution of customer group type, and the customer group type includes at least one of the following: individuals, enterprises, and related departments; calculating the geographical distance based on the first geographical location information and the second geographical location information; calculating the cosine similarity between the first feature and the second feature; and calculating the product of the cosine similarity and the geographical distance to obtain the first weight.

[0101] Optionally, in the task processing device for network outlets provided in this application embodiment, determining the second weight of the second type of edge based on the business customer group characteristics of the network outlet node and the solution characteristics of the problem node includes: determining the third network outlet node and the first problem node connected by the second type of edge; determining the third business customer group of the third network outlet node and the first solution of the first problem node; extracting the third feature of the third business customer group and the solution text of the first solution; converting the third feature into a demand vector and the solution text into a semantic vector through a pre-trained language model; concatenating the demand vector and the semantic vector to obtain a concatenated vector; and inputting the concatenated vector into an activation function to obtain the second weight.

[0102] Optionally, in the task processing device for network points provided in this application embodiment, determining the third weight of the third type of edge based on the scheme effect evaluation value and the matching degree between the business processing type and the problem type includes: determining the first maintenance node and the second problem node connected by the third type of edge; determining the target business processing type of the first maintenance node and the target scheme effect evaluation value and the target problem type of the second problem node; determining whether the target business processing type and the target problem type match; if the target business processing type and the target problem type match, determining the matching degree as 1; if the target business processing type and the target problem type do not match, determining the matching degree as 0; calculating the product of the target scheme effect evaluation value and the matching degree to obtain the third weight.

[0103] Optionally, in the task processing device for network points provided in this application embodiment, the extraction unit 402 includes: a conversion module, used to convert the queuing status of each network point node in the network point collaboration graph into a target feature vector when the network point task is a prediction of the number of people queuing at the network point; an extraction module, used to extract the network point feature vector of each network point node and the problem feature vector of each problem node from the network point collaboration graph when the network point task is a prediction of the probability of problem occurrence, and determine the network point feature vector and the problem feature vector as the target feature vector; and a first determination module, used to extract the expert feature vector of each operation and maintenance node and the problem feature vector of each problem node from the network point collaboration graph when the network point task is a solution recommendation, and determine the expert feature vector and the problem feature vector as the target feature vector.

[0104] Optionally, in the task processing device for network points provided in this application embodiment, the input unit 403 includes: a first input module, used to input the target feature vector into the target model when the network point task is to predict the number of people queuing at the network point, to obtain the predicted number of people queuing at each network point node within the target period, to determine the passenger flow allocation scheme based on the predicted number of people queuing at each network point node, and to determine the passenger flow allocation scheme as the processing result; a second input module, used to input the target feature vector into the target model when the network point task is to predict the probability of a problem occurring, to obtain the predicted probability of each network point node for each type of problem, to push the solution of the corresponding type of problem to the network point node when the predicted probability of occurrence is greater than or equal to the probability threshold, and to determine the predicted probability of occurrence and the solution as the processing result; a third input module, used to input the target feature vector into the target model when the network point task is to recommend a solution, to obtain the probability that the problem type corresponding to each problem node will be solved by the expert corresponding to each operation and maintenance node, and to determine the expert solution probability sequence of each problem node as the processing result, wherein the expert solution probability sequence is sorted from largest to smallest according to the probability of the expert solving the problem type corresponding to the problem node.

[0105] It should be noted that the first determining unit 401, the extraction unit 402, and the input unit 403 mentioned above correspond to steps S201 to S203 in Embodiment 1. The three units and the corresponding steps implement the same instances and application scenarios, but are not limited to the content disclosed in Embodiment 1. It should be noted that the above modules or units can be hardware or software components stored in memory (e.g., memory 104) and processed by one or more processors (e.g., processors 102a, 102b, ..., 102n). The above modules or units can also be part of a device and run in the computer terminal 10 provided in Embodiment 1.

[0106] Example 3

[0107] Embodiments of this application may provide an electronic device. Figure 5 This is a structural block diagram of an electronic device according to an embodiment of this application. Figure 5 As shown, the electronic device may include: one or more ( Figure 5 (Only one is shown) processor 502, memory 504, memory controller, and peripheral interface, wherein the peripheral interface is connected to the radio frequency module, audio module and display.

[0108] The memory can be used to store software programs and modules, such as the program instructions / modules corresponding to the methods and apparatus in the embodiments of this application. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, thereby implementing the above-described methods. The memory may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory may further include memory remotely located relative to the processor, and these remote memories can be connected to the terminal via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0109] The processor can access information and applications stored in memory via a transmission device to perform the following steps: Determine a branch collaboration graph of M branches of a financial institution, where M is a positive integer. The branch collaboration graph is constructed from relevant data of the M branches and includes various types of nodes and edges. Nodes include at least one of the following: branch nodes, problem nodes, and maintenance nodes. Upon receiving a branch task, extract a target feature vector associated with the branch task from the branch collaboration graph. The branch task includes at least one of the following: branch queue length prediction, problem occurrence probability prediction, and solution recommendation. Input the target feature vector into a target model to obtain the processing result of the branch task. The target model is trained using multiple sets of training samples, each set of training samples including historical feature vectors and historical processing results.

[0110] The processor can also call the information and application programs stored in the memory through the transmission device to perform the following steps: For each branch, extract the branch attribute data, problem data, and operation and maintenance data from the relevant data; construct branch nodes based on the branch attribute data, construct problem nodes based on the problem data, and construct operation and maintenance nodes based on the operation and maintenance data; connect all branch nodes with first-type edges, connect branch nodes and problem nodes with second-type edges based on the branch's historical problems, and connect operation and maintenance nodes and problem nodes with third-type edges based on the problems solved by the historical business processing cases of operation and maintenance nodes; determine the weight of each type of edge, and construct a branch collaboration graph from branch nodes, problem nodes, operation and maintenance nodes, first-type edges, second-type edges, third-type edges, and the weight of each type of edge.

[0111] The processor can also invoke information and applications stored in the memory via a transmission device to perform the following steps: Identify at least one of the following as branch attribute data: geographical location, customer base, and queuing status; identify at least one of the following as problem data: problem type, solution, and solution effectiveness evaluation value, where problem types include customer complaints, business system problems, and business risk problems; identify at least one of the following as operation and maintenance data: employee number, business processing type, and historical business processing cases; determine the weight of each type of edge, including: determining the first weight of the first type of edge based on the geographical distance between branches and the similarity of customer base; determining the second weight of the second type of edge based on the customer base characteristics of the branch node and the solution characteristics of the problem node; and determining the third weight of the third type of edge based on the solution effectiveness evaluation value and the matching degree between the business processing type and the problem type.

[0112] The processor can also invoke information and applications stored in the memory via a transmission device to perform the following steps: determining a first network node and a second network node connected by a first type of edge; determining a first business customer group and a first geographical location information for the first network node; determining a second business customer group and a second geographical location information for the second network node; extracting a first feature of the first business customer group and a second feature of the second business customer group, wherein the first feature or the second feature includes at least one of the following: customer group type, amount of funds, and quantity distribution of customer group type, and the customer group type includes at least one of the following: individuals, enterprises, and related departments; calculating the geographical distance based on the first geographical location information and the second geographical location information; calculating the cosine similarity between the first feature and the second feature; calculating the product of the cosine similarity and the geographical distance to obtain a first weight.

[0113] The processor can also call the information and application stored in the memory through the transmission device to perform the following steps: determine the third network node and the first problem node of the second type of edge connection, determine the third business customer group of the third network node and the first solution of the first problem node; extract the third feature of the third business customer group and the solution text of the first solution; convert the third feature into a demand vector and the solution text into a semantic vector through a pre-trained language model; concatenate the demand vector and the semantic vector to obtain a concatenated vector, and input the concatenated vector into an activation function to obtain the second weight.

[0114] The processor can also call the information and application stored in the memory through the transmission device to perform the following steps: determine the first maintenance node and the second problem node of the third type of edge connection; determine the target business processing type of the first maintenance node and the target solution effect evaluation value and target problem type of the second problem node; determine whether the target business processing type and the target problem type match; if the target business processing type and the target problem type match, determine the matching degree as 1; if the target business processing type and the target problem type do not match, determine the matching degree as 0; calculate the product of the target solution effect evaluation value and the matching degree to obtain the third weight.

[0115] The processor can also invoke information and applications stored in the memory via the transmission device to perform the following steps: when the network task is to predict the number of people queuing at the network, the queuing status of each network node in the network collaboration graph is transformed into a target feature vector; when the network task is to predict the probability of a problem occurring, the network feature vector of each network node and the problem feature vector of each problem node are extracted from the network collaboration graph, and the network feature vector and the problem feature vector are determined as the target feature vector; when the network task is to recommend a solution, the expert feature vector of each operation and maintenance node and the problem feature vector of each problem node are extracted from the network collaboration graph, and the expert feature vector and the problem feature vector are determined as the target feature vector.

[0116] The processor can also access information and applications stored in the memory via a transmission device to perform the following steps: When the task at a service point is to predict the number of people queuing at the service point, the target feature vector is input into the target model to obtain the predicted number of people queuing at each service point node within the target period. Based on the predicted number of people queuing at each service point node, a passenger flow allocation scheme is determined, and the passenger flow allocation scheme is determined as the processing result. When the task at a service point is to predict the probability of a problem occurring, the target feature vector is input into the target model to obtain the predicted probability of each type of problem at each service point node. If the predicted probability of occurrence is greater than or equal to a probability threshold, a solution for the corresponding type of problem is pushed to the service point node, and the predicted probability of occurrence and the solution are determined as the processing result. When the task at a service point is to recommend a solution, the target feature vector is input into the target model to obtain the probability that the problem type corresponding to each problem node will be solved by the expert corresponding to each operation and maintenance node. The expert solution probability sequence for each problem node is determined as the processing result, wherein the expert solution probability sequence is sorted from largest to smallest according to the probability of the expert solving the problem type corresponding to the problem node.

[0117] This application provides a method for determining a branch collaboration graph for M branches of a financial institution, where M is a positive integer. The branch collaboration graph is constructed from relevant data of the M branches and includes various types of nodes and edges. Nodes include at least one of the following: branch nodes, problem nodes, and maintenance nodes. Upon receiving a branch task, a target feature vector associated with the branch task is extracted from the branch collaboration graph. The branch task includes at least one of the following: branch queue length prediction, problem occurrence probability prediction, and solution recommendation. The target feature vector is input into a target model to obtain the processing result of the branch task. The target model is trained using multiple sets of training samples, each set of training samples including historical feature vectors and historical processing result schemes. By integrating multiple data sources from various branches to construct the branch collaboration graph, the matching between problem nodes and maintenance nodes ensures that when a branch encounters a specific problem, it can quickly obtain the most suitable expert or solution to solve the problem. This accelerates the reuse of excellent service cases, risk management plans, and strategies, and improves the overall service level and risk control capabilities of the branches. Through a network collaboration map, intelligent matching of resources and capabilities across networks can be achieved based on factors such as customer profile similarity, complementary service capabilities, and the spatiotemporal characteristics of business traffic. For example, geographically adjacent networks can share promotional strategies or allocate human resources when one network experiences peak customer traffic; networks with expertise can provide advice or support in specific business areas to other networks, achieving efficient task processing and synergistic effects between networks. This improves the technical efficiency of network task processing and solves the technical problem of low network task processing efficiency.

[0118] Those skilled in the art will understand that Figure 5 The structure shown is for illustrative purposes only. Electronic devices can also be smartphones, tablets, handheld computers, mobile internet devices (MIDs), PADs, and other terminal devices. Figure 5 This does not limit the structure of the aforementioned electronic device. For example, electronic devices may also include components that are more... Figure 5 The more or fewer components shown (such as network interfaces, display devices, etc.), or having the same Figure 5 The different configurations shown.

[0119] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing the hardware related to the terminal device. The program can be stored in a computer-readable storage medium, which may include: flash drive, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.

[0120] Example 4

[0121] Embodiments of this application also provide a storage medium. Optionally, in this embodiment, the storage medium can be used to store the program code executed by the task processing method for network points provided in Embodiment 1.

[0122] Optionally, in this embodiment, the storage medium may be located in any computer terminal in a group of computer terminals in a computer network, or in any mobile terminal in a group of mobile terminals.

[0123] This application also provides a computer program product that, when executed on a data processing device, is suitable for performing the task processing method steps of a branch office.

[0124] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0125] In the above embodiments of this application, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0126] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.

[0127] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0128] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0129] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard drive, magnetic disk, or optical disk.

[0130] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A task processing method for a service point, characterized in that, include: Determine a branch collaboration graph of M branches of a financial institution, where M is a positive integer. The branch collaboration graph is constructed from the relevant data of the M branches. The branch collaboration graph includes multiple types of nodes and edges. The nodes include at least one of the following: branch nodes, problem nodes, and operation and maintenance nodes. Upon receiving a task from a branch, a target feature vector associated with the task is extracted from the branch collaboration graph, wherein the branch task includes at least one of the following: branch queue length prediction, problem occurrence probability prediction, and solution recommendation. The target feature vector is input into the target model to obtain the processing result of the network task. The target model is trained by multiple sets of training samples, and each set of training samples includes historical feature vectors and historical processing results.

2. The method according to claim 1, characterized in that, The network point collaboration map is constructed in the following way: For each branch, extract the branch attribute data, problem data, and operation and maintenance data from the relevant data, construct the branch node based on the branch attribute data, construct the problem node based on the problem data, and construct the operation and maintenance node based on the operation and maintenance data; Connect all network nodes with first-type edges, connect the network node and the problem node with second-type edges based on the network node's historical problems, and connect the operation and maintenance node and the problem node with third-type edges based on the problems solved by the operation and maintenance node's historical business processing cases. The weight of each type of edge is determined, and the network node collaboration graph is constructed from the network node, the problem node, the operation and maintenance node, the first type of edge, the second type of edge, the third type of edge, and the weight of each type of edge.

3. The method according to claim 2, characterized in that, Extracting the network attribute data, problem data, and operation and maintenance data of the network points from the relevant data includes: The following shall be identified as the network attribute data: geographical location, customer group and queuing status; the following shall be identified as the problem data: problem type, solution and solution effectiveness evaluation value, wherein the problem type includes customer complaint problem, business system problem and business risk problem; and the following shall be identified as the operation and maintenance data: employee number, business processing type and historical business processing case. Determining the weight of each type of edge includes: The first weight of the first type of edge is determined based on the geographical distance between outlets and the similarity of business customer groups. The second weight of the second type of edge is determined based on the business customer group characteristics of the outlet node and the solution characteristics of the problem node. The third weight of the third type of edge is determined based on the solution effect evaluation value and the matching degree between the business processing type and the problem type.

4. The method according to claim 3, characterized in that, The first weight of the first type of edge is determined based on the geographical distance between outlets and the similarity of customer groups, including: Determine the first and second network nodes connected by the first type of edge, determine the first business customer group and the first geographical location information of the first network node, and determine the second business customer group and the second geographical location information of the second network node; Extract a first feature of the first business customer group and a second feature of the second business customer group, wherein the first feature or the second feature includes at least one of the following: customer group type, amount of funds and quantity distribution of customer group type, wherein the customer group type includes at least one of the following: individuals, enterprises and related departments; Calculate the geographical distance based on the first geographical location information and the second geographical location information, and calculate the cosine similarity between the first feature and the second feature; The first weight is obtained by calculating the product of the cosine similarity and the geographical distance.

5. The method according to claim 3, characterized in that, Determining the second weight of the second type of edge based on the business customer characteristics of the network node and the solution characteristics of the problem node includes: Identify the third network node and the first problem node connected by the second type of edge, and determine the third business customer group of the third network node and the first solution of the first problem node; Extract the third characteristic of the third business customer group and the solution text of the first solution; The third feature is transformed into a demand vector using a pre-trained language model, and the solution text is transformed into a semantic vector. The demand vector and the semantic vector are concatenated to obtain a concatenated vector. The concatenated vector is then input into an activation function to obtain the second weight.

6. The method according to claim 3, characterized in that, The third weight of the third type of edge is determined based on the evaluation value of the solution effect and the matching degree between the business processing type and the problem type, including: Determine the first maintenance node and the second problem node of the third type of edge connection, and determine the target business processing type of the first maintenance node and the target solution effect evaluation value and target problem type of the second problem node; Determine whether the target business processing type matches the target problem type; If the target business processing type matches the target problem type, the matching degree is determined to be 1; if the target business processing type does not match the target problem type, the matching degree is determined to be 0. The third weight is obtained by calculating the product of the target solution's effectiveness evaluation value and the matching degree.

7. The method according to claim 1, characterized in that, Extracting target feature vectors associated with the network task from the network collaborative graph includes: When the task at the network point is to predict the number of people queuing at the network point, the queuing status of each network point node in the network point collaboration graph is transformed into the target feature vector. When the task at the network point is to predict the probability of the problem occurring, the network point feature vector of each network point node and the problem feature vector of each problem node are extracted from the network point collaboration graph, and the network point feature vector and the problem feature vector are determined as the target feature vector. When the task at the network point is the solution recommended, the expert feature vector of each operation and maintenance node and the problem feature vector of each problem node are extracted from the network point collaboration graph, and the expert feature vector and the problem feature vector are determined as the target feature vector.

8. The method according to claim 1, characterized in that, The target feature vector is input into the target model to obtain the processing result of the network task, including: When the task at the network point is to predict the number of people queuing at the network point, the target feature vector is input into the target model to obtain the predicted number of people queuing at each network point node within the target period. Based on the predicted number of people queuing at each network point node, a passenger flow allocation scheme is determined, and the passenger flow allocation scheme is determined as the processing result. When the task at the network point is to predict the probability of the problem occurring, the target feature vector is input into the target model to obtain the predicted probability of each type of problem for each network point node. If the predicted probability of occurrence is greater than or equal to the probability threshold, a solution for the corresponding type of problem is pushed to the network point node, and the predicted probability of occurrence and the solution are determined as the processing result. When the task at the network point is the solution recommendation, the target feature vector is input into the target model to obtain the probability that the problem type corresponding to each problem node will be solved by the expert corresponding to each operation and maintenance node. The expert solution probability sequence of each problem node is determined as the processing result, wherein the expert solution probability sequence is sorted from largest to smallest according to the probability of the expert solving the problem type corresponding to the problem node.

9. A task processing device for a network of service points, characterized in that, include: The first determining unit is used to determine the branch collaboration graph of M branches of a financial institution, where M is a positive integer. The branch collaboration graph is constructed from the relevant data of the M branches. The branch collaboration graph includes multiple types of nodes and edges. The nodes include at least one of the following: branch nodes, problem nodes, and operation and maintenance nodes. An extraction unit is configured to extract a target feature vector associated with a network task from the network collaboration graph when a network task is received, wherein the network task includes at least one of the following: network queue length prediction, problem occurrence probability prediction, and solution recommendation. The input unit is used to input the target feature vector into the target model to obtain the processing result of the network task. The target model is trained by multiple sets of training samples, and each set of training samples includes historical feature vectors and historical processing results.

10. A computer program product comprising computer instructions, characterized in that, When the computer instructions are executed by the processor, they implement the steps of the task processing method for the network site as described in any one of claims 1 to 8.

Images