A bank intelligent authorization method, system, device and medium based on a shared operating system

Abstract

The present invention discloses a bank intelligent authorization method, system, device and medium based on a shared operating system. The method comprises: obtaining business data of pending business based on a business authorization request sent by a business counter, performing risk level assessment and risk profile assessment on the pending business based on the business data, matching the authorization review strategy corresponding to the pending business according to the risk profile and business risk level, performing authorization review on the pending business according to the authorization review strategy and business data, and obtaining an authorization review result; performing quality inspection on the authorization review result based on an abnormal time sequence behavior analysis result; correcting the authorization review result based on the quality inspection result, obtaining a target authorization result, and feeding back the target authorization result to the business counter, thereby effectively improving the efficiency and accuracy of business authorization, accurately identifying potential risks through risk level and risk profile assessment, and reducing the risk of mis-authorization through quality inspection through abnormal time sequence behavior analysis.

Description

Technical Field

[0001] The present invention relates to the field of computers, and in particular to a bank intelligent authorization method, system, device, medium and program product based on a shared operating system. Background Art

[0002] With the advancement of science and technology, banking authorization review is gradually shifting from manual review to intelligent centralized authorization review. Relying on artificial intelligence, optical character recognition (OCR), facial recognition, and big data technologies, intelligent centralized authorization review replaces traditional manual authorization through a remote centralized model, automating business processes. Combining intelligent risk control models with multimodal algorithms, the system can automatically compare transaction information, documents, and biometrics, enabling automated authorization of services. However, current intelligent centralized authorization review systems suffer from low accuracy, data silos, and a single risk analysis dimension, making it impossible to accurately identify potential risks and anomalies within the audited services. Furthermore, the quality control of audit results still relies on manual review, which is subject to the subjective nature of manual review, resulting in significant limitations in authorization quality control. Summary of the Invention

[0003] The main purpose of the present invention is to provide a bank intelligent authorization method, system, equipment, medium and program product based on a shared operating system, aiming to solve the technical problems of the current intelligent centralized authorization review in the existing technology with low accuracy, single risk analysis dimension, inability to accurately identify potential risks and anomalies in the review business, and limitations in authorization quality inspection.

[0004] To achieve the above objectives, the present invention provides a bank intelligent authorization method based on a shared operating system. The method is applied to the shared operating system and includes the following steps:

[0005] In response to a service authorization request sent by a service counter, obtaining service data of the service to be processed based on the service authorization request, the service data including the service type;

[0006] Performing a risk level assessment on the pending business based on the business type to determine the business risk level of the pending business;

[0007] Inputting the branch data of the business branch to which the business counter belongs, the customer data of the customer handling the pending business, and the teller data of the teller handling the pending business into a risk profile assessment model to generate a risk profile of the pending business, wherein the risk profile assessment model is constructed based on multidimensional feature data of multiple external audit nodes, the multidimensional feature data including customer feature data, branch feature data, and teller feature data;

[0008] Matching the authorization review strategy corresponding to the pending business according to the risk profile and the business risk level, the authorization review strategy includes a high-risk authorization review strategy and a medium- and low-risk authorization review strategy. The high-risk business processing authorization review strategy includes a target individual authorization review strategy and a target business authorization review strategy. The target individuals include agent customers and customers under a preset age threshold. The target businesses include cash flow business, loan business, and corporate business. The medium- and low-risk authorization review strategy includes determining whether to directly authorize based on the daily authorization ratio of the business counter;

[0009] Performing authorization review on the pending business according to the authorization review policy and the business data, and obtaining an authorization review result;

[0010] Performing abnormal time series behavior analysis based on the historical business data of the business outlets and the historical business data of the customers, and performing quality inspection on the authorization review results based on the abnormal time series behavior analysis results;

[0011] The authorization review result is corrected based on the quality inspection result to obtain the target authorization result, and the target authorization result is fed back to the business counter.

[0012] Optionally, before inputting the branch data of the business branch to which the business counter belongs, the customer data of the customer handling the pending business, and the teller data of the teller handling the pending business into the risk profile assessment model to generate the risk profile of the pending business, the method further includes:

[0013] Obtaining business logs, including branch business logs, customer business logs, and teller business logs;

[0014] Generate an authorization behavior time series based on the business log, and train a pre-built original portrait model based on the authorization behavior time series;

[0015] Obtain initial model parameters based on the trained original portrait model;

[0016] Performing hash encryption on the initial model parameters to generate encrypted model parameters, and sending the encrypted model parameters to each external audit node;

[0017] In response to the external node model parameters sent by each external audit node, the external node model parameters are aggregated to obtain global model parameters, wherein the external node model parameters are obtained by each external audit node decrypting the encrypted model parameters and training the local portrait model based on the initial model parameters obtained after decryption;

[0018] The trained original portrait model is trained based on the global model parameters to obtain a risk portrait assessment model.

[0019] Optionally, generating an authorization behavior time series based on the business log and training a pre-built original portrait model according to the authorization behavior time series includes:

[0020] Extracting log semantic information and timestamp information from the business log;

[0021] Acquire historical event information based on the log semantic information, the historical event information including branch event information, teller event information, and customer event information;

[0022] Performing event behavior analysis based on the historical event information to obtain event behavior characteristics of each historical event;

[0023] Constructing an event causal graph based on the event behavior characteristics;

[0024] Generate an authorization behavior time series according to the timestamp information, and construct an event graph based on the authorization behavior time series;

[0025] Establishing a path index based on the causal path in the causal graph and the temporal path in the event graph;

[0026] Extracting event nodes from the event graph and extracting causal pair nodes from the causal graph;

[0027] Merging the event graph with the causal graph based on the path index, the event node, and the causal pair node to generate a target hypergraph, wherein the target hypergraph includes a causal hyperedge and a temporal hyperedge;

[0028] Aggregating node information of neighboring nodes of each hyperedge in the target hypergraph to obtain a hypergraph embedding feature vector;

[0029] Performing feature fusion on the original features of each node in the target hypergraph and the hypergraph embedded feature vector to obtain a comprehensive feature vector;

[0030] A data set is generated based on the comprehensive feature vector, and a pre-built original portrait model is trained based on the data set.

[0031] Optionally, aggregating the external node model parameters to obtain global model parameters includes:

[0032] Determine the institutional type of each external audit node and obtain historical audit behavior information of each external audit node;

[0033] Performing feature analysis on the historical audit behavior information to obtain multi-dimensional behavior feature information, wherein the multi-dimensional behavior feature information includes response delay features, parameter noise features, historical abnormal behavior features, and network stability features of each external audit node;

[0034] Normalize the multi-dimensional behavior feature information, and calculate the direct trust of each external audit node based on the normalized multi-dimensional behavior feature information:

[0035]

[0036] in, Indicates direct trust. Represent the weight coefficient of each behavioral feature, represents the normalized response delay characteristics, represents the normalized parameter noise characteristic, Represents the normalized historical abnormal behavior characteristics, represents the normalized network stability characteristics;

[0037] Obtain the historical credibility of each external audit node and the evaluation information between each external audit node, and calculate the recommended trustworthiness of each external audit node based on the evaluation information:

[0038]

[0039] in, Represents an external audit node The recommendation trust Represents external audit nodes The set of other external audit nodes that have interacted, Represents an external audit node Historical trust, Represents an external audit node External audit nodes Evaluation information;

[0040] Aggregate the direct trust and the recommended trust to obtain the comprehensive trust of each external audit node:

[0041]

[0042] in, represents the fusion weight coefficient, Indicates comprehensive trust;

[0043] The external node model parameters are aggregated based on the comprehensive trust and the organization type to obtain global model parameters.

[0044] Optionally, aggregating the external node model parameters based on the comprehensive trust and the organization type to obtain global model parameters includes:

[0045] Determine the behavior time interval of historical audits of each external audit node based on the historical audit behavior information;

[0046] The time-sensitive trust of each external audit node is calculated based on the behavior time interval and the comprehensive trust:

[0047]

[0048]

[0049] in, Indicates time-sensitive trust, represents the time decay factor, Indicates the weight of the current review behavior. Indicates historical audit trust. represents the decay rate, Indicates the time interval between behaviors;

[0050] Classifying the external audit nodes according to the organization type and the time-sensitive trust to obtain a plurality of local node sets, wherein the local node sets include one or more external audit nodes;

[0051] Performing local aggregation on the external node model based on the comprehensive trust of each node in each local node set to obtain local model parameters of each local node set;

[0052] Obtaining the historical contribution of each node in the local node set described in the historical model parameters;

[0053] Determining the collective contribution of each local node set based on the historical contribution, and determining the collective trust of each local node set based on the comprehensive trust;

[0054] The local model parameters are aggregated based on the collective contribution and the comprehensive trust to obtain global model parameters:

[0055]

[0056] in, represents the global model parameters, represents the local model parameters, represents the balance coefficient, which is used to balance the comprehensive trust and the collective contribution. represents the collective contribution, Represents the collective confidence.

[0057] Optionally, performing authorization review on the pending business according to the authorization review policy and the business data to obtain an authorization review result includes:

[0058] If the authorization review policy is a target individual authorization review policy, then the individual type of the current customer of the pending business is obtained according to the business data;

[0059] If the individual type of the current customer is an agent customer, then obtaining the original image of the power of attorney for the business to be handled;

[0060] Perform grayscale processing on each pixel in the original image to obtain a grayscale image of the power of attorney:

[0061]

[0062] in, Represents the grayscale image of the power of attorney at coordinates The pixel value at Represent the weight coefficients of the red, green, and blue channels respectively, Represent the coordinates in the original image The pixel values ​​at are red, green and blue channels;

[0063] Obtaining the grayscale value range of the grayscale image of the power of attorney, and traversing each pixel point of the grayscale image of the power of attorney;

[0064] Performing a grayscale distribution analysis on each pixel in the grayscale image of the power of attorney according to the grayscale value range and the traversal result, and setting an initial grayscale threshold based on the grayscale distribution analysis result;

[0065] Based on the initial grayscale threshold and the grayscale distribution analysis result, the maximum inter-class variance calculation is performed on the grayscale image of the power of attorney to determine the target grayscale threshold:

[0066]

[0067] in, represents the target grayscale threshold, represents the initial grayscale threshold, Represents the initial grayscale threshold The percentage of pixels on the left side, Represents the initial grayscale threshold The pixel ratio on the right side, represents the average gray value of the pixels on the left, Represents the average gray value of the pixels on the right;

[0068] dividing the power of attorney grayscale image into a foreground layer and a background layer according to the target grayscale threshold;

[0069] Dividing the foreground layer into a handwritten font area and a printed font area;

[0070] Acquire a handwritten structural element of the handwritten font area according to the structural information of the handwritten font area, and acquire a printed structural element of the printed font area according to the structural information of the printed font area;

[0071] performing morphological opening operations on the handwritten font area and the printed font area based on the handwritten structure element and the printed structure element, respectively, to obtain a target handwritten image and a target printed image, wherein the morphological opening operations include an erosion operation and a dilation operation;

[0072] Performing content detection on the target handwritten image and the target printed image to obtain content information of the power of attorney for the pending business;

[0073] Based on the content information of the power of attorney, the pending business is authorized to be reviewed and the authorization review result is obtained.

[0074] Optionally, dividing the foreground layer into a handwritten font area and a printed font area includes:

[0075] Calculate the attribution cost of each pixel in the foreground layer, and divide the foreground layer into multiple superpixel blocks based on the attribution cost:

[0076]

[0077] in, Representing coordinates The attributed cost of the pixel at , Representing coordinates The gradient value of the pixel at represents the average gradient value of the superpixel cluster, represents the global gradient standard deviation of the foreground layer, Representing coordinates The spatial distance between the pixel at and the center of the superpixel cluster, represents the size parameter of the superpixel, Represent the balance weight coefficients of gradient and distance respectively;

[0078] Perform character target detection on each superpixel block, and annotate the border of each superpixel block based on the character target detection result to generate multiple character borders;

[0079] Performing feature extraction on each character frame and the characters within the character frame to obtain a frame feature vector and a character structure feature vector of each character frame, and determining the similarity of each character frame based on the frame feature vector and the character structure feature vector;

[0080] The information entropy of each superpixel block is calculated based on the similarity:

[0081]

[0082] in, Represents a superpixel block The information entropy of Indicates character bounding boxes whose similarity is lower than the preset similarity threshold In superpixel blocks The probability of occurrence in

[0083] The superpixel block is classified into a handwritten font area and a printed font area based on the information entropy, the information entropy of the handwritten font area is greater than a preset entropy threshold, and the information entropy of the printed font area is not greater than the preset entropy threshold.

[0084] In addition, to achieve the above-mentioned purpose, the present invention further proposes a shared operation system, which includes:

[0085] An authorization request response module is used to respond to a service authorization request sent by a service counter and obtain service data of the service to be processed based on the service authorization request, wherein the service data includes a service type;

[0086] A risk level assessment module, configured to assess the risk level of the pending business based on the business type and determine the business risk level of the pending business;

[0087] a risk profile generation module, configured to input the branch data of the business branch to which the business counter belongs, the customer data of the customer handling the pending business, and the teller data of the teller handling the pending business into a risk profile assessment model to generate a risk profile of the pending business, wherein the risk profile assessment model is constructed based on multidimensional feature data of multiple external audit nodes, the multidimensional feature data including customer feature data, branch feature data, and teller feature data;

[0088] An audit policy matching module is configured to match the authorization audit policy corresponding to the pending business based on the risk profile and the business risk level. The authorization audit policy includes a high-risk authorization audit policy and a medium- and low-risk authorization audit policy. The high-risk business processing authorization audit policy includes a target individual authorization audit policy and a target business authorization audit policy. The target individuals include agent clients and clients below a preset age threshold. The target businesses include cash flow business, loan business, and corporate business. The medium- and low-risk authorization audit policy includes determining whether to directly authorize based on the daily authorization ratio of the business counter.

[0089] An authorization review module is used to perform authorization review on the pending business according to the authorization review policy and the business data, and obtain an authorization review result;

[0090] An authorization quality inspection module is used to perform abnormal time series behavior analysis based on the historical business data of the business outlets and the historical business data of the customers, and to perform quality inspection on the authorization review results based on the abnormal time series behavior analysis results;

[0091] The authorization feedback module is used to correct the authorization review result based on the quality inspection result, obtain the target authorization result, and feed back the target authorization result to the business counter.

[0092] In addition, to achieve the above-mentioned purpose, the present application also proposes a bank intelligent authorization device based on a shared operating system, the device comprising: a memory, a processor, and a computer program stored on the memory and executable on the processor, the computer program being configured to implement the steps of the bank intelligent authorization method based on the shared operating system as described above.

[0093] In addition, to achieve the above-mentioned purpose, the present application also proposes a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the bank intelligent authorization method based on the shared operating system as described above are implemented.

[0094] In addition, to achieve the above-mentioned purpose, the present application also provides a computer program product, which includes a computer program, and when the computer program is executed by a processor, it implements the steps of the bank intelligent authorization method based on the shared operating system as described above.

[0095] The present invention obtains business data of pending business based on the business authorization request in response to the business authorization request sent by the business counter, wherein the business data includes the business type, and the risk level of the pending business is evaluated based on the business type to determine the business risk level of the pending business. The branch data of the business branch to which the business counter belongs, the customer data of the customer who handles the pending business, and the teller data of the teller who handles the pending business are input into the risk profile evaluation model to generate a risk profile of the pending business. The risk profile evaluation model is constructed based on multidimensional feature data of multiple external audit nodes, and the multidimensional feature data packet Including customer feature data, branch feature data and teller feature data, according to the risk profile and the business risk level, matching the authorization review strategy corresponding to the pending business, the authorization review strategy includes high-risk authorization review strategy and medium-low risk authorization review strategy, the medium-low risk authorization review strategy includes judging whether to directly authorize based on the authorization ratio of the business counter on the same day, the high-risk business handling authorization review strategy includes target individual authorization review strategy and target business authorization review strategy, the target individual includes agent customers and customers whose age is below the preset age threshold, the target business includes capital flow business, loan business and corporate business, according to the The authorization review strategy and the business data are used to conduct authorization review on the business to be processed, and an authorization review result is obtained. An abnormal time series behavior analysis is conducted based on the historical business data of the business outlet and the historical business data of the processing customer, and a quality inspection is conducted on the authorization review result based on the abnormal time series behavior analysis result. The authorization review result is corrected based on the quality inspection result to obtain a target authorization result, and the target authorization result is fed back to the business counter. Since the present invention conducts risk profiling assessment on the business from multiple dimensions, it effectively avoids the data island problem and overcomes the limitation problem of risk assessment of single-dimensional data. Dynamic assessment of risk levels and risk profiles accurately identifies potential risks and anomalies in the business, accurately matches the corresponding authorization review strategy for the pending business, and classifies business reviews into high-risk authorization reviews and medium- and low-risk authorization reviews, thereby strengthening the strictness of the review of high-risk businesses. It determines whether to directly authorize based on the authorization ratio of the day, thereby reducing redundant review links for medium- and low-risk businesses, thereby greatly improving the efficiency of authorization review. By analyzing historical business data for abnormal time series behavior analysis, accurate quality inspection of authorization results is achieved, and the degree of deviation between the customer's current behavior and historical behavior quality inspection is accurately identified. The risk of mis-authorization is effectively reduced through quality inspection and correction mechanisms. BRIEF DESCRIPTION OF THE DRAWINGS

[0096] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0097] Figure 1 This is a schematic diagram of the structure of a bank intelligent authorization device based on a shared operating system in a hardware operating environment according to an embodiment of the present invention;

[0098] Figure 2 A flowchart of a first embodiment of a bank intelligent authorization method based on a shared operating system according to the present invention;

[0099] Figure 3 A flowchart of a second embodiment of the bank intelligent authorization method based on a shared operating system according to the present invention;

[0100] Figure 4 A flowchart of a third embodiment of the bank intelligent authorization method based on a shared operating system according to the present invention;

[0101] Figure 5 This is a structural block diagram of the first embodiment of the shared operating system of the present invention.

[0102] The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings and in conjunction with the embodiments. DETAILED DESCRIPTION

[0103] It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0104] Reference Figure 1 , Figure 1 This is a schematic diagram of the structure of a bank intelligent authorization device based on a shared operating system in the hardware operating environment involved in the embodiment of the present invention.

[0105] like Figure 1As shown, the shared operating system-based bank smart authorization device may include: a processor 1001, such as a central processing unit (CPU), a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005. The communication bus 1002 is used to enable communication between these components. The user interface 1003 may include a display and an input unit, such as a keyboard. Optionally, the user interface 1003 may also include a standard wired interface or a wireless interface. The network interface 1004 may optionally include a standard wired interface or a wireless interface (such as a wireless fidelity (WI-FI) interface). The memory 1005 may be a high-speed random access memory (RAM) or a stable non-volatile memory (NVM), such as a disk drive. Optionally, the memory 1005 may be a storage system independent of the processor 1001.

[0106] Those skilled in the art will understand that Figure 1 The structure shown in the figure does not constitute a limitation on the bank intelligent authorization device based on the shared operating system, and may include more or fewer components than shown in the figure, or a combination of certain components, or a different arrangement of components.

[0107] like Figure 1 As shown, the memory 1005 as a computer-readable storage medium may include an operating system, a network communication module, a user interface module, and a bank intelligent authorization program based on a shared operating system.

[0108] exist Figure 1 In the bank intelligent authorization device based on the shared operating system shown, the network interface 1004 is mainly used for data communication with the network server; the user interface 1003 is mainly used for data interaction with the user; the processor 1001 and the memory 1005 in the bank intelligent authorization device based on the shared operating system of the present invention can be set in the bank intelligent authorization device based on the shared operating system, and the bank intelligent authorization device based on the shared operating system calls the bank intelligent authorization program based on the shared operating system stored in the memory 1005 through the processor 1001, and executes the bank intelligent authorization method based on the shared operating system provided by the embodiment of the present invention.

[0109] The embodiment of the present invention provides a bank intelligent authorization method based on a shared operating system, referring to Figure 2 , Figure 2 2 is a flow chart of a first embodiment of a bank intelligent authorization method based on a shared operating system according to the present invention.

[0110] In this embodiment, the bank intelligent authorization method based on the shared operating system includes the following steps:

[0111] Step S10: In response to the service authorization request sent by the service counter, the service data of the service to be processed is obtained based on the service authorization request.

[0112] It should be understood that the execution subject of this embodiment may be a computing service device with data processing, network communication, and program execution capabilities, such as a computer, server, or mobile phone, or a terminal electronic device capable of performing the aforementioned functions. This embodiment and the following embodiments will be described below using a bank intelligent authorization device (hereinafter referred to as the authorization device) based on a shared operating system as an example.

[0113] It should be noted that a business authorization request can be a business operation request initiated by a business counter (such as a bank manual counter or smart counter) that requires centralized authorization review and verification and authorized execution, such as account opening, transfer, loan approval and other services.

[0114] It should be noted that business data can be a data set or information set related to the pending business of the business authorization request. Business data includes but is not limited to business type, customer type, user biometric information (such as facial information or fingerprint information), OCR data, user identity information, transaction amount, historical behavior records, device information, geographic location, timestamp, etc.

[0115] In some embodiments, the service counter sends a service authorization request through a standardized interface (such as the HTTP protocol). After the authorization device receives the request through the API, it collects multi-source business data in real time (including structured data and unstructured data. Structured data includes user ID, transaction amount, account balance, historical transaction frequency, etc.; unstructured data may include scanned copies of ID cards uploaded by users, camera images of transaction scenes, voice call records, etc.).

[0116] In some embodiments, the authorized device can perform OCR recognition on image data in the business data, extract text data, and use natural language processing (NLP) (such as the BERT model) to extract keywords (such as "urgent transfer", "large withdrawal", etc.) from the text data.

[0117] Step S20: performing a risk level assessment on the pending business based on the business type to determine the business risk level of the pending business.

[0118] It should be noted that the business risk level can be the degree of business operation risk assessed based on business data, and is usually divided into low risk, medium risk, high risk and other levels. In some embodiments, the business risk level can be used to determine the strictness of subsequent authorization processes (such as manual review, real-time interception, etc.).

[0119] In the specific implementation, the authorization device conducts risk level assessment based on the business type. For example, risky businesses such as transfers, corporate business, and withdrawals are assessed as high-risk business risk levels; and businesses with lower risks such as balance checking and consulting business are assessed as low-risk or medium-risk business risk levels.

[0120] In some embodiments, the authorization device may assess the business risk level of the business to be processed from multiple dimensions based on business data, for example, based on the business type, large transfers, large withdrawals, loans, and corporate business may be assessed as high risk.

[0121] In some embodiments, the authorization device may calculate the risk probability based on a pre-built logistic regression model and perform a risk level assessment based on the risk probability. The risk probability calculation formula is as follows:

[0122]

[0123]

[0124] in, Indicates the risk probability of the pending business, represents the multidimensional risk feature vector, Represents the intercept term, which can be optimized by maximum likelihood estimation (MLE) or cross entropy loss function. represents the combined weight of the multidimensional risk feature vector, represents the time factor, Indicates the business risk level.

[0125] Step S30: Input the branch data of the business branch to which the business counter belongs, the customer data of the customer handling the pending business, and the teller data of the teller handling the pending business into the risk profile assessment model to generate a risk profile of the pending business.

[0126] It should be noted that the risk profile assessment model is constructed based on multi-dimensional feature data of multiple external audit nodes, and the multi-dimensional feature data includes customer feature data, branch feature data and teller feature data.

[0127] It should be noted that the risk profile can be a comprehensive assessment result of the business risk of pending business generated based on a comprehensive analysis of multi-dimensional data (customer data, branch data, teller data). For example, the risk profile may include risk labels (such as "fraud tendency" and "abnormal operation"), risk probability, risk source and other information.

[0128] It should be noted that branch data can be characteristic information of the branch to which the counter belongs, including historical risk records, geographic location, business volume, and equipment security level. The aforementioned customer data can be information about the customer who transacted, including identity information, transaction behavior data (such as historical transaction frequency and amount distribution), credit score, and anti-fraud tags. The aforementioned teller data can be characteristic information related to the teller who transacted, including, for example, years of service, historical operational compliance records, number of abnormal operations, and authority level.

[0129] In some embodiments, the authorization device may pre-process branch data, customer data, and teller data, and then extract features of each type of data to obtain multi-dimensional features. Then, a knowledge graph may be constructed to associate customer features, branch features, and teller features. The multi-dimensional features may be aggregated using a graph neural network (GNN), and a risk profile of the pending business may be generated based on the aggregated features.

[0130] Step S40: Match the authorization review policy corresponding to the pending business according to the risk profile and the business risk level.

[0131] It should be noted that the authorization review strategy includes a high-risk authorization review strategy and a medium- and low-risk authorization review strategy. The medium- and low-risk authorization review strategy includes determining whether to directly authorize based on the daily authorization ratio of the business counter. The high-risk business processing authorization review strategy includes a target individual authorization review strategy and a target business authorization review strategy. The target individuals include agent customers and customers whose age is below the preset age threshold. The target businesses include cash flow business, loan business and corporate business.

[0132] In some embodiments, the authorization device can classify pending businesses into high-risk businesses, medium-risk businesses, and low-risk businesses based on risk profiles and business risk levels; classify pending businesses into agent-handled businesses, minor-handled businesses, and elderly customer-handled businesses based on the individual type of the current customer handling the pending business; classify pending businesses into transfer businesses, loan businesses, corporate businesses, etc. based on business types, and match corresponding authorization review strategies for pending businesses based on the business classification results.

[0133] In some embodiments, the authorization review policy may include an agent business review policy, a high-risk business review policy, a young customer review policy, a loan review policy, a corporate business review policy, a transfer business review policy, an elderly customer review policy, etc.

[0134] It should be noted that for medium and low-risk businesses, the authorization device can perform authorization processing based on the medium and low-risk authorization review strategy. The medium and low-risk authorization review strategy may include automatic matching of basic business rules. If the verification is passed, authorization-free processing will be adopted. For authorization-free businesses, in order to prevent moral hazard, the risk score calculated according to the risk model is matched with the risk interval to obtain the corresponding system authorization ratio. The authorization device can dynamically choose whether to automatically authorize the business to be handled based on the actual authorization ratio of the day, thereby reducing redundant review processes.

[0135] Step S50: performing authorization review on the pending business according to the authorization review policy and the business data to obtain an authorization review result.

[0136] In some embodiments, the authorization review policy may include a basic review policy and a dynamic review policy. The basic review policy may be a basic review policy such as biometric recognition, OCR recognition, and NLP recognition; the above-mentioned dynamic review policy may be an review policy that is adaptively adjusted according to the business type, risk profile, risk level, and customer type. The authorization device may add a dynamic review policy on the basis of the basic review policy to generate an authorization review policy that adapts to the current scenario. For example, the business to be processed is a high-risk business, and the customer type is to be handled by an agent. Therefore, the authorization device may add an agent's power of attorney review policy on the basis of the basic review policy.

[0137] In some embodiments, low-risk businesses can be directly and automatically authorized; medium-risk businesses can be authorized and reviewed based on basic review policies; high-risk businesses can be reviewed by adding dynamic review policies to the basic review policies.

[0138] Step S60: performing abnormal time sequence behavior analysis based on the historical business data of the business outlets and the historical business data of the customers, and performing quality inspection on the authorization review results based on the abnormal time sequence behavior analysis results.

[0139] It should be noted that historical branch business data may include the branch's historical business volume, transaction amounts, risk events, and the corresponding event occurrence time. The aforementioned customer historical business data may include customer transaction records, risk tag time series, operating device change records, etc.

[0140] It should be noted that the above-mentioned abnormal time sequence behavior analysis results may include abnormal time sequence behavior analysis results of the business outlet and abnormal time sequence behavior analysis results of the customer. This embodiment can determine whether the business outlet and / or the customer have abnormal behavior based on the abnormal time sequence behavior analysis results of the business outlet and the customer.

[0141] In some embodiments, the authorization device can obtain the business volume, transaction amount, risk event occurrence time, etc. of the branch in the past 30 days from the time series database; the authorization device can extract the customer's transaction records, risk label time series, operating equipment change records, etc. in the past 180 days.

[0142] In some embodiments, the authorization device can align time series data of different granularities, fill in missing time series data, and then standardize the behavior and event data (such as transaction amount, number of transactions, business behavior, etc.), and use the Long Short Term Memory (LSTM) model to capture the behavioral dependencies between business outlets and customers (such as periodic fluctuations in holiday transaction volume, transition probability of transaction frequency, etc.), and detect anomalies through reconstruction error: if the error exceeds a threshold (such as 0.3), it is marked as an anomaly.

[0143] Step S70: Correct the authorization review result based on the quality inspection result to obtain a target authorization result, and feed back the target authorization result to the business counter.

[0144] In some embodiments, the authorization device may combine the original authorization review result and quality inspection result through a decision tree for correction to generate a final target authorization result: target authorization result = DecisionTree (authorization review result, quality inspection result, confidence level).

[0145] For example, if the original authorization review result is "passed" and the quality inspection result is "normal", the output target authorization result is "authorization passed";

[0146] If the original authorization review result is "passed", and the quality inspection result is "customer behavior abnormality", and the high confidence abnormality, the output target authorization result is "intercept authorization", and it can return to step S50 for review to form a quality inspection closed-loop review.

[0147] If the quality inspection results conflict with the original results but the confidence levels are similar, manual arbitration will be triggered and a manual arbitration request will be sent to the business counter or a branch or counter with higher authority.

[0148] In some embodiments, the authorization device obtains business logs, which include branch business logs, customer business logs, and teller business logs; generates an authorization behavior time series based on the business logs, and trains a pre-built original portrait model based on the authorization behavior time series; obtains initial model parameters based on the trained original portrait model; hashes the initial model parameters to generate encrypted model parameters, and sends the encrypted model parameters to each external audit node; in response to the external node model parameters sent by each external audit node, aggregates the external node model parameters to obtain global model parameters, and the external node model parameters are obtained by each external audit node decrypting the encrypted model parameters and training a local portrait model based on the initial model parameters obtained after decryption; trains the trained original portrait model based on the global model parameters to obtain a risk portrait assessment model.

[0149] In some embodiments, the authorization device extracts log semantic information and timestamp information from the business log; obtains historical event information based on the log semantic information, the historical event information including branch event information, teller event information, and customer event information; performs event behavior analysis based on the historical event information to obtain event behavior characteristics of each historical event; constructs an event causal graph based on the event behavior characteristics; generates an authorization behavior time series based on the timestamp information, and constructs an event graph based on the authorization behavior time series; establishes a path index based on the causal path in the causal graph and the temporal path in the event graph; and Event nodes are extracted from the event graph, and causal pair nodes are extracted from the causal graph; the event graph and the causal graph are merged based on the path index, the event nodes and the causal pair nodes to generate a target hypergraph, wherein the target hypergraph includes causal hyperedges and temporal hyperedges; the node information of each hyperedge neighbor node in the target hypergraph is aggregated to obtain a hypergraph embedding feature vector; the original features of each node in the target hypergraph are fused with the hypergraph embedding feature vector to obtain a comprehensive feature vector; a data set is generated based on the comprehensive feature vector, and a pre-built original portrait model is trained based on the data set.

[0150] In some embodiments, the authorization device determines the organization type of each external audit node and obtains historical audit behavior information of each external audit node; performs feature analysis on the historical audit behavior information to obtain multi-dimensional behavior feature information, wherein the multi-dimensional behavior feature information includes response delay characteristics, parameter noise characteristics, historical abnormal behavior characteristics, and network stability characteristics of each external audit node; normalizes the multi-dimensional behavior feature information, and calculates the direct trust of each external audit node based on the normalized multi-dimensional behavior feature information:

[0151]

[0152] in, Indicates direct trust. 、 、 、 Represent the weight coefficient of each behavioral feature, represents the normalized response delay characteristics, represents the normalized parameter noise characteristic, Represents the normalized historical abnormal behavior characteristics, represents the normalized network stability characteristics;

[0153] Obtain the historical credibility of each external audit node and the evaluation information between each external audit node, and calculate the recommended trustworthiness of each external audit node based on the evaluation information:

[0154]

[0155] in, Represents an external audit node The recommendation trust Represents external audit nodes The set of other external audit nodes that have interacted, Represents an external audit node Historical trust, Represents an external audit node External audit nodes Evaluation information;

[0156] Aggregate the direct trust and the recommended trust to obtain the comprehensive trust of each external audit node:

[0157]

[0158] in, represents the fusion weight coefficient, Indicates comprehensive trust;

[0159] The external node model parameters are aggregated based on the comprehensive trust and the organization type to obtain global model parameters.

[0160] In some embodiments, the authorization device determines the behavior time interval of historical audits of each external audit node based on the historical audit behavior information; and calculates the time-sensitive trust of each external audit node according to the behavior time interval and the comprehensive trust:

[0161]

[0162]

[0163] in, Indicates time-sensitive trust, represents the time decay factor, Indicates the weight of the current review behavior. Indicates historical audit trust. represents the decay rate, Indicates the time interval between behaviors;

[0164] The external audit nodes are classified according to the institution type and the time-sensitive trust to obtain a plurality of local node sets, wherein the local node sets include one or more external audit nodes; the external node model is locally aggregated based on the comprehensive trust of each node in each local node set to obtain local model parameters of each local node set; the historical contribution of each node in the local node set in the historical model parameters is obtained; the collective contribution of each local node set is determined based on the historical contribution, and the collective trust of each local node set is determined based on the comprehensive trust;

[0165] The local model parameters are aggregated based on the collective contribution and the comprehensive trust to obtain global model parameters:

[0166]

[0167] in, represents the global model parameters, represents the local model parameters, represents the balance coefficient, which is used to balance the comprehensive trust and the collective contribution. represents the collective contribution, Represents the collective confidence.

[0168] This embodiment effectively avoids the data island problem by conducting risk profiling of the business from multiple dimensions, overcomes the limitations of risk assessment of single-dimensional data, accurately identifies potential risks and anomalies in the business through dynamic assessment of risk levels and risk profiles, accurately matches corresponding authorization review strategies for pending businesses, and classifies business reviews into high-risk authorization reviews and medium- and low-risk authorization reviews, thereby strengthening the strictness of the review of high-risk businesses. It determines whether to directly authorize based on the authorization ratio of the day, thereby reducing redundant review links for medium- and low-risk businesses, thereby greatly improving the efficiency of authorization review. It performs abnormal time series behavior analysis through analysis of historical business data to achieve accurate quality inspection of authorization results, accurately identifies the degree of deviation between the customer's current behavior and the historical behavior quality inspection, and effectively reduces the risk of mis-authorization through quality inspection and correction mechanisms.

[0169] refer to Figure 3 , Figure 3 FIG2 is a flow chart of a second embodiment of the bank intelligent authorization method based on a shared operating system according to the present invention.

[0170] Based on the first embodiment above, in this embodiment, before step S30, the following steps are further included:

[0171] Step S31: Obtain service logs.

[0172] It should be noted that the business logs include branch business logs, customer business logs and teller business logs.

[0173] It should be noted that the branch business logs mentioned above may be logs that record the activities of the branch during its operations, and may include transaction flow, equipment status, risk events, teller operation time, business type distribution, etc. For example, the total number of transactions at a branch on a particular day, the proportion of high-risk business, and equipment failure alarm records.

[0174] The aforementioned customer business logs may be logs recording the customer's behavior when conducting business, and may include transaction time, amount, device fingerprint, IP address, number of input errors, business type preferences, etc. For example, a customer's transfer history within 30 days, number of account login failures, and device models used.

[0175] The teller business log can be a log that records the teller's business operations, and may include the type of business processed, duration, permission usage records, number of abnormal operations, and approval rate. For example, a teller processed 50 transfers on a certain day, of which 3 were marked as "delayed operations" and 2 triggered risk interception.

[0176] In some embodiments, the authorization device may pre-process the collected raw log data, such as performing data cleaning, feature extraction, and abnormal data processing on the raw log data.

[0177] Step S32: Generate an authorization behavior time series based on the business log, and train the pre-built original portrait model according to the authorization behavior time series.

[0178] It can be understood that this embodiment extracts time series features related to authorization behavior from business logs, such as extracting business type proportions, risk score fluctuations, authorization result distribution, etc. from business logs.

[0179] In some embodiments, the authorization device may use a sliding window to aggregate features of different dimensions to generate a multidimensional time series, referring to the following formula, where: Represents the feature vector of the t-th time window, including business type, risk score, authorization result, etc.

[0180]

[0181] Furthermore, in order to improve the timeliness and accuracy of risk assessment and provide effective timing support for model training, the above step S32 may include:

[0182] Step S3201: extracting log semantic information and timestamp information from the business log.

[0183] It should be noted that log semantic information can be structured semantic content extracted from business logs, such as "Customer A processed a transfer at branch B through teller C" or "Device D malfunctioned at time T." The above-mentioned timestamp information can be the timestamp corresponding to each log content or log action recorded in the business log.

[0184] Step S3202: Acquire historical event information based on the log semantic information.

[0185] It should be noted that the historical event information includes branch event information, teller event information, and customer event information. In this embodiment, historical events can be divided into three categories: branch events, teller events, and customer events. Branch events include abnormal branch transaction behavior and historical branch authorization frequency; teller events include teller operation delays, teller authority violations, and teller authorization transaction frequency; and customer events include remote login events, abnormal device login events, and abnormal transaction events.

[0186] Step S3203: Perform event behavior analysis based on the historical event information to obtain event behavior features of each historical event.

[0187] It should be noted that event behavior characteristics include branch behavior characteristics, teller behavior characteristics and customer behavior characteristics.

[0188] In some embodiments, branch behavior characteristics may include equipment failure rate, proportion of high-risk business, transaction volume fluctuation coefficient, etc.; teller behavior characteristics may include standard deviation of operation time, number of authority violations, review pass rate, etc.; customer behavior characteristics may include number of IP changes, number of equipment changes, number of abnormal inputs, number of authorization rejections, number of authorization requests, etc.

[0189] Step S3204: Construct an event causal graph based on the event behavior characteristics.

[0190] In some embodiments, the authorization device can use Granger causality testing or Bayesian networks to analyze the causal relationships between events and construct causal edges between events. For example, a branch device failure may lead to a customer transaction interruption, or a teller's excessive authority may lead to a customer engaging in a risky transaction. A causal graph is constructed based on the causal edges and event nodes.

[0191] Step S3205: Generate an authorization behavior time series according to the timestamp information, and construct an event graph based on the authorization behavior time series.

[0192] In some embodiments, the authorization device may construct event nodes and timing edges based on the authorization behavior time series, and construct an event graph based on the event nodes and timing edges.

[0193] Step S3206: Establish a path index based on the causal path in the causal graph and the temporal path in the event graph.

[0194] It can be understood that this embodiment combines the causal path of the causal graph (such as "equipment failure → transaction interruption") and the temporal path of the event graph (such as "T1 → T2 → T3") to establish a path index, where the path index = {causal path, temporal path, association weight}.

[0195] Step S3207: extracting event nodes from the event graph and extracting causal pair nodes from the causal graph;

[0196] Step S3208: Merge the event graph and the causal graph based on the path index, the event node, and the causal pair node to generate a target hypergraph.

[0197] In some embodiments, the target hypergraph includes causal hyperedges and temporal hyperedges. This embodiment can merge the causal edges (hyperedges) of the causal graph and the temporal edges (hyperedges) of the temporal graph into the target hypergraph based on a hypergraph neural network:

[0198]

[0199] in, represents the target hypergraph, Indicates event node combination, represents the set of causal hyperedges, Represents a set of temporal hyperedges.

[0200] Step S3209: Aggregate the node information of each hyperedge neighbor node in the target hypergraph to obtain a hypergraph embedding feature vector.

[0201] In some embodiments, the authorized device may aggregate neighbor node information via a hypergraph convolutional network:

[0202]

[0203] in, Representation node v The hypergraph embedding vector of Represents the original feature matrix of the node, represents the node v The node information of the hyperedge neighbor node.

[0204] Step S3210: performing feature fusion on the original features of each node in the target hypergraph and the hypergraph embedded feature vector to obtain a comprehensive feature vector;

[0205] Step S3211: Generate a data set based on the comprehensive feature vector, and train the pre-built original portrait model based on the data set.

[0206] In some embodiments, the authorization device may extract a comprehensive feature vector of a node from the target hypergraph, generate a data set based on the comprehensive feature vector, and train a pre-built original portrait model based on the data set.

[0207] Step S33: Obtain initial model parameters based on the trained original portrait model.

[0208] It should be noted that the initial model parameters may include weight parameters, bias, etc. of the original portrait model after training.

[0209] Step S34: Hash-encrypt the initial model parameters to generate encrypted model parameters, and send the encrypted model parameters to each external audit node.

[0210] In some embodiments, the authorized device may use homomorphic encryption to encrypt the initial model parameters and calculate a hash value to ensure integrity:

[0211]

[0212]

[0213] in, represents the encryption model parameters, represents the initial model parameters.

[0214] Step S35: In response to the external node model parameters sent by each external audit node, the external node model parameters are aggregated to obtain global model parameters.

[0215] It should be noted that the external node model parameters are obtained by each external audit node decrypting the encrypted model parameters and training the local portrait model based on the initial model parameters obtained after decryption.

[0216] In some embodiments, the authorization device is a central node, and the external audit node is an external node that is communicatively connected to the central node. The external audit node can be other outlets other than the business processing outlets or a third-party audit structure (such as a regulatory agency, etc.).

[0217] In some embodiments, the external audit node can use the private key to decrypt the received encrypted model parameters to obtain the initial model parameters, train the local model based on the initial model parameters and local private data (such as customer transaction records, the audit institution's own regulatory data and assessment data), and obtain the external node model parameters:

[0218]

[0219] in, The external node model parameters obtained by training the local model for the external audit node, Local privacy parameters for external audit nodes.

[0220] The external audit node extracts the external node model parameters from the trained local model, adds noise to the external node model parameters, and sends the noise-added external node model parameters to the central node (ie, the authorized device).

[0221] In some embodiments, the authorization device may use a federated averaging algorithm to aggregate the external node model parameters sent by each external audit node; or may aggregate the external node model parameters of each external audit node by weighted averaging.

[0222] Furthermore, in order to effectively aggregate the model parameters of multiple external review nodes, the above step S35 may include:

[0223] Step S3501: Determine the organization type of each external audit node and obtain historical audit behavior information of each external audit node;

[0224] Step S3502: Perform feature analysis on the historical audit behavior information to obtain multi-dimensional behavior feature information;

[0225] Step S3503: normalizing the multi-dimensional behavior feature information, and calculating the direct trustworthiness of each external audit node based on the normalized multi-dimensional behavior feature information;

[0226] Step S3504: Obtain the historical credibility of each external review node and the evaluation information between each external review node, and calculate the recommendation trust of each external review node based on the evaluation information;

[0227] Step S3505: Aggregate the direct trustworthiness and the recommended trustworthiness to obtain the comprehensive trustworthiness of each external review node;

[0228] Step S3506: Aggregate the external node model parameters based on the comprehensive trust and the organization type to obtain global model parameters.

[0229] It should be noted that the institution type may include a financial institution, a regulatory agency, a third-party certification structure, etc. In some embodiments, the authorization device may determine the institution type based on the name, certification qualifications, and filing information of the external audit node institution.

[0230] It should be noted that the multi-dimensional behavioral characteristic information includes the response delay characteristics, parameter noise characteristics, historical abnormal behavior characteristics, and network stability characteristics of each external audit node. Response delay characteristics can include the node's average response time, delay fluctuation rate, timeout rate, etc.; parameter noise characteristics can include the noise ratio, noise distribution, noise amplitude, and invalid data ratio of the parameters provided by the node; historical abnormal behavior characteristics can include the node's abnormal behavior ratio and the response mode abnormality ratio; and network stability characteristics can include the node's packet loss rate, bandwidth fluctuation, and network protocol compatibility.

[0231] It should be noted that the multi-dimensional behavior characteristic information is normalized according to the following formula:

[0232]

[0233] in, Indicates direct trust. Represent the weight coefficient of each behavioral feature, represents the normalized response delay characteristics, represents the normalized parameter noise characteristic, Represents the normalized historical abnormal behavior characteristics, Represents the normalized network stability characteristics.

[0234] It should be noted that the recommended trust level of each external review node is calculated using the following formula:

[0235]

[0236] in, Represents an external audit node The recommendation trust Represents external audit nodes The set of other external audit nodes that have interacted, Represents an external audit node Historical trust, Represents an external audit node External audit nodes evaluation information.

[0237] It should be noted that the comprehensive trustworthiness of each external audit node is calculated using the following formula:

[0238]

[0239] in, represents the fusion weight coefficient, Indicates the overall trust level.

[0240] It can be understood that this embodiment analyzes the direct trust and recommended trust of each external audit node, and aggregates parameters from the two dimensions of trust, so as to accurately configure the parameter weights of each node, and improve the efficiency and accuracy of global parameter aggregation through comprehensive trust-driven weight aggregation.

[0241] Furthermore, in order to improve parameter quality, global parameter aggregation is performed from a time dimension. The above step S3506 may include:

[0242] Step S35061: Determine the time interval of historical audit behavior of each external audit node based on the historical audit behavior information;

[0243] Step S35062: Calculate the time-sensitive trustworthiness of each external audit node based on the behavior time interval and the comprehensive trustworthiness;

[0244] Step S35063: Classify the external audit nodes according to the organization type and the time-sensitive trust to obtain multiple local node sets, where the local node sets include one or more external audit nodes;

[0245] Step S35064: performing local aggregation on the external node model based on the comprehensive trust of each node in each local node set to obtain local model parameters of each local node set;

[0246] Step S35065: Obtain the historical contribution of each node in the local node set in the historical model parameters;

[0247] Step S35066: determining the collective contribution of each local node set based on the historical contribution, and determining the collective trust of each local node set based on the comprehensive trust;

[0248] Step S35067: Aggregate the local model parameters based on the collective contribution and the comprehensive trust to obtain global model parameters.

[0249] It should be noted that the time-sensitive trust calculation formula is as follows:

[0250]

[0251]

[0252] in, Indicates time-sensitive trust, represents the time decay factor, Indicates the weight of the current review behavior. Indicates historical audit trust. represents the decay rate, Indicates the time interval between behaviors.

[0253] It should be noted that the following formula is used for the aggregation of local model parameters:

[0254]

[0255] in, represents the global model parameters, represents the local model parameters, represents the balance coefficient, which is used to balance the comprehensive trust and the collective contribution. represents the collective contribution, Represents the collective confidence.

[0256] It is understandable that the authorized device can identify the activity and response speed of the node by analyzing the historical behavior time interval. For example, the node that frequently participates in the audit may be more reliable, while the node that has not participated for a long time may have anomalies (such as hardware failure or malicious behavior).

[0257] It should be understood that the authorized device can combine the time interval and the comprehensive trust, introduce the time decay factor, and perform trust analysis from the time dimension. For example, a node that has not participated in the audit for a long time may be marked as "inactive", while a node that participates frequently but has abnormal parameters may be identified as "malicious", thereby reducing its negative impact on the global model.

[0258] It can be understood that this embodiment can perform contribution analysis at the level of local collections through local aggregation node parameters, evaluate the overall value of the collection based on the sum of the members' historical contributions, and evaluate the reliability of the collection through the weighted average of the members' comprehensive trust, thereby improving the efficiency and stability of global aggregation. For example, a collection with high contribution but low trust may be marked as "needs to be monitored", while a collection with high trust and high contribution may be aggregated first.

[0259] It should be understood that this embodiment achieves a dynamic balance between trust and contribution through time-sensitive trust and historical contribution, adapts to real-time changes in node behavior, and effectively filters malicious nodes through multi-layer screening (individual trust, collective contribution), reduces attack risks, and reduces global communication overhead through local aggregation. At the same time, it avoids the participation of inefficient nodes through contribution screening.

[0260] Step S36: Train the trained original portrait model based on the global model parameters to obtain a risk portrait assessment model.

[0261] It can be understood that this embodiment updates the model parameters of the trained original portrait model based on the global model parameters, generates a data set according to the business log and authorization behavior time series, trains the updated original portrait model based on the data set, and obtains a risk portrait assessment model.

[0262] This embodiment collects business logs and conducts multi-dimensional time series analysis on the authorization behaviors of outlets, customers, and tellers based on the business logs, thereby capturing the dynamic characteristics of business risks, and training the original portrait model according to the time of the authorization behavior, thereby providing time-dependent characteristics for model training, significantly improving the timeliness and accuracy of risk prediction, and performing global parameter aggregation on the external node model parameters obtained based on local parameter training by combining federated learning with multiple external audit nodes, ensuring data privacy while effectively avoiding the problem of data silos and improving the accuracy of risk assessment.

[0263] refer to Figure 4 , Figure 4 2 is a flow chart of a third embodiment of the bank intelligent authorization method based on a shared operating system according to the present invention.

[0264] Based on the above embodiments, in this embodiment, step S50 further includes:

[0265] Step S501: If the authorization review policy is a target individual authorization review policy, the individual type of the current customer of the pending business is obtained according to the business data.

[0266] It should be noted that the target individual authorization review strategy can be applied to the authorization review scenario of the target individual's customers handling business. The individual types of the target individual may include agent customers (that is, the individual entrusts the agent to handle the business), customers whose age is below the preset age threshold (that is, minor customers), older customers (for example, customers over 70 years old), etc.

[0267] Step S502: If the individual type of the current customer is an agent customer, then the original image of the power of attorney for the business to be handled is obtained.

[0268] It should be noted that if the current customer's individual type is an agent customer, the business to be handled is determined to be a business scenario in which the customer himself entrusts an agent to handle it (i.e., a scenario in which the customer does not handle the business himself). In this scenario, a power of attorney is required to be issued by the entrusting party, and the entrusted party (i.e., the agent) assists in handling the business.

[0269] It should be understood that since there is no standard template for the power of attorney, it is difficult for machine recognition to accurately identify the content of the power of attorney, and the efficiency and accuracy of manual recognition are low. Therefore, this embodiment performs grayscale processing and maximum inter-class variance calculation on the original image of the power of attorney, and divides the power of attorney image into a foreground layer and a background layer based on the target grayscale threshold, thereby effectively narrowing the recognition range. By dividing the foreground layer into a handwritten font area and a printed font area, and performing morphological opening operations on them respectively, the noise in the image is effectively reduced, and the missing and broken characters in the font are supplemented, thereby improving the recognition accuracy of handwritten fonts and printed fonts, thereby accurately identifying the content of the power of attorney and effectively improving the accuracy of authorization review in the agent's business scenario.

[0270] Step S503: grayscale processing is performed on each pixel in the original image to obtain a grayscale image of the power of attorney.

[0271] It should be noted that this embodiment can pre-process the original image to enhance image quality and remove noise. In some embodiments, the pre-processing may include grayscale processing of the original image, direction correction processing (for example, aligning and correcting the font direction of handwritten fonts with the font direction of printed fonts), Gaussian filtering smoothing processing, binarization processing, etc.

[0272] It is understood that in this embodiment, grayscale processing can be performed on the original image to calculate the grayscale value of each pixel in the original image, and the grayscale value calculation refers to the following formula:

[0273]

[0274] in, Represents the grayscale image of the power of attorney at coordinates The pixel value at Represent the weight coefficients of the red, green, and blue channels respectively, Represent the coordinates in the original image The pixel values ​​at the red, green, and blue channels are shown.

[0275] Step S504: obtaining the grayscale value range of the grayscale image of the power of attorney, and traversing each pixel point of the grayscale image of the power of attorney;

[0276] Step S505: performing a grayscale distribution analysis on each pixel in the grayscale image of the power of attorney according to the grayscale value range and the traversal result, and setting an initial grayscale threshold based on the grayscale distribution analysis result.

[0277] In some embodiments, the authorization device may determine the grayscale value range [0, L-1] of the grayscale image of the power of attorney, and initialize the grayscale value t to 0, starting from t=0, gradually increasing t until t=L-2, and for each t, ​​calculating the proportion of pixels to the left of the threshold t, the proportion of pixels to the right of the threshold t, the average grayscale value of the pixels on the left, the average grayscale value of the pixels on the right, and the inter-class variance, and finding the maximum inter-class variance among all possible t.

[0278] Step S506: performing maximum inter-class variance calculation on the power of attorney grayscale image based on the initial grayscale threshold and the grayscale distribution analysis result to determine a target grayscale threshold.

[0279] It should be noted that the maximum between-class variance is calculated according to the following formula:

[0280]

[0281] in, represents the target grayscale threshold, represents the initial grayscale threshold, Represents the initial grayscale threshold The percentage of pixels on the left side, Represents the initial grayscale threshold The pixel ratio on the right side, represents the average gray value of the pixels on the left, Represents the average gray value of the pixels on the right.

[0282] Step S507: dividing the power of attorney grayscale image into a foreground layer and a background layer according to the target grayscale threshold.

[0283] It should be noted that this embodiment divides the pixels in the power of attorney grayscale image into two categories by determining the target grayscale threshold: pixels greater than the target grayscale threshold are marked as foreground (usually white), and pixels less than the target grayscale threshold are marked as background (usually black), thereby achieving layer segmentation.

[0284] Step S508: Divide the foreground layer into a handwritten font area and a printed font area.

[0285] It should be noted that the handwritten font area can be the distribution area of ​​the client's handwritten font in the power of attorney, and the printed font area can be the distribution area of ​​the font printed by the printer in the power of attorney. In actual power of attorney, there will be both the client's own and the agent's handwritten Chinese characters. Due to the significant differences in writing styles between different individuals, direct recognition is difficult. Printed fonts are generally standardized fonts (such as Kaiti and Songti). Therefore, this embodiment can improve the accuracy of power of attorney content recognition by partitioning the handwritten and printed fonts in the power of attorney.

[0286] In some embodiments, the authorization device can obtain multi-dimensional feature information by performing texture feature extraction, edge feature extraction, and property feature extraction on the foreground layer, classify the fonts based on the multi-dimensional feature information, and locate the handwritten font area and the printed font area.

[0287] For example, for each region in an image, the authorized device can calculate the mean and variance of its LBP histogram to obtain the texture characteristics of the region. It can then compare the texture characteristics of different regions to distinguish between handwritten and printed text. Generally speaking, handwritten text has more complex texture characteristics and a larger variance in its LBP histogram, while printed text has simpler texture characteristics and a smaller variance in its LBP histogram.

[0288] For example, for each region in an image, the authorized device can calculate the mean and variance of its edge strength to obtain the edge characteristics of the region. It can then compare the edge characteristics of different regions to distinguish between handwritten and printed text. Generally speaking, handwritten text regions have stronger edge strength and more complex edge directions, while printed text regions have weaker edge strength and simpler edge directions.

[0289] For example, for each region in an image, the authorized device can calculate the mean and variance of its area, perimeter, and aspect ratio to obtain the region's shape characteristics. This can then be compared to distinguish between handwritten and printed text. Generally speaking, handwritten text has more complex shape characteristics, resulting in greater variance in area, perimeter, and aspect ratio; whereas printed text has simpler shape characteristics, resulting in smaller variance in area, perimeter, and aspect ratio.

[0290] Step S509: obtaining handwritten structural elements of the handwritten font area according to the structural information of the handwritten font area, and obtaining printed structural elements of the printed font area according to the structural information of the printed font area.

[0291] It should be noted that the structural element is a binary image and can be a simple shape such as a square, a circle or a cross, which is used to define the neighborhood of the operation.

[0292] It should be noted that the above structural information may be the structural information of handwritten fonts and printed fonts, and may be the font structural feature information of handwritten fonts and printed fonts.

[0293] It will be appreciated that this embodiment can select appropriate structuring elements based on the characteristics of handwritten and printed fonts. Generally, printed fonts have thicker and more regular strokes, while handwritten fonts have thinner and more irregular strokes. Therefore, a larger structuring element can be used to process printed font areas, and a smaller structuring element can be used to process handwritten font areas. For example, a 5x5 square structuring element can be selected to process printed font areas, and a 3x3 square structuring element can be selected to process handwritten font areas.

[0294] Step S510: performing morphological opening operations on the handwritten font area and the printed font area based on the handwritten structural elements and the printed structural elements, respectively, to obtain a target handwritten image and a target printed image.

[0295] It should be noted that the morphological opening operation includes erosion and dilation. Erosion can be used to eliminate small objects in an image, separate connections between objects, and smooth object boundaries. In this embodiment, it can be used to eliminate invalid strokes, fonts, and noise in a letter of authorization. Dilation can be used to fill small holes in an image, connect breakpoints between objects, and smooth object boundaries. In this embodiment, it can be used to connect broken fonts and strokes and repair corroded fonts.

[0296] It is understood that the erosion operation may include: traversing each pixel in the handwritten font area and the printed font area, aligning the center of the structuring element with the pixel, and checking the values ​​of all pixels in the image area covered by the structuring element. If all pixel values ​​covered by the structuring element are 1, then the pixel is retained as 1; otherwise, the pixel is set to 0.

[0297] It should be understood that the dilation operation may include traversing each pixel in the handwritten font area and the printed font area, aligning the center of the structuring element with the pixel, and checking the values ​​of all pixels in the image area covered by the structuring element. If any pixel value in the area covered by the structuring element is 1, then the pixel is set to 1; otherwise, the pixel is left as 0.

[0298] In some embodiments, the authorized device can perform an opening operation on the printed text area using a larger structuring element to remove noise and connect broken strokes. For example, an erosion operation can use a 5x5 square structuring element to erode the printed text area to remove noise and small objects in the image. A dilation operation can then use the same 5x5 square structuring element to dilate the eroded image to connect broken strokes, making the printed text area more complete and clear.

[0299] In some embodiments, the authorized device may perform an opening operation on the handwritten area using a smaller structuring element to remove noise while preserving handwritten details. For example, an erosion operation may use a 3x3 square structuring element to erode the handwritten area, removing noise and small objects in the image. A dilation operation may then use the same 3x3 square structuring element to dilate the eroded image, preserving handwritten details and making the handwritten area clearer.

[0300] Step S511: performing content detection on the target handwritten image and the target printed image to obtain content information of the power of attorney for the pending business.

[0301] In some embodiments, the authorization device may extract print features of the target printed image, perform content recognition on the target printed image based on the print features and a pre-built deep learning model (eg, a CRNN model), and obtain the print content text.

[0302] In some embodiments, the authorized device may extract handwriting features of the target handwriting image, perform content recognition on the target handwriting image based on the handwriting features and a pre-built attention mechanism model (such as a Transformer model), and obtain the handwritten content text.

[0303] Step S512: performing authorization review on the pending business based on the content information of the power of attorney and obtaining the authorization review result.

[0304] In some embodiments, the authorization device can extract the key information of the entrustment (such as the entrusted business, entrusted amount, entrustment date, signature, etc.) based on the content information of the power of attorney, match and compare the key information of the entrustment with the key business information in the business data, and detect whether there is key information that fails to match (such as the entrustment amount does not match the business amount, etc.).

[0305] In some embodiments, the authorization device can locate the principal's signature area and the agent's (i.e., the trustee's) signature area based on the target printed image, detect the handwriting density of the principal's signature area and the agent's signature area, and determine whether there is valid handwriting in both the principal's signature area and the agent's signature area based on the handwriting density. If there is valid handwriting, the principal's historical signature and the agent's historical signature are obtained, and the principal's historical signature is compared with the signature in the principal's signature area, and the agent's historical signature is compared with the signature in the agent's signature area to detect whether there is a risk of forgery of the signature in the target handwritten image.

[0306] This embodiment performs grayscale processing and foreground segmentation on the power of attorney image, thereby accurately classifying the handwritten font area and printed font area in the power of attorney image, and performs morphological opening operations on the handwritten font area and the printed font area respectively, thereby effectively reducing the noise in the image, and completing the missing and broken characters in the font, thereby improving the recognition accuracy of handwritten fonts and printed fonts, thereby accurately identifying the content of the power of attorney and effectively improving the accuracy of authorization review in the agent handling business scenario.

[0307] In addition, an embodiment of the present invention also proposes a computer-readable storage medium, on which a bank intelligent authorization program based on a shared operating system is stored. When the bank intelligent authorization program based on a shared operating system is executed by a processor, the steps of the bank intelligent authorization method based on a shared operating system as described above are implemented.

[0308] The computer-readable storage medium provided herein may be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, systems, or devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to, an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, system, or device. The program code contained on the computer-readable storage medium may be transmitted using any suitable medium, including, but not limited to, wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.

[0309] The computer-readable storage medium may be included in the bank intelligent authorization device based on the shared operating system; or it may exist independently without being assembled into the bank intelligent authorization device based on the shared operating system.

[0310] In addition, an embodiment of the present invention also proposes a computer program product, including a bank intelligent authorization program based on a shared operating system, which implements the steps of the bank intelligent authorization method based on a shared operating system when executed by a processor.

[0311] The specific implementation of the computer program product of the present invention is basically the same as the above-mentioned embodiments of the bank intelligent authorization method based on the shared operating system, and will not be repeated here.

[0312] Reference Figure 5 , Figure 5 This is a structural block diagram of the first embodiment of the shared operating system of the present invention.

[0313] like Figure 5 As shown, the shared operation system proposed in the embodiment of the present invention includes:

[0314] The authorization request response module 10 is used to respond to the service authorization request sent by the service counter and obtain the service data of the service to be processed based on the service authorization request, wherein the service data includes the service type;

[0315] A risk level assessment module 20 is configured to assess the risk level of the pending business based on the business type and determine the business risk level of the pending business;

[0316] A risk profile generating module 30 is configured to input the branch data of the business branch to which the business counter belongs, the customer data of the customer handling the pending business, and the teller data of the teller handling the pending business into a risk profile assessment model to generate a risk profile of the pending business. The risk profile assessment model is constructed based on multidimensional feature data of multiple external audit nodes, the multidimensional feature data including customer feature data, branch feature data, and teller feature data.

[0317] An audit policy matching module 40 is configured to match the authorization audit policy corresponding to the pending business based on the risk profile and the business risk level. The authorization audit policy includes a high-risk authorization audit policy and a medium- and low-risk authorization audit policy. The high-risk business authorization audit policy includes a target individual authorization audit policy and a target business authorization audit policy. The target individuals include agent clients and clients below a preset age threshold. The target businesses include cash flow business, loan business, and corporate business. The medium- and low-risk authorization audit policy includes determining whether to directly authorize based on the daily authorization ratio of the business counter.

[0318] The authorization review module 50 is used to perform authorization review on the pending business according to the authorization review policy and the business data, and obtain the authorization review result;

[0319] An authorization quality inspection module 60 is configured to perform abnormal time series behavior analysis based on the historical business data of the business outlet and the historical business data of the customer, and to perform quality inspection on the authorization review result based on the abnormal time series behavior analysis result;

[0320] The authorization feedback module 70 is used to correct the authorization review result based on the quality inspection result, obtain the target authorization result, and feed back the target authorization result to the business counter.

[0321] This embodiment effectively avoids the data island problem by conducting risk profiling of the business from multiple dimensions, overcomes the limitations of risk assessment of single-dimensional data, accurately identifies potential risks and anomalies in the business through dynamic assessment of risk levels and risk profiles, accurately matches corresponding authorization review strategies for pending businesses, and classifies business reviews into high-risk authorization reviews and medium- and low-risk authorization reviews, thereby strengthening the strictness of the review of high-risk businesses. It determines whether to directly authorize based on the authorization ratio of the day, thereby reducing redundant review links for medium- and low-risk businesses, thereby greatly improving the efficiency of authorization review. It performs abnormal time series behavior analysis through analysis of historical business data to achieve accurate quality inspection of authorization results, accurately identifies the degree of deviation between the customer's current behavior and the historical behavior quality inspection, and effectively reduces the risk of mis-authorization through quality inspection and correction mechanisms.

[0322] The shared operating system provided in this application utilizes the bank intelligent authorization method based on the shared operating system described in the aforementioned embodiment, thereby resolving the technical issues surrounding bank intelligent authorization. Compared to the prior art, the shared operating system provided in this application offers the same beneficial effects as the bank intelligent authorization method based on the shared operating system described in the aforementioned embodiment. Other technical features of the shared operating system are the same as those disclosed in the aforementioned embodiment and are not further elaborated upon here.

[0323] It should be understood that the above is only an example and does not constitute any limitation to the technical solution of the present invention. In specific applications, those skilled in the art can make settings as needed, and the present invention does not impose any limitation on this.

[0324] It should be noted that the workflow described above is merely illustrative and does not limit the scope of protection of the present invention. In practical applications, technicians in this field can select part or all of it according to actual needs to achieve the purpose of the embodiment scheme, and no limitation is made here.

[0325] In addition, for technical details not fully described in this embodiment, please refer to the bank intelligent authorization method based on the shared operating system provided in any embodiment of the present invention, and will not be repeated here.

[0326] It should be noted that, in this document, the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, article, or system comprising a series of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or system. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article, or system comprising the element.

[0327] The serial numbers of the above embodiments of the present invention are for description only and do not represent the advantages or disadvantages of the embodiments.

[0328] Through the above description of the embodiments, those skilled in the art will clearly understand that the above-mentioned embodiments and methods can be implemented by means of software plus the necessary general-purpose hardware platform. Of course, hardware can also be used, but in many cases the former is a more preferred embodiment. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as read-only memory / random access memory, a magnetic disk, or an optical disk) and includes a number of instructions for enabling a terminal device (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0329] The above are only preferred embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention description and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A bank intelligent authorization method based on a shared operating system, applied to a shared operating system, characterized in that: The method comprises: In response to a service authorization request sent by a service counter, obtaining service data of the service to be processed based on the service authorization request, the service data including the service type; Performing a risk level assessment on the pending business based on the business type to determine the business risk level of the pending business; Inputting the branch data of the business branch to which the business counter belongs, the customer data of the customer handling the pending business, and the teller data of the teller handling the pending business into a risk profile assessment model to generate a risk profile of the pending business, wherein the risk profile assessment model is constructed based on multidimensional feature data of multiple external audit nodes, the multidimensional feature data including customer feature data, branch feature data, and teller feature data; Matching the authorization review strategy corresponding to the pending business according to the risk profile and the business risk level, the authorization review strategy includes a high-risk authorization review strategy and a medium- and low-risk authorization review strategy. The high-risk authorization review strategy includes a target individual authorization review strategy and a target business authorization review strategy. The target individuals include agent customers and customers under a preset age threshold. The target businesses include cash flow business, loan business, and corporate business. The medium- and low-risk authorization review strategy includes determining whether to directly authorize based on the daily authorization ratio of the business counter; Performing authorization review on the pending business according to the authorization review policy and the business data, and obtaining an authorization review result; Performing abnormal time series behavior analysis based on the historical business data of the business outlets and the historical business data of the customers, and performing quality inspection on the authorization review results based on the abnormal time series behavior analysis results; The authorization review result is corrected based on the quality inspection result to obtain the target authorization result, and the target authorization result is fed back to the business counter.

2. The bank intelligent authorization method based on the shared operating system as claimed in claim 1, characterized in that: Before inputting the branch data of the business branch to which the business counter belongs, the customer data of the customer handling the pending business, and the teller data of the teller handling the pending business into the risk profile assessment model and generating the risk profile of the pending business, the method further includes: Obtaining business logs, including branch business logs, customer business logs, and teller business logs; Generate an authorization behavior time series based on the business log, and train a pre-built original portrait model based on the authorization behavior time series; Obtain initial model parameters based on the trained original portrait model; Performing hash encryption on the initial model parameters to generate encrypted model parameters, and sending the encrypted model parameters to each external audit node; In response to the external node model parameters sent by each external audit node, the external node model parameters are aggregated to obtain global model parameters, wherein the external node model parameters are obtained by each external audit node decrypting the encrypted model parameters and training the local portrait model based on the initial model parameters obtained after decryption; The trained original portrait model is trained based on the global model parameters to obtain a risk portrait assessment model.

3. The bank intelligent authorization method based on the shared operating system as claimed in claim 2, characterized in that: Generating an authorization behavior time series based on the business log and training a pre-built original portrait model according to the authorization behavior time series includes: Extracting log semantic information and timestamp information from the business log; Acquire historical event information based on the log semantic information, the historical event information including branch event information, teller event information, and customer event information; Performing event behavior analysis based on the historical event information to obtain event behavior characteristics of each historical event; Constructing an event causal graph based on the event behavior characteristics; Generate an authorization behavior time series according to the timestamp information, and construct an event graph based on the authorization behavior time series; Establishing a path index based on the causal path in the causal graph and the temporal path in the event graph; Extracting event nodes from the event graph and extracting causal pair nodes from the causal graph; Merging the event graph with the causal graph based on the path index, the event node, and the causal pair node to generate a target hypergraph, wherein the target hypergraph includes a causal hyperedge and a temporal hyperedge; Aggregating node information of neighboring nodes of each hyperedge in the target hypergraph to obtain a hypergraph embedding feature vector; Performing feature fusion on the original features of each node in the target hypergraph and the hypergraph embedded feature vector to obtain a comprehensive feature vector; A data set is generated based on the comprehensive feature vector, and a pre-built original portrait model is trained based on the data set.

4. The bank intelligent authorization method based on a shared operating system as claimed in claim 3, characterized in that: The aggregating the external node model parameters to obtain global model parameters includes: Determine the institutional type of each external audit node and obtain historical audit behavior information of each external audit node; Performing feature analysis on the historical audit behavior information to obtain multi-dimensional behavior feature information, wherein the multi-dimensional behavior feature information includes response delay features, parameter noise features, historical abnormal behavior features, and network stability features of each external audit node; Normalize the multi-dimensional behavior feature information, and calculate the direct trust of each external audit node based on the normalized multi-dimensional behavior feature information: in, Indicates direct trust. Represent the weight coefficient of each behavioral feature, represents the normalized response delay characteristics, represents the normalized parameter noise characteristic, Represents the normalized historical abnormal behavior characteristics, represents the normalized network stability characteristics; Obtain the historical credibility of each external audit node and the evaluation information between each external audit node, and calculate the recommended trustworthiness of each external audit node based on the evaluation information: in, Represents an external audit node The recommendation trust Represents external audit nodes The set of other external audit nodes that have interacted, Represents an external audit node Historical trust, Represents an external audit node External audit nodes Evaluation information; Aggregate the direct trust and the recommended trust to obtain the comprehensive trust of each external audit node: in, represents the fusion weight coefficient, Indicates comprehensive trust; The external node model parameters are aggregated based on the comprehensive trust and the organization type to obtain global model parameters.

5. The bank intelligent authorization method based on the shared operating system as claimed in claim 4, characterized in that: The aggregating the external node model parameters based on the comprehensive trust and the organization type to obtain global model parameters includes: Determine the behavior time interval of historical audits of each external audit node based on the historical audit behavior information; The time-sensitive trust of each external audit node is calculated based on the behavior time interval and the comprehensive trust: in, Indicates time-sensitive trust, represents the time decay factor, Indicates the weight of the current review behavior. Indicates historical audit trust. represents the decay rate, Indicates the time interval between behaviors; Classifying the external audit nodes according to the organization type and the time-sensitive trust to obtain a plurality of local node sets, wherein the local node sets include one or more external audit nodes; Performing local aggregation on the external node model based on the comprehensive trust of each node in each local node set to obtain local model parameters of each local node set; Obtaining the historical contribution of each node in the local node set described in the historical model parameters; Determining the collective contribution of each local node set based on the historical contribution, and determining the collective trust of each local node set based on the comprehensive trust; The local model parameters are aggregated based on the collective contribution and the collective confidence to obtain global model parameters: in, represents the global model parameters, represents the local model parameters, represents the balance coefficient, which is used to balance the set trust and the set contribution. represents the collective contribution, Represents the collective confidence.

6. The bank intelligent authorization method based on a shared operating system according to any one of claims 1 to 5, characterized in that: The performing authorization review on the pending business according to the authorization review policy and the business data to obtain the authorization review result includes: If the authorization review policy is a target individual authorization review policy, then the individual type of the current customer of the pending business is obtained according to the business data; If the individual type of the current customer is an agent customer, then obtaining the original image of the power of attorney for the business to be handled; Perform grayscale processing on each pixel in the original image to obtain a grayscale image of the power of attorney: in, Represents the grayscale image of the power of attorney at coordinates The pixel value at Represent the weight coefficients of the red, green, and blue channels respectively, Represent the coordinates in the original image The pixel values ​​of the red, green, and blue channels at ; Obtaining the grayscale value range of the grayscale image of the power of attorney, and traversing each pixel point of the grayscale image of the power of attorney; Performing a grayscale distribution analysis on each pixel in the grayscale image of the power of attorney according to the grayscale value range and the traversal result, and setting an initial grayscale threshold based on the grayscale distribution analysis result; Based on the initial grayscale threshold and the grayscale distribution analysis result, the maximum inter-class variance calculation is performed on the grayscale image of the power of attorney to determine the target grayscale threshold: in, represents the target grayscale threshold, represents the initial grayscale threshold, Represents the initial grayscale threshold The percentage of pixels on the left side, Represents the initial grayscale threshold The pixel ratio on the right side, represents the average gray value of the pixels on the left, Represents the average gray value of the pixels on the right; dividing the power of attorney grayscale image into a foreground layer and a background layer according to the target grayscale threshold; Dividing the foreground layer into a handwritten font area and a printed font area; Acquire a handwritten structural element of the handwritten font area according to the structural information of the handwritten font area, and acquire a printed structural element of the printed font area according to the structural information of the printed font area; performing morphological opening operations on the handwritten font area and the printed font area based on the handwritten structure element and the printed structure element, respectively, to obtain a target handwritten image and a target printed image, wherein the morphological opening operations include an erosion operation and a dilation operation; Performing content detection on the target handwritten image and the target printed image to obtain content information of the power of attorney for the pending business; Based on the content information of the power of attorney, the pending business is authorized to be reviewed and the authorization review result is obtained.

7. The bank intelligent authorization method based on a shared operating system as claimed in claim 6, characterized in that: The step of dividing the foreground layer into a handwritten font area and a printed font area includes: Calculate the attribution cost of each pixel in the foreground layer, and divide the foreground layer into multiple superpixel blocks based on the attribution cost: in, Representing coordinates The attributed cost of the pixel at , Representing coordinates The gradient value of the pixel at represents the average gradient value of the superpixel cluster, represents the global gradient standard deviation of the foreground layer, Representing coordinates The spatial distance between the pixel at and the center of the superpixel cluster, represents the size parameter of the superpixel, Represent the balance weight coefficients of gradient and distance respectively; Perform character target detection on each superpixel block, and annotate the border of each superpixel block based on the character target detection result to generate multiple character borders; Performing feature extraction on each character frame and the characters within the character frame to obtain a frame feature vector and a character structure feature vector of each character frame, and determining the similarity of each character frame based on the frame feature vector and the character structure feature vector; The information entropy of each superpixel block is calculated based on the similarity: in, Represents a superpixel block The information entropy of Indicates character bounding boxes whose similarity is lower than the preset similarity threshold In superpixel blocks The probability of occurrence in The superpixel block is classified into a handwritten font area and a printed font area based on the information entropy, the information entropy of the handwritten font area is greater than a preset entropy threshold, and the information entropy of the printed font area is not greater than the preset entropy threshold.

8. A shared operating system, characterized in that: The shared operating system includes: An authorization request response module is used to respond to a service authorization request sent by a service counter and obtain service data of the service to be processed based on the service authorization request, wherein the service data includes a service type; A risk level assessment module, configured to assess the risk level of the pending business based on the business type and determine the business risk level of the pending business; a risk profile generation module, configured to input the branch data of the business branch to which the business counter belongs, the customer data of the customer handling the pending business, and the teller data of the teller handling the pending business into a risk profile assessment model to generate a risk profile of the pending business, wherein the risk profile assessment model is constructed based on multidimensional feature data of multiple external audit nodes, the multidimensional feature data including customer feature data, branch feature data, and teller feature data; An audit policy matching module is configured to match the authorization audit policy corresponding to the pending business according to the risk profile and the business risk level, wherein the authorization audit policy includes a high-risk authorization audit policy and a medium- and low-risk authorization audit policy. The high-risk authorization audit policy includes a target individual authorization audit policy and a target business authorization audit policy. The target individuals include agent clients and clients below a preset age threshold. The target businesses include cash flow business, loan business, and corporate business. The medium- and low-risk authorization audit policy includes determining whether to directly authorize based on the daily authorization ratio of the business counter. An authorization review module is used to perform authorization review on the pending business according to the authorization review policy and the business data, and obtain an authorization review result; An authorization quality inspection module is used to perform abnormal time series behavior analysis based on the historical business data of the business outlets and the historical business data of the customers, and to perform quality inspection on the authorization review results based on the abnormal time series behavior analysis results; The authorization feedback module is used to correct the authorization review result based on the quality inspection result, obtain the target authorization result, and feed back the target authorization result to the business counter.

9. A bank intelligent authorization device based on a shared operating system, characterized in that: The bank intelligent authorization device based on the shared operating system includes: a memory, a processor, and a bank intelligent authorization program based on the shared operating system stored on the memory and executable on the processor. The bank intelligent authorization program based on the shared operating system is configured to implement the bank intelligent authorization method based on the shared operating system as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a bank intelligent authorization program based on a shared operating system, and when the bank intelligent authorization program based on a shared operating system is executed by a processor, the bank intelligent authorization method based on a shared operating system as described in any one of claims 1 to 7 is implemented.

Images