A dynamic consistency checking system for medical heterogeneous data

Abstract

The application relates to a dynamic consistency verification system for medical heterogeneous data and belongs to the technical field, which solves the problem that consistency verification cannot be efficiently and stably performed in the prior art. The system comprises: a state monitoring module for collecting state information of a source system in real time; a large model interaction module for constructing a prompt word based on the state information, calling a large model inference to obtain a recommended extraction amount, a risk level and a confidence level; an extraction amount determination module for determining whether to perform alarm processing based on the risk level and the confidence level, if yes, sending alarm information to a warning module, otherwise, dynamically determining a final extraction amount according to the recommended extraction amount, the risk level and the confidence level; a heterogeneous data extraction module for extracting corresponding amount of data from a heterogeneous data source according to the determined final extraction amount; a verification module for performing consistency verification on the extracted data; and a warning module for performing alarm based on the risk level and the confidence level. Efficient and stable consistency verification is realized.

Description

Technical Field

[0001] This invention relates to the field of consistency verification technology, and in particular to a dynamic consistency verification system for heterogeneous medical data. Background Technology

[0002] In the process of building a smart healthcare system, various medical institutions have, through years of IT development, formed a diversified data resource system encompassing basic patient data, laboratory test data, medical imaging data, electronic medical records, and drug distribution information. This medical data exhibits significant heterogeneity in terms of type, structure, and storage method: structured diagnostic and treatment data is typically stored in distributed relational databases, semi-structured medical records are stored in distributed document databases, and unstructured medical texts and retrieval data are managed using distributed index databases. This multi-source, heterogeneous storage architecture has led to severe fragmentation of medical data resources, creating data barriers that are difficult to overcome.

[0003] In implementing medical data quality control, cross-system data consistency verification for distributed relational databases, document databases, and indexed databases faces significant challenges. Traditional consistency verification methods typically employ fixed time interval division strategies (such as dividing processing batches by month or year). This coarse-grained batch processing approach reveals significant shortcomings in medical scenarios: First, static time windows cannot accurately reflect the real-time generation patterns of medical data, easily leading to uneven allocation of system resources, resulting in system overload during peak processing periods and resource idleness during off-peak periods. Second, medical operations are characterized by continuous operation and real-time response; fixed batch processing models are ill-suited to the sudden surge in data in scenarios such as emergency rooms and intensive care units, potentially affecting the timeliness of clinical decisions due to data processing delays. Third, medical data has extremely high requirements for accuracy and completeness; traditional methods lack the ability to dynamically perceive the state of the source system during data extraction, easily causing performance fluctuations in the source database and even leading to chain reactions in business systems. Summary of the Invention

[0004] Based on the above analysis, the present invention aims to provide a dynamic consistency verification system for heterogeneous medical data to solve the problem that existing systems cannot perform consistency verification efficiently and stably.

[0005] On one hand, embodiments of the present invention provide a dynamic consistency verification system for heterogeneous medical data, comprising: The status monitoring module is used to collect the status information of the source system in real time; The large model interaction module is used to construct prompt words based on the state information and call the large model inference to obtain the suggested extraction quantity, risk level and confidence level. The extraction quantity determination module is used to determine whether to initiate an alarm based on the risk level and confidence level. If so, an alarm message is sent to the early warning module; otherwise, the final extraction quantity is dynamically determined based on the suggested extraction quantity, risk level, and confidence level. The heterogeneous data extraction module is used to connect to heterogeneous data sources and extract a corresponding amount of data from the heterogeneous data sources according to the determined final extraction amount. The verification module is used to perform consistency verification on the extracted data. The early warning module is used to issue alerts based on risk level and confidence level.

[0006] Based on further improvements to the above system, the extraction quantity determination module includes a decision tree model, a judgment module, a decay calculation module, and an optimization module. The decision tree model is used to determine the current decision based on risk level and confidence level; The judgment module is used to call the early warning module, the attenuation calculation module, or the optimization module based on the current decision. The attenuation calculation module is used to perform attenuation calculation on the suggested extraction amount to obtain the final extraction amount; The optimization module is used to optimize and adjust the suggested extraction amount to obtain the final extraction amount.

[0007] Based on further improvements to the above system, the attenuation calculation module attenuates the suggested extraction amount to obtain the final extraction amount in the following manner: ; ; in, Indicates the final extraction amount. Indicates the suggested extraction amount. Indicates the attenuation coefficient. Indicates the confidence level. The confidence decay function is represented by the function. The weights representing the confidence level. This represents the real-time risk decay value. Indicates real-time risk weights. This represents the risk decay value based on historical experience. This represents the risk weight based on historical experience.

[0008] Based on further improvements to the above system, the attenuation calculation module obtains the real-time risk attenuation value using the following method: Calculate the risk score for each key indicator based on the current status information; The real-time risk decay value is obtained based on the risk score of each key indicator.

[0009] Based on further improvements to the above system, the attenuation calculation module obtains the attenuation value of historical experience risk using the following method: Filter high-risk operations whose similarity to the current status information exceeds a preset threshold; Calculate the average similarity between the selected high-risk operations and the current state information; The historical experience risk decay value is obtained based on the average similarity and the maximum historical risk decay ratio.

[0010] Based on further improvements to the above system, the confidence decay function is expressed as: ; in, This indicates the maximum decay rate of the confidence level.

[0011] Based on further improvements to the above system, the large model interaction module obtains the confidence score in the following way: The variance is calculated based on the probability of each type of extraction predicted by the large model; The confidence level is calculated based on the variance.

[0012] Based on further improvements to the above system, the confidence level is calculated using the following formula based on the variance: ; ; ; in, Indicates the number of risk levels. Indicates the maximum variance. Represents the normalized variance. Indicates the confidence level. Indicates variance.

[0013] Based on further improvements to the above system, the system also includes a historical operation record management module, which is used to persistently store the operation information of each data extraction and verification operation.

[0014] Based on further improvements to the above system, the optimization module uses a reinforcement learning model to optimize and adjust the suggestion extraction amount.

[0015] Compared with existing technologies, this invention collects the status information of the source system in real time through a status monitoring module. Based on the status information, it interacts with a large model to obtain suggested extraction volume, risk level, and confidence level. The extraction volume determination module further dynamically determines the final extraction volume or performs alarm processing based on the large model's suggestions. The heterogeneous data extraction module extracts the data corresponding to the final extraction volume for consistency verification, thereby realizing intelligent dynamic adjustment of the medical heterogeneous data extraction batch. This fundamentally eliminates the source database system crash or severe performance fluctuations caused by data extraction and verification operations, ensuring the high availability of the medical business system and effectively balancing the efficiency of data consistency verification and the stability of the source system.

[0016] In this invention, the above-described technical solutions can be combined with each other to achieve more preferred combinations. Other features and advantages of this invention will be set forth in the following description, and some advantages may become apparent from the description or be learned by practicing the invention. The objects and other advantages of this invention can be realized and obtained from what is particularly pointed out in the description and drawings. Attached Figure Description

[0017] The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Throughout the drawings, the same reference numerals denote the same parts. Figure 1 This is a block diagram of a dynamic consistency verification system for heterogeneous medical data according to an embodiment of the present invention. Detailed Implementation

[0018] Preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings, which form part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not intended to limit the scope of the present invention.

[0019] A specific embodiment of the present invention discloses a dynamic consistency verification system for heterogeneous medical data, such as... Figure 1 As shown, it includes: The status monitoring module is used to collect the status information of the source system in real time; The large model interaction module is used to construct prompt words based on the state information and call the large model inference to obtain the suggested extraction quantity, risk level and confidence level. The extraction quantity determination module is used to determine whether to initiate an alarm based on the risk level and confidence level. If so, an alarm message is sent to the early warning module; otherwise, the final extraction quantity is dynamically determined based on the suggested extraction quantity, risk level, and confidence level. The heterogeneous data extraction module is used to connect to heterogeneous data sources and extract a corresponding amount of data from the heterogeneous data sources according to the determined final extraction amount. The verification module is used to perform consistency verification on the extracted data. The early warning module is used to issue alerts based on risk level and confidence level.

[0020] It should be noted that heterogeneous data sources include distributed relational databases, distributed document databases, distributed index databases, and other heterogeneous data sources. The source system includes the heterogeneous data source to be verified and its host machine.

[0021] When the status monitoring module is implemented, it collects status information of the source system, including key performance indicators of heterogeneous data sources and the host machine, thereby building a comprehensive system status assessment system.

[0022] The collected status information includes the following key performance indicators: CPU utilization monitoring includes: total system CPU utilization, database / index service process CPU utilization, CPU load balancing, CPU temperature, etc.

[0023] Memory usage monitoring: Total system memory usage, database / index service process memory usage, memory overflow (OOM) events, memory fragmentation, cache hit rate, etc.

[0024] Disk I / O monitoring: Read / write throughput, I / O wait queue length, disk space utilization, disk response time, IOPS performance metrics, etc.

[0025] Network I / O monitoring: Incoming and outgoing traffic between the data source and the extraction service, network latency, packet loss rate, number of connections, etc.

[0026] Internal database metrics include: active connection count, transaction / query lock status, internal queue depth, slow query log, exception / warning log, and index usage.

[0027] Business metrics monitoring: Key business-level metrics such as data processing speed, error rate, success rate, and average response time.

[0028] The status monitoring module sends the real-time collected status information to the large model interaction module, which then constructs prompts based on the status information.

[0029] During implementation, necessary discretization or normalization processing is performed on continuous indices to construct standardized state vectors for use as input to large models.

[0030] During implementation, a general large model can be used directly (without fine-tuning). Therefore, it is necessary to encapsulate all the necessary background information (such as historical operations, real-time indicators, database configuration, etc.) and the problem to be solved in a context through sophisticated prompt word engineering, and hand it over to the general large model for reasoning, so that the large model can learn and reason based on the context.

[0031] To ensure accurate results from large models, constructing a cue word with high information density and a clear structure is crucial. Below is a cue word template that can be directly generated by the system and input into the LLM.

[0032] # 1. Role and Task Definition # Define the identity, target, and output format of the LLM to ensure its behavior meets expectations.

[0033] You are a top-tier database architect and operations expert for healthcare information systems, specializing in performance tuning and risk control of heterogeneous data systems such as Greenplum, Elasticsearch, and MongoDB.

[0034] Your task is to accurately assess the potential risks of an upcoming data operation based on the extremely detailed "decision context" provided below, and to provide specific, actionable optimization recommendations.

[0035] Your analysis must be data-driven and clearly explain your decision-making logic.

[0036] # 2. Decision Context # Contains all the information needed to perform this assessment.

[0037] # 2.1 Business and Task Profile # Describe the business context of the operation, which determines the risk tolerance.

[0038] # 2.2 Operations to be evaluated # Clearly state what we want to achieve; this is the core objective of the assessment. This application requires the large model to provide a suggested extraction size and a quantified risk level.

[0039] # 2.3 Current System Real-Time Status # A snapshot of the system at the moment before the operation is executed, showing the current status information collected in real time.

[0040] # 2.4 Relevant Historical Operation Records # Provides the most recent operation record that is most similar to the current state, including the previous state -> action -> subsequent state -> result.

[0041] # 3. Evaluation Request # Issue specific instructions to the LLM.

[0042] Based on the complete decision-making context above, please conduct a comprehensive risk assessment of the "operation to be evaluated".

[0043] When implementing this, the risk level is divided into four levels: severe, high risk, medium risk, and low risk.

[0044] During implementation, prompts can be used to instruct the large model to output its assessment confidence level. Alternatively, the confidence level can be calculated based on the large model's predicted probabilities. During implementation, prompts can be used to instruct the large model to output its calculated sampling amount, corresponding probability, risk level, and suggested batch size, and then the confidence level can be calculated based on the probability corresponding to each sampling amount.

[0045] Specifically, the confidence level is obtained using the following method: The variance is calculated based on the probability of each type of extraction predicted by the large model; The confidence level is calculated based on the variance.

[0046] Within the large model, the selection of each digital token (for each specified extraction amount) is accompanied by a probability distribution. The degree of certainty can be quantified by analyzing these probabilities. If the model's selection probabilities are similar across multiple alternative parameters, it indicates high uncertainty, and the confidence level should be lowered.

[0047] Specifically, the raw, unnormalized prediction scores of a large language model before the final output layer (usually a Softmax activation function layer) represent its probability of obtaining each digital token.

[0048] Calculate the variance of these probability distributions .

[0049] The confidence level is calculated based on the variance using the following formula: ; ; ; in, Indicates the number of risk levels. Indicates the maximum variance. Represents the normalized variance. Indicates the confidence level. Indicates variance.

[0050] The smaller the variance, the more concentrated the probability distribution, and the higher the confidence level of the model's conclusions should be; conversely, the larger the variance, the more dispersed the probability distribution, and the lower the confidence level of the model should be.

[0051] After obtaining the suggested sample size, risk level, and confidence level, the sample size determination module determines the sample size based on the suggested sample size, risk level, and confidence level.

[0052] Specifically, the extraction volume determination module includes a decision tree model, a judgment module, a decay calculation module, and an optimization module; The decision tree model is used to determine the current decision based on risk level and confidence level; The judgment module is used to call the early warning module, the attenuation calculation module, or the optimization module based on the current decision. The attenuation calculation module is used to perform attenuation calculation on the suggested extraction amount to obtain the final extraction amount; The optimization module is used to optimize and adjust the suggested extraction amount to obtain the final extraction amount.

[0053] In practice, the decision tree model determines the current decision based on the risk level and confidence level, that is, it determines the current decision based on the number of suggestions extracted from the large model, the corresponding risk level, and the confidence level.

[0054] In practice, the risk level is the root node of the decision tree model, and the confidence level is the first-level node of the decision tree model. The corresponding decision is determined based on different risk levels and confidence levels.

[0055] Specifically, the decision tree model's judgment rules are as follows: If the risk level is severe, the current decision will be made manually. If the risk level is high, and the confidence level is greater than or equal to the first threshold, the current decision is to adopt the recommendation; if the confidence level is less than the first threshold but greater than or equal to the second threshold, the current decision is to use a decay strategy with a high decay intensity; if the confidence level is less than the second threshold, the current decision is to make a manual decision. If the risk level is medium risk, and the confidence level is greater than or equal to the first threshold, the current decision is an optimization strategy; if the confidence level is less than the first threshold but greater than or equal to the second threshold, the current decision is a decay strategy with a medium decay strength; if the confidence level is less than the second threshold, the current decision is an empirical strategy. If the risk level is low and the confidence level is greater than or equal to the second threshold, the current decision is to optimize the strategy; otherwise, the current decision is to adopt the recommendation.

[0056] In practice, the first threshold can be set to 0.85 and the second threshold can be set to 0.6.

[0057] Once the large model determines that the risk has reached the "severe" level, it means that a catastrophic event such as service interruption may occur. Therefore, the most conservative strategy must be adopted, requiring manual intervention. In implementation, the highest level alert can be sent to the DBA and operations team, along with the complete decision context, requiring manual analysis and decision-making.

[0058] If the large model indicates a high risk level, the corresponding decision is determined based on the confidence level. If the confidence level is high (greater than or equal to the first threshold), the recommended value given by the LLM is fully adopted, and no optimization exploration is conducted. Simultaneously, due to the high risk, an advanced alert can be triggered to notify operations personnel to pay attention to this batch of tasks. If the confidence level is medium (less than the first threshold but greater than or equal to the second threshold), the large model considers there to be a high risk, but lacks confidence in this judgment. Therefore, the current decision is a decay strategy, that is, the suggested extraction amount given by the large model is reduced to obtain execution parameters that are more conservative than the large LLM's recommended value. If the confidence level is low, the large model vaguely senses a high risk, which is a very dangerous situation. Therefore, the current decision is a manual decision, and an alert is sent to operations personnel.

[0059] If the large model assigns a medium risk level, the corresponding decision is determined based on the confidence level. If the confidence level is high, the large model clearly identifies a medium risk and provides a clear suggestion. Since the risk level is not high, further numerical optimization can be performed; that is, the current decision is an optimization strategy, and subsequent optimization exploration can be carried out using an RL agent. If the confidence level is medium (the most common scenario), the large model provides a suggestion but there is some uncertainty. Therefore, the suggestion extraction amount given by the large model is conservatively adjusted; that is, the decision is a decay strategy, and the decay strength can be medium. If the confidence level is low, the judgment of the large model is very ambiguous. To ensure safety, the current strategy is an empirical strategy. During implementation, the extraction amount corresponding to operations similar to the current state information can be queried from the historical operation record management module as the final extraction amount.

[0060] If the risk level is low and the confidence level exceeds the second threshold, the current environment is considered safe, and the decision is to optimize the strategy. If the confidence level is less than the second threshold, the current decision is to adopt the suggestion and not optimize.

[0061] The judgment module calls different processing modules to perform alarm processing, attenuation calculation or optimization calculation based on the current decision given by the decision tree model.

[0062] If the current decision involves human intervention, an alarm message is sent to the early warning module, which includes the complete decision context, requiring human analysis and decision-making.

[0063] If the current decision includes adopting the suggestion, the decision module will use the suggested extraction amount as the final extraction amount.

[0064] If the current decision includes a decay strategy, the decay calculation module performs decay calculations based on the suggested extraction amount to obtain the final extraction amount.

[0065] If the current decision includes an optimization strategy, the optimization module will adjust the extraction amount based on the suggested extraction amount to obtain the final extraction amount.

[0066] If the current decision includes experience-based decisions, the judgment module retrieves the extraction quantity corresponding to the operation most similar to the current state information from the historical records and uses it as the final extraction quantity.

[0067] Specifically, the early warning module can issue alerts by sending messages to the maintenance terminal or text messages to maintenance personnel. It can also alert maintenance personnel by sounding an alarm.

[0068] Specifically, the attenuation calculation module attenuates the suggested extraction amount to obtain the final extraction amount in the following way: ; ; in, Indicates the final extraction amount. Indicates the suggested extraction amount. Indicates the attenuation coefficient. Indicates the confidence level. The confidence decay function is represented by the function. The weights representing the confidence level. This represents the real-time risk decay value. Indicates real-time risk weights. This represents the risk decay value based on historical experience. This represents the risk weight based on historical experience.

[0069] Specifically, the confidence decay function is expressed as follows: ; in, This indicates the maximum decay rate of the confidence level.

[0070] The confidence level of a large model reflects its self-awareness. Within the medium confidence range [0.6, 0.85), the lower the score, the more ambiguous the scenario encountered by the large model and the less historical data support it has. Therefore, the system should apply greater conservatism.

[0071] The maximum attenuation percentage that may result solely from insufficient confidence is defined (e.g., 0.4, indicating a maximum reduction of 40%). In practice, the maximum attenuation percentage varies depending on the intensity of attenuation; the maximum attenuation percentage for medium-intensity attenuation is less than that for high-intensity attenuation.

[0072] No matter how credible the recommendations of a large model are, the current actual health of the system cannot be ignored. A system under high load and pressure is fragile, and any large operation could be the straw that breaks the camel's back. Therefore, the impact of real-time risks needs to be considered.

[0073] Specifically, the attenuation calculation module obtains the real-time risk attenuation value in the following way: Calculate the risk score for each key indicator based on the current status information; The real-time risk decay value is obtained based on the risk score of each key indicator.

[0074] During implementation, health and danger thresholds can be set for each key performance indicator, and then the risk score of each key indicator in the current status information can be calculated.

[0075] Specifically, the risk score for each key indicator is calculated using the following formula: ; in, This indicates the value of the key indicator in the current status information. This indicates the health threshold for this key indicator. This indicates the danger threshold for this key indicator.

[0076] The real-time risk decay value is obtained by weighted averaging of the risk scores of all key indicators.

[0077] In order to learn from past failures, if the current operating environment is highly similar to a historical operating environment that caused system delays or even failures, a strong "punitive" attenuation must be applied to avoid repeating the same mistakes.

[0078] Specifically, the attenuation calculation module obtains historical experience risk attenuation values ​​using the following method: Filter high-risk operations whose similarity to the current status information exceeds a preset threshold; Calculate the average similarity between the selected high-risk operations and the current state information; The historical experience risk decay value is obtained based on the average similarity and the maximum historical risk decay ratio.

[0079] During implementation, the system also includes a historical operation record management module, which is used to persistently store the operation information of each data extraction and verification operation.

[0080] During implementation, after each extraction and verification, the following information is stored: the source system status before extraction, the final extraction amount, the risk level and confidence level of the large model, the operation result, and the source system performance information during the data extraction operation. The operation result includes success, failure, and source system anomaly / crash. The source system performance information includes the source system status information collected by the status monitoring module during the operation.

[0081] In implementation, the current status information and the status information of high-risk operations from the historical operation record management module can be converted into vectors. The similarity between the two status information pairs can then be calculated using a similarity calculation method, such as cosine similarity. Finally, the average similarity is multiplied by the maximum historical risk attenuation ratio to obtain the historical experience risk attenuation value.

[0082] The optimization module is used to optimize and adjust the suggested extraction amount to obtain the final extraction amount.

[0083] During implementation, reinforcement learning models can be used to optimize and adjust the suggestion extraction amount.

[0084] In practice, the state space of a reinforcement learning model includes current state information, historical information, risk level, and confidence level.

[0085] During implementation, historical information includes the source system status information of the previous operation, the final extraction amount, and the risk level and confidence level obtained from the large model.

[0086] The action space defines all possible operations that a reinforcement learning model can perform. In this application, the action space is a discrete set of actions from which the agent of the reinforcement learning model needs to select an optimal action at each decision point. For example, the action space may include INCREASE_BATCH (+2000 actions), MAINTAIN_BATCH (unchanged), and DECREASE_BATCH (-1000 actions).

[0087] In practice, reinforcement learning models can employ reinforcement learning algorithms such as deep Q-networks and Q-Learning.

[0088] Input the current state information, historical information, risk level, and confidence level into the reinforcement learning model, and the model will give the current action. If the action is MAINTAIN_BATCH (unchanged), the suggested extraction amount will be used as the final extraction amount. If the action is INCREASE_BATCH (+2000 items), the suggested extraction amount will be increased by 2000 items and used as the final extraction amount. If the action is DECREASE_BATCH (-1000 items), the suggested extraction amount will be decreased by 1000 items and used as the final extraction amount.

[0089] After obtaining the final extraction volume, the heterogeneous data extraction module performs data extraction.

[0090] During implementation, the heterogeneous data extraction module connects to heterogeneous data sources, performs data queries, extracts the corresponding amount of data, and the verification module performs consistency verification on the extracted data.

[0091] During implementation, training samples can be built from the information stored in the historical operation record management module to train the reinforcement learning model.

[0092] To accurately and effectively guide the behavior of reinforcement learning models, this invention designs a comprehensive multi-dimensional reward function. This function quantifies multiple expected outcomes (efficiency, stability, error-free, health) of a single operation into a single scalar reward value. After each operation, the extraction amount is determined based on the action given by the reinforcement learning model, and the reward value is calculated after comparing the data according to this extraction amount.

[0093] Specifically, the reward function of a reinforcement learning model is expressed as: ; in, , , and These represent the weighting coefficients for efficiency, stability, error penalty, and health, respectively. For example, they could be set to 0.4, 0.3, 0.2, and 0.1. During implementation, these values ​​can be adjusted based on the priorities of specific business scenarios. Indicates an efficiency reward. Indicates a stability reward. Indicates punishment for mistakes. This indicates a health reward.

[0094] Specifically, the goal of efficiency rewards is to reward higher data processing volume per unit time (throughput) while also taking into account lower costs.

[0095] Specifically, the efficiency bonus is calculated using the following formula: ; ; ; in, This represents the actual throughput of this operation, calculated as batch size / operation time. This represents the baseline throughput, a value pre-set based on business data. The hyperbolic tangent function is represented. When the throughput is equal to the benchmark, the reward is 50 points; when it is significantly higher than the benchmark, the reward approaches 100 points; and when it is lower than the benchmark, the reward approaches 0 points. This indicates the average CPU load during this operation. This indicates the average memory usage during this operation. This indicates the average I / O latency during this operation. , and This represents the weighting coefficient.

[0096] This represents the resource consumption cost factor, which measures how many system resources are consumed to achieve the current throughput. The more resources consumed, the larger this factor becomes, and the greater the "discount" on the basic efficiency reward. This represents the normalization function.

[0097] The target of the stability reward is the smooth behavior that has a small impact on the source system during operation.

[0098] ; ;

[0099] in, Indicates CPU load volatility. The weighting coefficients can be represented by... The calculation measures the fluctuation of CPU peak values ​​relative to the average value. This indicates the highest CPU load value during this operation, representing the maximum instantaneous pressure this operation exerted on CPU resources. This represents the average CPU load during this operation. Indicates I / O wait volatility. This represents the corresponding weighting coefficient, which can be determined according to... The calculated value is used to measure the jitter level of I / O wait. This indicates the maximum I / O wait time during this operation, representing the average workload of the disk system during this operation. This represents the average I / O wait time during this operation, indicating the maximum instantaneous blocking effect this operation has on the disk system.

[0100] This indicates the internal pressure penalty item of the database.

[0101] This represents the total database lock wait time during this operation. Lock wait affects online business response and is therefore included in the reward function. The weighting coefficient represents the database lock wait time.

[0102] This represents the number of queries queued in the database. This represents the weighting coefficient for the number of sorted queries.

[0103] This indicates a fluctuation penalty.

[0104] The goal of error penalty is to severely punish operational failures and data errors. Specifically, error penalties are calculated as follows: ; in, This indicates the error type that represents the most serious error that occurred during this operation. This indicates the number of retries performed for the most serious error in this operation. Indicates the retry penalty weight. This represents the basic penalty function, used to map different error types to penalty values.

[0105] In practice, the basic penalty function can be defined as follows: ; In practice, a base score is found through a basic penalty function, and then the base score is weighted and amplified according to the number of retries to obtain a comprehensive penalty score that accurately reflects the "worst outcome" and "recovery cost".

[0106] The goal of health rewards is to reward behaviors that allow the system to quickly return to a healthy state after an action is performed.

[0107] ; ; ; in, This indicates the average CPU load after this operation. This indicates the CPU load health threshold. This indicates the average memory usage after this operation. This indicates the healthy threshold for average memory utilization. Indicates CPU health. Indicates memory health status.

[0108] In practice, reinforcement learning models can be trained using existing training methods.

[0109] This embodiment provides a consistency verification system for heterogeneous medical data. By dynamically determining the amount of data extracted based on the performance of the source system, it achieves intelligent dynamic adjustment of the amount of heterogeneous medical data extracted. This fundamentally eliminates the source database system crash or severe performance fluctuations caused by data extraction and comparison operations, ensuring the high availability of the medical business system, effectively balancing the efficiency of data consistency verification and the stability of the source system, and achieving efficient and stable consistency verification.

[0110] Those skilled in the art will understand that all or part of the processes of the methods described in the above embodiments can be implemented by a computer program instructing related hardware, and the program can be stored in a computer-readable storage medium. The computer-readable storage medium may be a disk, optical disk, read-only memory, or random access memory, etc.

[0111] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A dynamic consistency verification system for heterogeneous medical data, characterized in that, include: The status monitoring module is used to collect the status information of the source system in real time; The large model interaction module is used to construct prompt words based on the state information and call the large model inference to obtain the suggested extraction quantity, risk level and confidence level. The extraction volume determination module is used to determine whether to trigger an alarm based on the risk level and confidence level. If so, it sends an alarm message to the early warning module. Otherwise, the final extraction amount will be dynamically determined based on the suggested extraction amount, risk level, and confidence level. The heterogeneous data extraction module is used to connect to heterogeneous data sources and extract a corresponding amount of data from the heterogeneous data sources according to the determined final extraction amount. The verification module is used to perform consistency verification on the extracted data. The early warning module is used to issue alerts based on risk level and confidence level; The extraction volume determination module includes a decision tree model, a judgment module, a decay calculation module, and an optimization module; The decision tree model is used to determine the current decision based on risk level and confidence level; The judgment module is used to call the early warning module, the attenuation calculation module, or the optimization module based on the current decision. The attenuation calculation module is used to perform attenuation calculation on the suggested extraction amount to obtain the final extraction amount; The optimization module is used to optimize and adjust the suggested extraction amount to obtain the final extraction amount; The attenuation calculation module attenuates the suggested extraction amount to obtain the final extraction amount in the following manner: ； ； in, Indicates the final extraction amount. Indicates the suggested extraction amount. Indicates the attenuation coefficient. Indicates the confidence level. The confidence decay function is represented by the function. The weights representing the confidence level. This represents the real-time risk decay value. Indicates real-time risk weights. This represents the risk decay value based on historical experience. This represents the risk weight based on historical experience.

2. The dynamic consistency verification system for heterogeneous medical data according to claim 1, characterized in that, The attenuation calculation module obtains the real-time risk attenuation value in the following way: Calculate the risk score for each key indicator based on the current status information; The real-time risk decay value is obtained based on the risk score of each key indicator.

3. The dynamic consistency verification system for heterogeneous medical data according to claim 1, characterized in that, The attenuation calculation module obtains the attenuation value of historical risk using the following method: Filter high-risk operations whose similarity to the current status information exceeds a preset threshold; Calculate the average similarity between the selected high-risk operations and the current state information; The historical experience risk decay value is obtained based on the average similarity and the maximum historical risk decay ratio.

4. The dynamic consistency verification system for heterogeneous medical data according to claim 1, characterized in that, The confidence decay function is expressed as: ； in, This indicates the maximum decay rate of the confidence level.

5. The dynamic consistency verification system for heterogeneous medical data according to claim 1, characterized in that, The large model interaction module obtains confidence scores using the following method: The variance is calculated based on the probability of each type of extraction predicted by the large model; The confidence level is calculated based on the variance.

6. The dynamic consistency verification system for heterogeneous medical data according to claim 5, characterized in that, The confidence level is calculated based on the variance using the following formula: ； ； ； in, Indicates the number of risk levels. Indicates the maximum variance. Represents the normalized variance. Indicates the confidence level. Indicates variance.

7. The dynamic consistency verification system for heterogeneous medical data according to claim 1, characterized in that, The system also includes a historical operation record management module, which is used to persistently store the operation information of each data extraction and verification operation.

8. The dynamic consistency verification system for heterogeneous medical data according to claim 2, characterized in that, The optimization module uses a reinforcement learning model to optimize and adjust the suggested extraction amount.

Images