Business function update methods, devices, systems, computer equipment, and storage media

Abstract

This application relates to a method, apparatus, system, computer equipment, and storage medium for updating business functions. The method includes: acquiring business data generated during the business process execution phase; identifying abnormal business data within the business data and acquiring corrected business data corresponding to the abnormal business data; determining a target function model from a function model pool corresponding to the target function category to which the abnormal business data belongs; the function model pool includes multiple function models corresponding to different function categories; and automatically searching for the function model corresponding to the target function category based on the target function category of the abnormal business data. The corrected business data is used as training samples, and the target function model is incrementally trained using these training samples to obtain the updated target function model. This method can improve the efficiency of updating function models and data security in business systems.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to a method, apparatus, system, computer equipment, storage medium, and computer program product for updating business functions. Background Technology

[0002] With the development of deep learning technology, it has been widely used in various fields. After the AI ​​functions based on deep learning are deployed to the actual production environment, if function upgrades are needed, field staff must first collect data from different hospitals that cause the AI ​​functions to malfunction, then send it to the developers, who will centrally perform the function upgrades before deploying it to various user terminals.

[0003] However, currently, when updating algorithms or models for AI functions based on deep learning, additional field personnel are required, increasing the company's personnel costs. Furthermore, the workload of field personnel in collecting data and upgrading functions is large, and the update projects are unclear, resulting in low update efficiency. Summary of the Invention

[0004] Therefore, it is necessary to provide a method, apparatus, system, computer equipment, computer-readable storage medium, and computer program product that can improve the efficiency and security of business function updates, addressing the aforementioned technical problems.

[0005] Firstly, this application provides a method for updating business functions. The method includes:

[0006] Acquire business data generated during the business process execution phase;

[0007] Identify abnormal business data in the business data and obtain the corrected business data corresponding to the abnormal business data;

[0008] From the functional model pool, determine the target functional model corresponding to the target functional category to which the abnormal business data belongs; the functional model pool includes multiple functional models corresponding to different functional categories.

[0009] The corrected business data is used as training samples, and the target function model is incrementally trained using the training samples to obtain the updated target function model.

[0010] In one embodiment, acquiring business data generated during the business process execution phase includes:

[0011] Obtain business function call requests during the business process execution phase and identify the target function category corresponding to the business function call request;

[0012] Based on the target function category, obtain the target function model corresponding to the target function category from the function model pool;

[0013] The system retrieves input data from the business function call request, calls the target function model to process the input data, and obtains the business data output by the target function model.

[0014] In one embodiment, identifying abnormal business data in the business data and obtaining corrected business data corresponding to the abnormal business data includes:

[0015] Identify the target function category corresponding to the business process and obtain the reference data corresponding to the target function category;

[0016] Based on the reference data, identify abnormal business data from the business data;

[0017] Based on the abnormal business data and reference data, generate corrected business data corresponding to the abnormal business data.

[0018] In one embodiment, the method further includes, from the functional model pool:

[0019] Multiple correction business data are classified and stored according to their respective functional models;

[0020] If the stored corrected business data corresponding to the target functional model meets the preset conditions, the step of using the corrected business data as training samples is executed.

[0021] In one embodiment, before performing the step of using the modified business data as training samples, provided that the stored modified business data corresponding to the target function category meets preset conditions, the method further includes:

[0022] Obtain the number of training samples contained in the original training set corresponding to the target functional model;

[0023] Based on the stored corrected business data corresponding to the target functional model, and the abnormal business data corresponding to each corrected business data, calculate the error parameters corresponding to each corrected business data.

[0024] Obtain the number of corrected business data stored for the target functional model, and calculate the average error parameter based on the number of corrected data and the error parameter.

[0025] The ratio of the number of corrected data to the number of training samples is obtained. If the ratio is greater than the first threshold and the average error parameter is greater than the second threshold, then the stored corrected business data corresponding to the target functional model is determined to meet the preset conditions.

[0026] In one embodiment, the method further includes:

[0027] Send heartbeat packets to the functional model pool;

[0028] If no communication response is received from the functional model pool in response to the heartbeat packet within a preset time period, multiple heartbeat packets are sent to the functional model pool within a preset period.

[0029] If no communication response is received within the preset period, a restart command is generated to instruct the functional model pool to restart the network.

[0030] In one embodiment, the method further includes:

[0031] The target functional model is randomly initialized from the functional model pool to obtain the original functional model;

[0032] Obtain business data corresponding to the target functional model from the business data generated during the business process operation phase;

[0033] Labels are added to the business data corresponding to the original functional model to obtain training labels. The original functional model is then trained using these training labels to obtain the updated target functional model.

[0034] Secondly, this application also provides a business function update apparatus. The apparatus includes:

[0035] The data acquisition module is used to acquire business data generated during the business process execution phase.

[0036] The acquisition module is used to identify abnormal business data in the business data and obtain the corrected business data corresponding to the abnormal business data.

[0037] The matching module is used to determine the target functional model corresponding to the target functional category to which the abnormal business data belongs from the functional model pool; the functional model pool includes multiple functional models corresponding to different functional categories.

[0038] The update module is used to use the corrected business data as training samples and to incrementally train the target function model using the training samples to obtain the updated target function model.

[0039] Thirdly, this application also provides a business function update system, characterized in that the system includes:

[0040] Business systems are used to generate business function call requests during the business process execution phase.

[0041] The engine system is used to acquire business function call requests during the business process execution phase and identify the target function category corresponding to the business function call request; based on the target function category, it retrieves the target function model corresponding to the target function category from the function model pool; it obtains input data from the business function call request, calls the target function model to process the input data, obtains the business data output by the target function model, and transmits the business data to the business system and the audit system;

[0042] The audit system is used to identify the target function category corresponding to the business process and obtain the reference data corresponding to the target function category; based on the reference data, it identifies abnormal business data from the business data; based on the abnormal business data and the reference data, it generates corrected business data corresponding to the abnormal business data and transmits the corrected business data to the engine system.

[0043] A storage system is used to store multiple correction business data.

[0044] The functional model pool is used to determine the target functional model corresponding to the target functional category to which the abnormal business data belongs; the corrected business data is used as training samples, and the target functional model is incrementally trained using the training samples to obtain the updated target functional model.

[0045] Specifically, when the stored corrected business data corresponding to the target functional model meets the preset conditions, the engine system notifies the functional model pool to perform incremental training on the target functional model.

[0046] In one embodiment, the engine system is also used to classify multiple correction business data in the storage system according to their respective functional models.

[0047] In one embodiment, the engine system is further configured to obtain the number of training samples contained in the original training set corresponding to the target functional model; calculate the error parameter corresponding to each corrected business data based on the stored corrected business data corresponding to the target functional model and the abnormal business data corresponding to each corrected business data; obtain the number of corrected data of the stored corrected business data corresponding to the target functional model, and calculate the average error parameter according to the number of corrected data and the error parameter; obtain the ratio of the number of corrected data to the number of training samples, and if the ratio is greater than a first threshold and the average error parameter is greater than a second threshold, then determine that the stored corrected business data corresponding to the target functional model meets the preset conditions.

[0048] In one embodiment, the engine system is further configured to send heartbeat packets to the functional model pool; if no communication response is received from the functional model pool in response to the heartbeat packets within a preset time period, multiple heartbeat packets are sent to the functional model pool within a preset period; if no communication response is received within the preset period, a restart command is generated to instruct the functional model pool to restart the network.

[0049] In one embodiment, the functional model pool is further used to randomly initialize the target functional model to obtain the original functional model; obtain business data corresponding to the target functional model from the business data generated during the business process operation phase; add labels to the business data corresponding to the original functional model to obtain training labels, and train the original functional model using the training labels to obtain the updated target functional model.

[0050] Fourthly, this application also provides a computer device. The computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to perform the following steps:

[0051] Acquire business data generated during the business process execution phase;

[0052] Identify abnormal business data in the business data and obtain the corrected business data corresponding to the abnormal business data;

[0053] From the functional model pool, determine the target functional model corresponding to the target functional category to which the abnormal business data belongs; the functional model pool includes multiple functional models corresponding to different functional categories.

[0054] The corrected business data is used as training samples, and the target function model is incrementally trained using the training samples to obtain the updated target function model.

[0055] Fifthly, this application also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program thereon, which, when executed by a processor, performs the following steps:

[0056] Acquire business data generated during the business process execution phase;

[0057] Identify abnormal business data in the business data and obtain the corrected business data corresponding to the abnormal business data;

[0058] From the functional model pool, determine the target functional model corresponding to the target functional category to which the abnormal business data belongs; the functional model pool includes multiple functional models corresponding to different functional categories.

[0059] The corrected business data is used as training samples, and the target function model is incrementally trained using the training samples to obtain the updated target function model.

[0060] Sixthly, this application also provides a computer program product. The computer program product includes a computer program that, when executed by a processor, performs the following steps:

[0061] Acquire business data generated during the business process execution phase;

[0062] Identify abnormal business data in the business data and obtain the corrected business data corresponding to the abnormal business data;

[0063] From the functional model pool, determine the target functional model corresponding to the target functional category to which the abnormal business data belongs; the functional model pool includes multiple functional models corresponding to different functional categories.

[0064] The corrected business data is used as training samples, and the target function model is incrementally trained using the training samples to obtain the updated target function model.

[0065] The aforementioned business function update methods, devices, systems, computer equipment, storage media, and computer program products acquire business data generated during the business process execution phase; identify abnormal business data within the business data; and obtain the corresponding corrected business data. This enables real-time detection and correction of abnormal business data. From the functional model pool, a target functional model corresponding to the target functional category of the abnormal business data is determined. Furthermore, based on the target functional category of the abnormal business data, the corresponding functional model is automatically searched. The corrected business data is used as training samples, and the target functional model is incrementally trained using these training samples to obtain the updated target functional model. This system can automatically detect abnormal business data during the business process execution phase, automatically generate training samples based on the abnormal business data, match the corresponding target functional model, and then automatically perform incremental training on the target functional model using the training samples, completing a model update. The entire process eliminates the need for manual updates of the functional model by technical personnel and avoids external transmission of business data, thus improving the efficiency of functional model updates and data security in the business system. Attached Figure Description

[0066] Figure 1 This is a flowchart illustrating a business function update method in one embodiment;

[0067] Figure 2 This is a flowchart illustrating the process of determining whether a preset condition is met in one embodiment.

[0068] Figure 3This is a schematic diagram of a process for delaying the transmission of heartbeat packets in one embodiment;

[0069] Figure 4 This is a flowchart illustrating a business function update method in one embodiment;

[0070] Figure 5 This is a schematic diagram of the structure of an AI function self-updating system in one embodiment;

[0071] Figure 6 This is a functional classification diagram of the AI ​​model pool in one embodiment;

[0072] Figure 7 This is a schematic diagram of the AI ​​Engine update cycle in one embodiment;

[0073] Figure 8 This is a schematic diagram of the AI ​​Engine structure in one embodiment;

[0074] Figure 9 This is a schematic diagram of the AI ​​Engine data flow in one embodiment;

[0075] Figure 10 This is a flowchart illustrating the AI ​​Monitor in one embodiment;

[0076] Figure 11 This is a schematic diagram illustrating the communication between the AI ​​Monitor and the AI ​​model pool in one embodiment;

[0077] Figure 12 This is a schematic diagram of the distributed storage structure of a storage system in one embodiment;

[0078] Figure 13 This is a flowchart illustrating the manual review process in one embodiment;

[0079] Figure 14 This is a schematic diagram illustrating two deployment methods of a manual review system in one embodiment;

[0080] Figure 15 This is a schematic diagram of a semi-automatic update process for an AI model in one embodiment;

[0081] Figure 16 This is a schematic diagram of the physical deployment of an AI function self-updating system in one embodiment;

[0082] Figure 17 This is a structural block diagram of a service function update system in one embodiment;

[0083] Figure 18 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0084] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0085] In one embodiment, such as Figure 1 As shown, a method for updating business functions is provided. This embodiment illustrates the application of this method to a computer device, which can be a terminal or a server. The terminal can be, but is not limited to, various personal computers, laptops, smartphones, tablets, IoT devices, and portable wearable devices. IoT devices can be smart speakers, smart TVs, smart air conditioners, smart medical devices, etc. Portable wearable devices can be smartwatches, smart bracelets, head-mounted devices, etc. The server can be a standalone server or a server cluster consisting of multiple servers. In this embodiment, the method includes the following steps:

[0086] Step 102: Obtain business data generated during the business process execution phase.

[0087] In this context, a business process refers to the data processing procedure of calling a functional model to process input data, obtain model output data, and then reviewing and correcting the output data. For example, business processes in the medical field include, but are not limited to, business processes that utilize preoperative data from patient data, such as medical images and laboratory results, for preoperative processing (e.g., bone segmentation, knee target detection, disease category diagnosis, etc.); and business processes that utilize surgical operation data implemented on patients from patient data for intraoperative processing (e.g., deep learning-based robotic arm control, etc.).

[0088] Optionally, for a business process, the computer device obtains the output data of the functional model called by the business process as the business data corresponding to the business process.

[0089] In one feasible implementation, for a business process, the computer device obtains the input data and output data of the functional model called by the business process, which together serve as the business data corresponding to the business process.

[0090] Step 104: Identify abnormal business data in the business data and obtain the corrected business data corresponding to the abnormal business data.

[0091] Abnormal business data refers to business data that is marked as changed or abandoned during the business process execution phase. Correcting business data includes, but is not limited to, deleting, modifying, and adding information related to abnormal business data.

[0092] Optionally, the computer equipment audits the business data, pre-configuring different audit standards according to different business types, and determines the accuracy of the business data based on the corresponding audit standards. If a piece of business data is inaccurate, an exception tag is added to mark the business data as abnormal, and corrected business data after modification is obtained.

[0093] In one feasible implementation, after collecting business data, the data is transmitted to the corresponding review system for review by designated personnel. For example, the reviewer could be a doctor responsible for processing hospital business data. The doctor reviews the data on the review system to determine its accuracy. If a piece of data is inaccurate, an anomaly tag is added, marking it as abnormal, and the abnormal data is corrected and entered into the system.

[0094] Step 106: Determine the target functional model corresponding to the target functional category to which the abnormal business data belongs from the functional model pool; the functional model pool includes multiple functional models corresponding to different functional categories.

[0095] In this context, functional categories are labels used to characterize the data processing capabilities of a functional model, with a one-to-one correspondence between functional categories and functional models. Functional categories include, but are not limited to, segmentation categories, detection categories, and classification categories. For example, in the medical field, functional categories include, but are not limited to, bone segmentation categories, organ segmentation categories, key point detection categories, joint detection categories, and disease classification categories.

[0096] Optionally, the computer equipment determines the target function type to which the abnormal business data belongs based on its source. Specifically, the function model pool includes multiple function models, each corresponding to a function category label. Since abnormal business data originates from business data, which is the output of the function model after processing input data, the generated abnormal business data will have a matching function category label with the function model. After identifying the abnormal business data, the computer can determine its target function category based on the function category label and then find the target function model that matches the target function category from the function model pool.

[0097] Step 108: Use the corrected business data as training samples, and incrementally train the target function model using the training samples to obtain the updated target function model.

[0098] Optionally, the computer device uses a piece of corrected business data as a training sample and adds the same target function category as the abnormal business data to the training sample. This target function category is also the same as the target function model. The target function model in the function model pool is incrementally trained using the training sample to obtain the updated target function model, thus completing one update of the target function model. In subsequent business process execution phases, if abnormal business data corresponding to the target function model is identified again, steps 104 to 108 are repeated to update the updated target function model again.

[0099] The aforementioned business function update method involves acquiring business data generated during the business process execution phase; identifying abnormal business data within the business data; and obtaining the corresponding corrected business data. This allows for real-time detection and correction of abnormal business data. From the functional model pool, a target functional model corresponding to the target functional category of the abnormal business data is determined. The functional model pool includes multiple functional models corresponding to different functional categories. The method automatically searches for the functional model corresponding to the target functional category of the abnormal business data. The corrected business data is used as training samples, and the target functional model is incrementally trained using these samples to obtain the updated target functional model. This method automatically detects abnormal business data during the business process execution phase, automatically generates training samples based on the abnormal business data, matches the corresponding target functional model, and then automatically performs incremental training on the target functional model using the training samples, completing a model update. The entire process eliminates the need for manual updates of the functional model by technical personnel and avoids external transmission of business data, thus improving the efficiency of functional model updates and data security in the business system.

[0100] In one embodiment, obtaining business data generated during the business process execution phase includes: obtaining business function call requests during the business process execution phase and identifying the target function category corresponding to the business function call request; based on the target function category, obtaining the target function model corresponding to the target function category from the function model pool; obtaining input data from the business function call request and calling the target function model to process the input data to obtain the business data output by the target function model.

[0101] Optionally, after receiving a business function call request during the business process execution phase, the computer device identifies the target function category corresponding to the business function call request, calls the target function model corresponding to the target function category from the function model pool, processes the input data in the business function call request using the target function model, and obtains the business data generated during this business process execution phase.

[0102] In this embodiment, business function call requests during the business process execution phase are obtained, and the target function category corresponding to the business function call request is identified. Based on the target function category, the target function model corresponding to the target function category is obtained from the function model pool. Input data is obtained from the business function call request, and the target function model is called to process the input data to obtain the business data output by the target function model. This allows for automatic matching of the appropriate function model to process the business function call request and complete the business processing flow.

[0103] In one embodiment, identifying abnormal business data in business data and obtaining corrected business data corresponding to the abnormal business data includes: identifying a target function category corresponding to the business process and obtaining reference data corresponding to the target function category; identifying abnormal business data from the business data based on the reference data; and generating corrected business data corresponding to the abnormal business data based on the abnormal business data and the reference data.

[0104] Optionally, the computer equipment identifies the target function category corresponding to each business process and obtains reference data corresponding to different target function categories. Using this reference data as a review standard, it determines whether each piece of business data matches the corresponding reference data. If a piece of business data does not match the corresponding reference data, it is marked as abnormal business data. Based on the reference data corresponding to the abnormal business data, corrected business data is generated. Alternatively, the doctor reviews the business data corresponding to each business process to determine its accuracy. If a piece of business data is inaccurate, an abnormal tag is added, marking the business data as abnormal, and the current abnormal business data is modified, with corrected business data entered.

[0105] In this embodiment, the target function category corresponding to the business process is identified, and reference data corresponding to the target function category is obtained; based on the reference data, abnormal business data is identified from the business data; based on the abnormal business data and the reference data, corrected business data corresponding to the abnormal business data is generated. This allows for the automatic identification and correction of abnormal business data from the business data, resulting in corrected business data.

[0106] In one embodiment, the method further includes: classifying and storing multiple correction business data according to their respective corresponding functional models; and, if the stored correction business data corresponding to the target functional model meets preset conditions, performing the step of using the correction business data as training samples.

[0107] Optionally, for each piece of corrective business data acquired, the computer device determines the corresponding functional model based on the functional category of the corrective business data (i.e., the functional category of the abnormal business data before correction). Multiple pieces of corrective business data corresponding to the same functional model are stored together, while multiple pieces of corrective business data corresponding to different functional models are stored separately. Essentially, for each functional model, the computer device configures a separate storage node for each functional model, and each storage node contains multiple training samples (corrective business data) corresponding to its respective functional model.

[0108] In this embodiment, multiple corrected service data are categorized and stored according to their respective functional models. When the stored corrected service data corresponding to the target functional model meets preset conditions, the step of using the corrected service data as training samples is executed. This allows for the categorization and storage of corrected service data according to functional models, facilitating subsequent incremental training for each functional model.

[0109] In one embodiment, before executing the step of using the corrected business data as training samples, if the stored corrected business data corresponding to the target function category meets preset conditions, the method further includes: obtaining the number of training samples contained in the original training set corresponding to the target function model; calculating the error parameter corresponding to each corrected business data based on the stored corrected business data corresponding to the target function model and the abnormal business data corresponding to each corrected business data; obtaining the number of corrected data in the stored corrected business data corresponding to the target function model, and calculating the average error parameter based on the number of corrected data and the error parameter; obtaining the ratio of the number of corrected data to the number of training samples, and determining that the stored corrected business data corresponding to the target function model meets preset conditions if the ratio is greater than a first threshold and the average error parameter is greater than a second threshold.

[0110] Optional, such as Figure 2 As shown, after storing a corrected business data corresponding to a target functional model, the computer device counts the stored corrected business data N, calculates the average error σ between the corrected data and the prediction result, and obtains the number of training samples M contained in the original training set of the target functional model. It then determines whether σ and N / M both exceed 0.01. If both σ and N / M exceed 0.01, it is determined that the stored corrected business data corresponding to the target functional model meets the preset conditions, and incremental training is performed to update the model using the stored corrected business data corresponding to the target functional model. Otherwise, it returns to the first step and continues to store the stored corrected business data corresponding to the target functional model.

[0111] Specifically, the calculation of the average error is related to the corresponding target function model, which is σ=L(x_i,y_i), where x_i is the corrected business data, y_i is the abnormal business data, and L is the distance metric function. Different L functions are defined according to different target function models. L functions include, but are not limited to, the L2 distance for calculating the distance between coordinate points, the intersection-union ratio between two masks in semantic segmentation tasks, and the L1 distance between two one-hot codes in classification tasks.

[0112] In this embodiment, the number of training samples in the original training set corresponding to the target functional model is obtained; based on the stored corrected business data corresponding to the target functional model and the abnormal business data corresponding to each corrected business data, the error parameter corresponding to each corrected business data is calculated; the number of corrected data in the stored corrected business data corresponding to the target functional model is obtained, and the average error parameter is calculated according to the number of corrected data and the error parameter; the ratio of the number of corrected data to the number of training samples is obtained, and if the ratio is greater than a first threshold and the average error parameter is greater than a second threshold, then it is determined that the stored corrected business data corresponding to the target functional model meets the preset conditions. This allows for automatic determination of whether the target functional model meets the update conditions, and if the update conditions are met, automatic use of the stored corrected business data corresponding to the target functional model as training samples for incremental training of the target functional model.

[0113] In one embodiment, the method further includes: sending a heartbeat packet to the functional model pool; if no communication response is received from the functional model pool in response to the heartbeat packet within a preset time period, sending multiple heartbeat packets to the functional model pool within a preset period; if no communication response is received within the preset period, generating a restart command to instruct the functional model pool to restart the network.

[0114] Optional, by Figure 3 As shown, if packet loss occurs during communication between the computer device and the functional model pool, the functional model pool service is not immediately activated. Instead, retransmissions are performed at a series of time intervals (1s, 2s, 4s, 8s, 16s). The functional model pool service is only restarted when no response data is received from all retransmitted heartbeat packets. This is because the lack of packet reception could be due to network instability causing delays, or it could be because the functional model pool service is temporarily blocked while processing a task. Therefore, a certain delay in sending heartbeat packets is used to rule out special cases.

[0115] In this embodiment, a heartbeat packet is sent to the functional model pool. If no communication response is received from the functional model pool within a preset time period, multiple heartbeat packets are sent to the functional model pool within a preset period. If no communication response is received within the preset period, a restart command is generated to instruct the functional model pool to restart the network. This allows for real-time monitoring of the communication status between the computer device and the functional model pool, and the elimination of special cases by configuring a delayed heartbeat packet transmission, thus accurately determining the communication status between the computer device and the functional model pool.

[0116] In one embodiment, a business function update method includes:

[0117] Obtain business function call requests during the business process execution phase and identify the target function category corresponding to the business function call request; based on the target function category, obtain the target function model corresponding to the target function category from the function model pool; obtain input data from the business function call request, and call the target function model to process the input data to obtain the business data output by the target function model.

[0118] Identify the target function category corresponding to the business process and obtain the reference data corresponding to the target function category; based on the reference data, identify abnormal business data from the business data; based on the abnormal business data and the reference data, generate corrective business data corresponding to the abnormal business data.

[0119] From the functional model pool, determine the target functional model corresponding to the target functional category to which the abnormal business data belongs; the functional model pool includes multiple functional models corresponding to different functional categories; classify and store multiple corrected business data according to their respective corresponding functional models.

[0120] Obtain the number of training samples in the original training set corresponding to the target functional model; calculate the error parameter corresponding to each corrected business data based on the stored corrected business data corresponding to the target functional model and the abnormal business data corresponding to each corrected business data; obtain the number of corrected data in the stored corrected business data corresponding to the target functional model, and calculate the average error parameter based on the number of corrected data and the error parameter; obtain the ratio of the number of corrected data to the number of training samples, and if the ratio is greater than the first threshold and the average error parameter is greater than the second threshold, then determine that the stored corrected business data corresponding to the target functional model meets the preset conditions.

[0121] If the stored corrected business data corresponding to the target functional model meets the preset conditions, the corrected business data is used as a training sample, and the target functional model is incrementally trained using the training sample to obtain the updated target functional model.

[0122] During the above process, heartbeat packets are sent to the functional model pool in real time; if no communication response is received from the functional model pool in response to the heartbeat packets within a preset time period, multiple heartbeat packets are sent to the functional model pool within a preset period; if no communication response is received within the preset period, a restart command is generated to instruct the functional model pool to restart the network.

[0123] In addition to the above process, the target functional model can be randomly initialized directly from the functional model pool to obtain the original functional model; business data corresponding to the target functional model can be obtained from the business data generated during the business process operation phase; labels can be added to the business data corresponding to the original functional model to obtain training labels, and the original functional model can be trained using the training labels to obtain the updated target functional model.

[0124] In one embodiment, such as Figure 4 As shown, a business function update method is applied to, for example, Figure 5 Taking the AI ​​function update system shown as an example, the methods include:

[0125] The business system initiates a business function call request to the AI ​​Engine through the AI ​​interface. The AI ​​Engine obtains the business function call request in the business process operation phase and identifies the target function category corresponding to the business function call request. Based on the target function category, it obtains the target AI model corresponding to the target function category from the AI ​​model pool. It obtains the input data from the business function call request and calls the target AI model to process the input data to obtain the business data output by the target AI model.

[0126] The AI ​​Monitor in the AI ​​Engine sends heartbeat packets to the AI ​​model pool in real time. If no communication response is received from the AI ​​model pool in response to the heartbeat packet within a preset time period, multiple heartbeat packets are sent to the AI ​​model pool within a preset period. If no communication response is received within the preset period, a restart command is generated to instruct the AI ​​model pool to restart the network.

[0127] AI Engine transmits business data to the business system, which in turn transmits it to the human review system. The human review system identifies the target function category corresponding to the business process and obtains reference data for that category. Based on the reference data, it identifies anomalous business data from the business data. Then, based on the anomalous business data and the reference data, it generates corrected business data corresponding to the anomalous data. The human review system transmits the corrected business data to AI Engine.

[0128] AI Engine determines the target AI model corresponding to the target function category of the abnormal business data from the AI ​​model pool; the AI ​​model pool includes multiple AI models corresponding to different function categories; and the multiple corrected business data are classified and stored in the storage system according to their respective corresponding AI models.

[0129] The AI ​​Monitor in AI Engine obtains the number of training samples contained in the original training set corresponding to the target AI model; based on the stored corrected business data corresponding to the target AI model and the abnormal business data corresponding to each corrected business data, it calculates the error parameters corresponding to each corrected business data; it obtains the number of corrected data of the stored corrected business data corresponding to the target AI model, and calculates the average error parameter based on the number of corrected data and the error parameters; it obtains the ratio of the number of corrected data to the number of training samples, and if the ratio is greater than the first threshold and the average error parameter is greater than the second threshold, it determines that the stored corrected business data corresponding to the target AI model meets the preset conditions.

[0130] If the stored corrected business data corresponding to the target AI model meets the preset conditions, the AI ​​model pool retrieves the stored corrected business data corresponding to the target AI model from the storage system, uses the corrected business data as training samples, and incrementally trains the target AI model using the training samples to obtain the updated target AI model.

[0131] Specifically, in the aforementioned AI function self-updating system:

[0132] The AI ​​model pool is equivalent to the functional model pool, used to provide AI functions (equivalent to functional models) for various business systems, such as semantic segmentation networks for organ segmentation, or classification networks for lesion site examination.

[0133] The AI ​​model pool provides the actual AI functionality for the entire system. The pool offers a unified API to the outside world and employs a layered architecture, categorized according to the different functions it provides. For example... Figure 6 As shown, AI is divided into three main modules based on its functions: segmentation network, detection network, and classification network. Each module is further subdivided; for example, the segmentation network includes a bone segmentation network that provides 3D reconstruction of bone models.

[0134] like Figure 7 As shown, the update cycle of the AI ​​model pool is divided into two main tasks:

[0135] 1. For prediction tasks, the system obtains the requested data and task type from the AI ​​Engine, loads the corresponding latest model, executes the prediction task, and returns the prediction results. To reduce unnecessary system I / O, the AI ​​model file can be loaded only after a model update or upon the first system startup, and then used directly during prediction.

[0136] 2. Update Task: AI Engine initiates an update task based on the data returned by the review system and the model type. The model undergoes incremental training, and after training, it outputs the latest model parameters for use in the next prediction task. The update task is only initiated when the amount of incorrectly predicted data reaches a certain threshold; it does not need to be executed after every review.

[0137] AI Engine, the central node of the system, is used to provide AI function interfaces to various business systems and receive corrected data returned by the manual review system. Internally, it is used to call AI functions and use the received corrected data to make incremental training condition judgments for the AI ​​model.

[0138] like Figure 8 As shown, AI Engine consists of two main parts: the server and the client. The server receives function requests from the client, distributes these requests through AI Monitor, and invokes AI models for prediction. The client is deployed in the business system, responsible for communicating with the server, requesting various services, and displaying the results. Depending on the business system, the client can be integrated as a plugin.

[0139] like Figure 9 As shown, the AI ​​Engine data stream is divided into two parts.

[0140] 1. Data Prediction Data Stream: The AI ​​Engine receives requests from the client, calls the corresponding AI model based on the request, and returns the prediction results to the client.

[0141] 2. Model update data stream: Collect correction results from domain experts. Once a certain number of corrections are collected, the model update process is initiated, and the new training results are saved.

[0142] The AI ​​Monitor in AI Engine is a core component of the system, responsible for calling AI models during the data prediction phase; monitoring whether the AI ​​model pool remains online; and initiating incremental training tasks. For example... Figure 10 The diagram shows a flowchart of the data prediction phase, where data is obtained from the AI ​​interface, distributed according to category, and invoked by the model. For example... Figure 11As shown, AIMonitor communicates with the AI ​​model pool via heartbeat packets. Once the packet loss rate exceeds a certain number of times, the AI ​​model pool is restarted, thereby ensuring that the business system remains online when AI function requests are made.

[0143] A storage system provides the necessary data storage functionality to all functional systems; for example, a DICOM gateway is typically a distributed storage system. Figure 12 As shown, the storage system uses distributed storage, with the underlying data managed by the Data Server, providing services to the outside world. Alternatively, it can be deployed without a separate server and directly integrated with existing storage resources.

[0144] Business systems are systems used to directly run business processes, such as surgical planning software and auxiliary diagnostic software.

[0145] The manual review system is used to review the prediction results of the AI ​​model and the feedback from business system users on the prediction results. It corrects erroneous predictions (equivalent to abnormal business data) and notifies the AI ​​Engine of correct results (equivalent to corrected business data). It also adds erroneous prediction cases (containing correct results) to the incremental training dataset. The manual review process is as follows: Figure 13 As shown, the review system obtains the prediction results from the AI ​​model, which are then reviewed by domain experts. During the review, any unqualified prediction results are directly modified online, and the results are then submitted to the AI ​​Engine for saving.

[0146] There are two deployment methods for the manual review system, such as Figure 14 As shown, there are two deployment methods: one is a standalone deployment independent of the business system. This method offers high flexibility, as it operates independently of the business system, requiring only minor modifications to the original business system for integration. The other method is a plug-in deployment, which integrates seamlessly with the original system. This is particularly beneficial when users themselves can act as auditing experts, providing a user-friendly experience. However, this requires some modifications to the original system.

[0147] This system can also perform semi-automatic training and updating of AI models directly from the AI ​​model pool, such as... Figure 15 As shown, the functional model is randomly initialized to obtain the original functional model. The original functional model is then gradually trained through the use of the business system. This approach is typically used in scenarios with strict data requirements, where AI functionality needs to be deployed within the hospital premises.

[0148] During the business process execution phase of the business system, business data is continuously collected. All business data is treated as anomalous data, and a label is added to each piece of business data (equivalent to correcting the business data). The tagged business data is then stored. When the amount of stored business data corresponding to the original functional model reaches a certain quantity, the stored business data and labels are used to train the functional model, outputting a preliminary model. This process is repeated until the output of the functional model meets the business requirements. This training method places certain demands on the model; overly complex models cannot be used, otherwise, insufficient initial data will cause the model to fail to converge, resulting in training failure. In a feasible implementation, the physical deployment of the AI ​​function self-updating system is as follows: Figure 16 As shown, the system is deployed within the hospital, with all systems connected via a high-speed network. Within the system, deployments can be on the same server or separate, depending on the specific needs. For example, if the number of AI models is small, the specific models and interfaces can be deployed on the same server. Whether the business system and the review system are deployed on the same server depends on the system size and actual circumstances. For example, if the users of the business system are experts in the field, the review function can be integrated into the business system, allowing for real-time adjustments to prediction results during user interaction.

[0149] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0150] Based on the same inventive concept, this application also provides a business function update apparatus for implementing the business function update method described above. The solution provided by this apparatus is similar to the implementation scheme described in the above method; therefore, the specific limitations in one or more business function update apparatus embodiments provided below can be found in the limitations of the business function update method described above, and will not be repeated here.

[0151] In one embodiment, such as Figure 17 As shown, a service function update device 1700 is provided, including: a collection module 1701, an acquisition module 1702, a matching module 1703, and an update module 1704, wherein:

[0152] The data acquisition module 1701 is used to acquire business data generated during the business process execution phase.

[0153] The acquisition module 1702 is used to identify abnormal business data in the business data and acquire the corrected business data corresponding to the abnormal business data.

[0154] The matching module 1703 is used to determine the target functional model corresponding to the target functional category to which the abnormal business data belongs from the functional model pool; the functional model pool includes multiple functional models corresponding to different functional categories.

[0155] The update module 1704 is used to use the corrected business data as training samples and to incrementally train the target function model using the training samples to obtain the updated target function model.

[0156] In one embodiment, the acquisition module 1701 is further configured to acquire business function call requests during the business process operation phase and identify the target function category corresponding to the business function call request; based on the target function category, acquire the target function model corresponding to the target function category from the function model pool; acquire input data from the business function call request and call the target function model to process the input data to obtain the business data output by the target function model.

[0157] In one embodiment, the acquisition module 1702 is further configured to identify the target function category corresponding to the business process and acquire reference data corresponding to the target function category; identify abnormal business data from the business data based on the reference data; and generate corrected business data corresponding to the abnormal business data based on the abnormal business data and the reference data.

[0158] In one embodiment, the matching module 1703 is further configured to classify and store multiple corrected business data according to their respective corresponding functional models; and, if the stored corrected business data corresponding to the target functional model meets the preset conditions, execute the step of using the corrected business data as training samples.

[0159] In one embodiment, the update module 1704 is further configured to obtain the number of training samples contained in the original training set corresponding to the target functional model; calculate the error parameter corresponding to each corrected business data based on the stored corrected business data corresponding to the target functional model and the abnormal business data corresponding to each corrected business data; obtain the number of corrected data of the stored corrected business data corresponding to the target functional model, and calculate the average error parameter according to the number of corrected data and the error parameter; obtain the ratio of the number of corrected data to the number of training samples, and if the ratio is greater than a first threshold and the average error parameter is greater than a second threshold, then determine that the stored corrected business data corresponding to the target functional model meets the preset conditions.

[0160] In one embodiment, the apparatus further includes:

[0161] The communication module is used to send heartbeat packets to the functional model pool; if no communication response is received from the functional model pool in response to the heartbeat packet within a preset time period, multiple heartbeat packets are sent to the functional model pool within a preset period; if no communication response is received within the preset period, a restart command is generated to instruct the functional model pool to restart the network.

[0162] Each module in the aforementioned business function update device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the operations corresponding to each module.

[0163] In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 18As shown, the computer device includes a processor, memory, input / output interfaces, a communication interface, a display unit, and an input device. The processor, memory, and input / output interfaces are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interfaces. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The input / output interfaces are used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, NFC (Near Field Communication), or other technologies. When the computer program is executed by the processor, it implements a business function update method. The display unit of the computer device is used to form a visually visible image. It can be a display screen, a projection device, or a virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of the computer device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads set on the casing of the computer device, or external keyboards, touchpads, or mice, etc.

[0164] Those skilled in the art will understand that Figure 18 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0165] In one embodiment, a business function update system is provided, characterized in that the system includes:

[0166] Business systems are used to generate business function call requests during the business process execution phase.

[0167] The engine system is used to acquire business function call requests during the business process execution phase and identify the target function category corresponding to the business function call request; based on the target function category, it retrieves the target function model corresponding to the target function category from the function model pool; it obtains input data from the business function call request, calls the target function model to process the input data, obtains the business data output by the target function model, and transmits the business data to the business system and the audit system;

[0168] The audit system is used to identify the target function category corresponding to the business process and obtain the reference data corresponding to the target function category; based on the reference data, it identifies abnormal business data from the business data; based on the abnormal business data and the reference data, it generates corrected business data corresponding to the abnormal business data and transmits the corrected business data to the engine system.

[0169] The storage system is used to store multiple correction business data; when the stored correction business data corresponding to the target functional model meets the preset conditions, the engine system notifies the functional model pool to perform incremental training on the target functional model.

[0170] The functional model pool is used to determine the target functional model corresponding to the target functional category to which the abnormal business data belongs; the corrected business data is used as training samples, and the target functional model is incrementally trained using the training samples to obtain the updated target functional model.

[0171] In one embodiment, the engine system is also used to classify multiple correction business data in the storage system according to their respective functional models.

[0172] In one embodiment, the engine system is further configured to: obtain the number of training samples contained in the original training set corresponding to the target functional model; calculate the error parameter corresponding to each corrected business data based on the stored corrected business data corresponding to the target functional model and the abnormal business data corresponding to each corrected business data; obtain the number of corrected data of the stored corrected business data corresponding to the target functional model, and calculate the average error parameter based on the number of corrected data and the error parameter; obtain the ratio of the number of corrected data to the number of training samples, and determine that the stored corrected business data corresponding to the target functional model meets the preset conditions if the ratio is greater than a first threshold and the average error parameter is greater than a second threshold.

[0173] In one embodiment, the engine system is further configured to send heartbeat packets to the functional model pool; if no communication response is received from the functional model pool in response to the heartbeat packets within a preset time period, multiple heartbeat packets are sent to the functional model pool within a preset period; if no communication response is received within the preset period, a restart command is generated to instruct the functional model pool to restart the network.

[0174] In one embodiment, the functional model pool is also used to randomly initialize the target functional model to obtain the original functional model; obtain business data corresponding to the target functional model from the business data generated during the business process operation phase; add labels to the business data corresponding to the original functional model to obtain training labels, and train the original functional model using the training labels to obtain the updated target functional model.

[0175] In one embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to perform the following steps: acquiring business data generated during the operation phase of a business process; identifying abnormal business data in the business data and acquiring corrected business data corresponding to the abnormal business data; determining a target function model from a function model pool that corresponds to the target function category to which the abnormal business data belongs; the function model pool includes multiple function models corresponding to different function categories; using the corrected business data as training samples, and incrementally training the target function model using the training samples to obtain an updated target function model.

[0176] In one embodiment, when the processor executes the computer program, it further performs the following steps: obtaining a business function call request in the business process execution phase and identifying the target function category corresponding to the business function call request; based on the target function category, obtaining the target function model corresponding to the target function category from the function model pool; obtaining input data from the business function call request and calling the target function model to process the input data to obtain the business data output by the target function model.

[0177] In one embodiment, when the processor executes the computer program, it further performs the following steps: identifying the target function category corresponding to the business process and obtaining reference data corresponding to the target function category; identifying abnormal business data from the business data based on the reference data; and generating corrected business data corresponding to the abnormal business data based on the abnormal business data and the reference data.

[0178] In one embodiment, when the processor executes the computer program, it further performs the following steps: classifying and storing multiple correction business data according to their respective corresponding functional models; and, if the stored correction business data corresponding to the target functional model meets the preset conditions, performing the step of using the correction business data as training samples.

[0179] In one embodiment, when the processor executes the computer program, it further performs the following steps: obtaining the number of training samples contained in the original training set corresponding to the target functional model; calculating the error parameter corresponding to each corrected business data based on the stored corrected business data corresponding to the target functional model and the abnormal business data corresponding to each corrected business data; obtaining the number of corrected data of the stored corrected business data corresponding to the target functional model, and calculating the average error parameter according to the number of corrected data and the error parameter; obtaining the ratio of the number of corrected data to the number of training samples, and determining that the stored corrected business data corresponding to the target functional model meets the preset conditions if the ratio is greater than a first threshold and the average error parameter is greater than a second threshold.

[0180] In one embodiment, when the processor executes the computer program, it further performs the following steps: sending a heartbeat packet to the functional model pool; if no communication response is received from the functional model pool in response to the heartbeat packet within a preset time period, sending multiple heartbeat packets to the functional model pool within a preset period; if no communication response is received within the preset period, generating a restart command to instruct the functional model pool to restart the network.

[0181] In one embodiment, when the processor executes the computer program, it further performs the following steps: randomly initializes the target functional model from the functional model pool to obtain the original functional model; obtains the business data corresponding to the target functional model from the business data generated during the business process operation phase; adds labels to the business data corresponding to the original functional model to obtain training labels, and trains the original functional model using the training labels to obtain the updated target functional model.

[0182] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, it performs the following steps: acquiring business data generated during the business process execution phase; identifying abnormal business data in the business data and acquiring corrected business data corresponding to the abnormal business data; determining, from a functional model pool, a target functional model corresponding to the target functional category to which the abnormal business data belongs; the functional model pool includes multiple functional models corresponding to different functional categories; using the corrected business data as training samples, and incrementally training the target functional model using the training samples to obtain an updated target functional model.

[0183] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: obtaining a business function call request in the business process execution phase and identifying the target function category corresponding to the business function call request; based on the target function category, obtaining the target function model corresponding to the target function category from the function model pool; obtaining input data from the business function call request and calling the target function model to process the input data to obtain the business data output by the target function model.

[0184] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: identifying the target function category corresponding to the business process and obtaining reference data corresponding to the target function category; identifying abnormal business data from the business data based on the reference data; and generating corrected business data corresponding to the abnormal business data based on the abnormal business data and the reference data.

[0185] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: classifying and storing multiple correction business data according to their respective corresponding functional models; and, if the stored correction business data corresponding to the target functional model meets the preset conditions, performing the step of using the correction business data as training samples.

[0186] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: obtaining the number of training samples contained in the original training set corresponding to the target functional model; calculating the error parameter corresponding to each corrected business data based on the stored corrected business data corresponding to the target functional model and the abnormal business data corresponding to each corrected business data; obtaining the number of corrected data of the stored corrected business data corresponding to the target functional model, and calculating the average error parameter according to the number of corrected data and the error parameter; obtaining the ratio of the number of corrected data to the number of training samples, and determining that the stored corrected business data corresponding to the target functional model meets the preset conditions if the ratio is greater than a first threshold and the average error parameter is greater than a second threshold.

[0187] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: sending a heartbeat packet to the functional model pool; if no communication response is received from the functional model pool in response to the heartbeat packet within a preset time period, sending multiple heartbeat packets to the functional model pool within a preset period; if no communication response is received within the preset period, generating a restart command to instruct the functional model pool to restart the network.

[0188] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: randomly initializes the target functional model from the functional model pool to obtain the original functional model; obtains the business data corresponding to the target functional model from the business data generated during the business process operation phase; adds labels to the business data corresponding to the original functional model to obtain training labels, and trains the original functional model using the training labels to obtain the updated target functional model.

[0189] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, performs the following steps:

[0190] Acquire business data generated during the business process execution phase; identify abnormal business data in the business data and obtain the corrected business data corresponding to the abnormal business data; determine the target function model corresponding to the target function category to which the abnormal business data belongs from the function model pool; the function model pool includes multiple function models corresponding to different function categories; use the corrected business data as training samples, and incrementally train the target function model using the training samples to obtain the updated target function model.

[0191] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: obtaining a business function call request in the business process execution phase and identifying the target function category corresponding to the business function call request; based on the target function category, obtaining the target function model corresponding to the target function category from the function model pool; obtaining input data from the business function call request and calling the target function model to process the input data to obtain the business data output by the target function model.

[0192] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: identifying the target function category corresponding to the business process and obtaining reference data corresponding to the target function category; identifying abnormal business data from the business data based on the reference data; and generating corrected business data corresponding to the abnormal business data based on the abnormal business data and the reference data.

[0193] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: classifying and storing multiple correction business data according to their respective corresponding functional models; and, if the stored correction business data corresponding to the target functional model meets the preset conditions, performing the step of using the correction business data as training samples.

[0194] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: obtaining the number of training samples contained in the original training set corresponding to the target functional model; calculating the error parameter corresponding to each corrected business data based on the stored corrected business data corresponding to the target functional model and the abnormal business data corresponding to each corrected business data; obtaining the number of corrected data of the stored corrected business data corresponding to the target functional model, and calculating the average error parameter according to the number of corrected data and the error parameter; obtaining the ratio of the number of corrected data to the number of training samples, and determining that the stored corrected business data corresponding to the target functional model meets the preset conditions if the ratio is greater than a first threshold and the average error parameter is greater than a second threshold.

[0195] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: sending a heartbeat packet to the functional model pool; if no communication response is received from the functional model pool in response to the heartbeat packet within a preset time period, sending multiple heartbeat packets to the functional model pool within a preset period; if no communication response is received within the preset period, generating a restart command to instruct the functional model pool to restart the network.

[0196] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: randomly initializes the target functional model from the functional model pool to obtain the original functional model; obtains the business data corresponding to the target functional model from the business data generated during the business process operation phase; adds labels to the business data corresponding to the original functional model to obtain training labels, and trains the original functional model using the training labels to obtain the updated target functional model.

[0197] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data shall comply with the relevant laws, regulations and standards of the relevant countries and regions.

[0198] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0199] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0200] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for updating business functions, characterized in that, The method includes: Acquire business data generated during the operation phase of the business process; the business process includes a business process for preoperative processing using preoperative data in the patient data, and a business process for intraoperative processing using surgical operation data implemented on the patient in the patient data. Identify abnormal business data in the business data and obtain the corrected business data corresponding to the abnormal business data; From the functional model pool, determine the target functional model corresponding to the target functional category to which the abnormal business data belongs; the functional model pool includes multiple functional models corresponding to different functional categories. The corrected business data is used as a training sample, and the target function model is incrementally trained using the training sample to obtain the updated target function model.

2. The method according to claim 1, characterized in that, The acquisition of business data generated during the business process execution phase includes: Obtain business function call requests during the business process execution phase, and identify the target function category corresponding to the business function call request; Based on the target function category, obtain the target function model corresponding to the target function category from the function model pool; The system obtains input data from the business function call request, and calls the target function model to process the input data to obtain the business data output by the target function model.

3. The method according to claim 1, characterized in that, The step of identifying abnormal business data in the business data and obtaining the corrected business data corresponding to the abnormal business data includes: Identify the target function category corresponding to the business process, and obtain the reference data corresponding to the target function category; Based on the reference data, abnormal business data is identified from the business data; Based on the abnormal business data and the reference data, corrected business data corresponding to the abnormal business data is generated.

4. The method according to claim 1, characterized in that, The method further includes: Multiple correction business data are classified and stored according to their respective functional models; If the stored corrected business data corresponding to the target functional model meets the preset conditions, the step of using the corrected business data as training samples is executed.

5. The method according to claim 4, characterized in that, Before executing the step of using the corrected service data as training samples when the stored corrected service data corresponding to the target function category meets preset conditions, the method further includes: Obtain the number of training samples contained in the original training set corresponding to the target functional model; Based on the stored corrected service data corresponding to the target functional model, and the abnormal service data corresponding to each corrected service data, calculate the error parameters corresponding to each corrected service data. Obtain the number of corrected data points of the stored corrected business data corresponding to the target functional model, and calculate the average error parameter based on the number of corrected data points and the error parameter; If the ratio of the number of corrected data to the number of training samples is obtained, and the ratio is greater than a first threshold and the average error parameter is greater than a second threshold, then it is determined that the stored corrected business data corresponding to the target functional model meets the preset conditions.

6. The method according to any one of claims 1 to 5, characterized in that, The method further includes: Send heartbeat packets to the functional model pool; If no communication response is received from the functional model pool in response to the heartbeat packet within a preset time period, multiple heartbeat packets are sent to the functional model pool within a preset period. If no communication response is received within the preset period, a restart command is generated to instruct the functional model pool to restart the network.

7. The method according to claim 1, characterized in that, The method further includes: The target functional model is randomly initialized from the functional model pool to obtain the original functional model; Obtain business data corresponding to the target functional model from the business data generated during the business process operation phase; Labels are added to the business data corresponding to the original functional model to obtain training labels, and the original functional model is trained using the training labels to obtain the updated target functional model.

8. A business function update device, characterized in that, The device includes: The data acquisition module is used to acquire business data generated during the operation phase of the business process; the business process includes a business process for preoperative processing using preoperative data in the patient data, and a business process for intraoperative processing using surgical operation data implemented on the patient in the patient data. The acquisition module is used to identify abnormal business data in the business data and acquire the corrected business data corresponding to the abnormal business data. The matching module is used to determine the target functional model corresponding to the target functional category to which the abnormal business data belongs from the functional model pool; the functional model pool includes multiple functional models corresponding to different functional categories respectively; The update module is used to use the corrected business data as training samples and to incrementally train the target function model using the training samples to obtain the updated target function model.

9. A business function update system, characterized in that, The system includes: The business system is used to generate business function call requests during the business process execution phase; the business process includes a business process for preoperative processing using preoperative data in the patient data, and a business process for intraoperative processing using surgical operation data implemented on the patient in the patient data. The engine system is used to acquire business function call requests during the business process execution phase and identify the target function category corresponding to the business function call request; based on the target function category, acquire the target function model corresponding to the target function category from the function model pool; acquire input data from the business function call request, call the target function model to process the input data, obtain the business data output by the target function model, and transmit the business data to the business system and the audit system; The audit system is used to identify target function categories corresponding to business processes and obtain reference data corresponding to the target function categories; based on the reference data, identify abnormal business data from the business data; based on the abnormal business data and the reference data, generate corrected business data corresponding to the abnormal business data, and transmit the corrected business data to the engine system. A storage system is used to store multiple correction business data. A functional model pool is used to determine the target functional model corresponding to the target functional category to which the abnormal business data belongs; the corrected business data is used as a training sample, and the target functional model is incrementally trained using the training sample to obtain the updated target functional model; Specifically, when the stored corrected business data corresponding to the target functional model meets preset conditions, the engine system notifies the functional model pool to perform incremental training on the target functional model.

10. The system according to claim 9, characterized in that, The engine system is also used to classify the multiple correction business data in the storage system according to their respective functional models.

11. The system according to claim 9, characterized in that, The engine system is also used to obtain the number of training samples contained in the original training set corresponding to the target functional model; and to calculate the error parameters corresponding to each corrected business data based on the stored corrected business data corresponding to the target functional model and the abnormal business data corresponding to each corrected business data. Obtain the number of corrected data points of the stored corrected business data corresponding to the target functional model, and calculate the average error parameter based on the number of corrected data points and the error parameter; If the ratio of the number of corrected data to the number of training samples is obtained, and the ratio is greater than a first threshold and the average error parameter is greater than a second threshold, then it is determined that the stored corrected business data corresponding to the target functional model meets the preset conditions.

12. The system according to claim 9, characterized in that, The engine system is also used to send heartbeat packets to the functional model pool; if no communication response is received from the functional model pool in response to the heartbeat packet within a preset time period, multiple heartbeat packets are sent to the functional model pool within a preset period; if no communication response is received within the preset period, a restart command is generated to instruct the functional model pool to restart the network.

13. The system according to claim 9, characterized in that, The functional model pool is also used to randomly initialize the target functional model to obtain the original functional model; obtain business data corresponding to the target functional model from the business data generated during the business process operation phase; add labels to the business data corresponding to the original functional model to obtain training labels, and train the original functional model using the training labels to obtain the updated target functional model.

14. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

15. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images