A data governance system based on data quality assessment

Abstract

The application provides a kind of data governance system based on data quality evaluation, it is related to data governance technical field.The system data monitoring module intercepts data packet and transmits after cutting according to time window, evaluation rule generation module calculates instantaneous deviation percentage and is mapped as trusted mark, obtains stability index by statistical turnover frequency, generates executable vector.Real-time evaluation module compares data packet field value with vector boundary, and outputs report if it is out of range.Data governance module covers abnormal field with boundary value and marks, and sends back downstream link.Feedback optimization module adjusts stability index to optimize evaluation rule.The application improves the timeliness of capturing abnormal data, accurately determines the eligibility of data through three progressive determinations, improves the fault tolerance rate of quality evaluation, realizes self-optimization iteration of evaluation rule using feedback mechanism, and enhances the long-term operation adaptability of the system.

Description

Technical Field

[0001] This invention relates to the field of data governance technology, specifically a data governance system based on data quality assessment. Background Technology

[0002] In today's digital age, data has become a core asset for various industries, and its quality is directly related to the operational efficiency, decision-making accuracy, and system reliability of enterprises. In fields such as industrial automation, community healthcare, and financial transactions, the real-time generation and flow of massive amounts of data provides a fundamental support for the intelligent operation of related systems.

[0003] While traditional data governance has played a vital role in basic data processing, its limitations have become increasingly apparent when faced with the real-time, dynamic, and complex nature of data. Existing data governance technologies struggle to provide real-time and accurate assessments and dynamic governance of data quality. In real-time data streams, the rapid generation and changes of data make it difficult for traditional static governance methods to capture data anomalies in a timely manner and process abnormal data immediately, leading to the spread of abnormal data within the system and triggering a chain reaction. Summary of the Invention

[0004] The purpose of this invention is to provide a data governance system based on data quality assessment, thereby solving the above-mentioned technical problems. This objective can be achieved through the following technical solutions:

[0005] The data monitoring module continuously captures data packets carrying field length and sampling frequency information at a time synchronized with the device clock at the moment the data is generated. These packets are then segmented into time windows and sent to the evaluation rule generation module.

[0006] The evaluation rule generation module calculates the instantaneous percentage of the measured value of each field relative to the nominal range in the device registry for the data packets after being segmented within the same window, and maps it to a trusted marker. The flipping frequency of the trusted marker in the continuous window is used as a stability index. Then, the percentage of offset is added to the stability index to obtain a dynamic threshold. Finally, the stability index is added to the dynamic threshold to obtain an executable vector containing three elements as the final evaluation rule. Each element of the executable vector defines the qualified boundary of the field in real time.

[0007] The real-time evaluation module compares the qualified boundaries of the current executable vector with the field values ​​of the data packet bit by bit when the data packet flows through the evaluation rule generation module. If any field value exceeds the corresponding boundary, a data evaluation report is output within the same clock cycle.

[0008] Upon receiving a data report, the data governance module immediately performs operations on the abnormal field values ​​in the data report, namely, overwriting the original value with a boundary value and marking the replacement bit, and then sending the replaced data packet back to the downstream link of the system.

[0009] The feedback optimization module sends the replacement bit information back to the evaluation rule generation module in the next clock cycle after each replacement action is completed. This triggers the replacement bit information to be used as the weight adjustment amount of the stability index. The weight adjustment amount is then multiplied by the stability index and re-involved in the calculation of the executable vector.

[0010] As a further technical solution, the data monitoring module is configured to initiate an interception action at the instant of a physical signal jump in the data generated by the data source device, based on a local clock beat that is hard-synchronized with the device's clock source. During the interception action, the data monitoring module first parses the physical layer frame structure of the data stream, identifies and extracts complete data packets containing original field length identifiers and system-specific sampling frequency codes. For example, in community IoT, data packets generated by sensors often contain fields such as temperature, humidity, and pressure. Each field has a fixed length identifier, and the sampling frequency code indicates the speed of data acquisition. The data monitoring module divides the continuous data packet sequence into time window units according to the sampling frequency. Each unit contains an integer number of complete data fields to maintain data integrity. Its output is directly connected to the input buffer queue of the evaluation rule generation module. The segmented data packets are sent to the evaluation rule generation module to ensure efficient data flow.

[0011] As a further technical solution, the evaluation rule generation module calculates the instantaneous offset percentage between the measured value and the nominal range in the device registry for the data packet fields segmented within the time window unit. The nominal range is dynamically generated based on the statistical distribution characteristics of the device's historical operating data in the device registry. Using the effective measured values ​​of the field within a certain historical period as samples, a probability density distribution curve is calculated, and the first quartile is taken as the lower limit of the dynamic range, and the third quartile as the upper limit. For example, in an intelligent transportation system, the historical vehicle speed data of the vehicle speed sensor can be used to dynamically generate a reasonable vehicle speed range to adapt to vehicle speed changes under different time periods and road conditions. The instantaneous offset percentage reflects the degree to which the measured value of the field deviates from the nominal range, providing basic data for the subsequent generation of evaluation rules.

[0012] As a further technical solution, when mapping instantaneous offset percentage to a trusted marker, an asymmetric acceptable range is set according to the field data type. For numerical fields, the positive floating threshold is greater than the negative one, because positive deviations are often more risky, such as pressure sensor values ​​exceeding the limit, which can easily lead to safety accidents. For enumerated fields, the negative floating threshold is zero, and the positive one is bound to the increment of the enumerated item. For example, the equipment operating status field is "0" for normal and "1" for abnormal. The threshold is only triggered when the status is abnormal, avoiding misjudgment of normal status.

[0013] As a further technical solution, the evaluation rule generation module sets evaluation rules based on trusted markers; it statistically analyzes the stability index of the flip frequency of trusted markers within a continuous time window, and dynamically adjusts the instantaneous offset percentage weight according to the stability index to generate a boundary benchmark value. A higher stability index reduces the weight, and a lower stability index increases the weight. The stability index is then amplified by a preset gain coefficient and superimposed on the boundary benchmark value to generate a dynamically expanded tolerance boundary value. Finally, an executable vector is generated, consisting of the boundary benchmark value, the dynamically expanded tolerance boundary value, and the stability index in sequence. The boundary benchmark value serves as the first element, defining the theoretical reference benchmark for field qualification. The dynamically expanded tolerance boundary value serves as the second element, incorporating the buffer space necessary for the inherent fluctuations of the device, forming a real-time execution boundary for field qualification. The stability index serves as the third element and is used as feedback input to the dynamic correlation processing step of the next time window to adjust the contribution weight of the instantaneous offset percentage within that time window. For example, in a meteorological monitoring system, data such as wind speed and humidity fluctuate greatly due to environmental influences. Based on this rule, qualified boundaries adapted to different weather conditions can be generated to accurately identify abnormal data.

[0014] As a further technical solution, the executable vector serves as the final evaluation rule. The real-time evaluation module compares the data packet field value with the boundary benchmark value of the first element of the executable vector. If the value does not exceed the boundary benchmark value, the system is deemed qualified. If the value exceeds the boundary benchmark value, a second evaluation is conducted based on the stability index of the third element. If the stability index is lower than the emergency fluctuation threshold, the system is deemed unqualified. Otherwise, it is temporarily qualified. The emergency fluctuation threshold is dynamically set based on the historical stability index distribution characteristics in the device registry. For example, in a power system, the emergency fluctuation thresholds for parameters such as grid voltage and frequency are generated based on historical stability data, which can accurately distinguish between normal fluctuations and sudden faults, ensuring the stable operation of the power grid.

[0015] As a further technical solution, the real-time evaluation module performs a three-level progressive evaluation on the data packet fields after receiving the executable vector. The first level compares the data packet with the boundary benchmark value; if the value is within the bounds, it is considered qualified. If the value is exceeded, the second level is initiated to compare the tolerance boundary value; if the value is exceeded, a data report is generated. If the value is within the bounds, the third level is triggered to compare the stability index with the emergency fluctuation threshold. The data report details the field position, measured value, and boundary violation level. The three levels of evaluation are interconnected: the first level sets the theoretical baseline, the second level captures sudden distortions, and the third level uses the stability index and historical thresholds to dynamically decouple inherent fluctuations from real anomalies. Even when the device experiences high jitter, it still maintains millisecond-level anomaly identification accuracy, which is superior to traditional static or dual-threshold schemes.

[0016] As a further technical solution, the data governance module extracts the corresponding boundary value from the executable vector to replace the abnormal value according to the location of the abnormal field and the type of boundary violation in the report, and marks the replacement bit in the header of the data packet; for example, in a financial transaction system, if the transaction amount field exceeds the limit abnormally, the module replaces it with the boundary value and marks the location; after replacement, the data packet is sent back to the downstream link to ensure the continuity of data flow; the boundary value of the executable vector is directly reused to avoid secondary calculation delay, and the replacement bit is marked to record the boundary type, providing accurate information for the feedback optimization module.

[0017] As a further technical solution, the feedback optimization module parses the data packet header replacement bit marker, extracts the abnormal field location code and boundary type identifier; generates weight adjustment amounts according to boundary type differences, with tolerance boundaries corresponding to high weights and benchmark boundaries corresponding to low weights; the weight adjustment amounts are sent back to the evaluation rule generation module as correction factors for the stability index; the stability index is updated through multiplication operations and incorporated into the next time window's executable vector calculation; for example, in industrial automated production, equipment parameters are frequently corrected and fed back, and the system adjusts the evaluation rules accordingly to gradually adapt to complex production environments and improve evaluation accuracy.

[0018] This invention provides a data governance system based on data quality assessment, which has the following beneficial effects:

[0019] This invention achieves zero-delay triggering of data quality assessment by activating a hard clock synchronization interception mechanism at the moment of data generation, thereby improving the timeliness of abnormal data capture.

[0020] This invention generates an executable vector containing boundary baseline values, dynamic tolerance boundary values, and stability indices by fusing instantaneous offset percentages and stability indices. Based on this vector, a three-level progressive judgment is performed, achieving adaptive precision in data qualification judgment and significantly improving the evaluation error tolerance rate in complex scenarios.

[0021] This invention achieves self-optimization and iteration of evaluation rules by replacing the tag-driven feedback mechanism, effectively improving the adaptability of the system in long-term operation. Attached Figure Description

[0022] Figure 1 This is a schematic diagram of the process of the present invention. Detailed Implementation

[0023] The technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0024] The data monitoring module is the foundation of the entire data governance system. In a community healthcare environment, it primarily monitors data generated by various medical devices in real time, including but not limited to electronic blood pressure monitors, blood glucose meters, electrocardiographs, and thermometers. The module initiates an interception process the instant data is generated by the medical device, based on a local clock that is hard-synchronized with the device's clock source. For example, when measuring a patient's blood pressure with an electronic blood pressure monitor, the module intercepts data packets carrying field length and sampling frequency information at the instant the physical signal of the blood pressure value changes, according to the local clock that is hard-synchronized with the device's clock source. Specifically, the module parses the physical layer frame structure in the data stream in real time, identifying and extracting data packets containing the original field length identifier and the specified sampling frequency. The system provides complete data packets with a built-in sampling frequency encoding. For electronic blood pressure monitors, the data packets contain fields such as systolic and diastolic blood pressure, each with a fixed length identifier. The sampling frequency encoding indicates the time interval for data collection. The data monitoring module divides the continuous data packet sequence into time window units based on the sampling frequency. Each unit contains an integer number of complete data fields to ensure data integrity. For example, if the sampling frequency is once per minute, the time window can be set to 1 minute, and each time window unit will contain one complete blood pressure data field. The output of the time window unit is directly connected to the input buffer queue of the evaluation rule generation module, enabling the segmented data packets to be quickly and efficiently sent to the evaluation rule generation module, ensuring the efficiency of data flow.

[0025] After the data monitoring module completes the interception of data packets and the construction of time window units, the data is promptly sent to the evaluation rule generation module. The evaluation rule generation module is the core of data quality assessment; it generates corresponding evaluation rules based on the device's historical data and the actual situation of the current data. The evaluation rule generation module first calculates the instantaneous percentage deviation of the measured value relative to the nominal range in the device registry for each data packet field after segmentation within the time window unit. The device registry stores the device's historical operating data, which forms the basis for calculating the nominal range. Taking an electronic blood pressure monitor as an example, it extracts all valid measured values ​​of the systolic blood pressure field for a specific historical period, such as the past month, and calculates its probability density distribution curve. The first quartile (Q1) of the distribution curve is used as the lower limit of the dynamic range, and the third quartile (Q3) is used as the upper limit of the dynamic range. Assuming that the historical data of the electronic blood pressure monitor over the past month shows that the systolic blood pressure Q... 1 is 110 mmHg, Q3 is 130 mmHg; when the measured systolic blood pressure is 140 mmHg, the instantaneous offset percentage is calculated as follows: (measured value - nominal upper limit) / (nominal upper limit - nominal lower limit) × 1 = (140 - 130) / (130 - 110) × 1 = 0.5; next, the instantaneous offset percentage is mapped to a confidence level; for numerical blood pressure fields, an asymmetric acceptable range is set, with a positive floating threshold of 20% and a negative floating threshold of 15%; this is because high blood pressure is usually more risky than low blood pressure, for example, high systolic blood pressure may lead to serious complications such as hypertensive crisis; when the instantaneous offset percentage falls between -15% and +20%, it is marked as a confidence level, otherwise it is marked as an unconfidential level; in this example, the instantaneous offset percentage of the measured systolic blood pressure is 50%, which exceeds the positive floating threshold of 20%, so it is marked as an unconfidential level.

[0026] Simultaneously, the evaluation rule generation module statistically analyzes the state flipping frequency of the trust marker for the same field within consecutive time window units to obtain a stability index. For example, if the trust marker for the shrinkage field undergoes four state flips within ten consecutive time window units (e.g., from trust to untrustworthy, then back to trustworthy, back to untrustworthy, and finally back to trustworthy), the stability index is calculated as a flipping frequency value of 0.4, calculated by dividing the number of flips by the time window. Then, the instantaneous offset percentage is dynamically correlated with the stability index, and the contribution weight of the instantaneous offset percentage in generating boundary baseline values ​​is dynamically adjusted according to the magnitude of the stability index. The weighting coefficient is set to 0.7 when the stability index is less than 0.3, and 0.6 when it is between 0.3 and 0.6. When the value is between 0.3 and 0.6, the weighting factor is 0.85; when it is greater than 0.6, the weighting factor is 1.0. In this example, the stability index is 0.4, which is between 0.3 and 0.6, so the weighting factor is 0.85. The boundary reference value is then calculated as instantaneous offset percentage × weighting factor = 0.5 × 0.85 = 0.42.5. Next, the stability index is amplified by a preset gain factor, such as 0.6, and superimposed on the boundary reference value to generate a dynamically expanded tolerance boundary value. That is, tolerance boundary value = boundary reference value + stability index × gain factor = 0.42.5 + 0.4 × 0.6 = 0.665. Finally, an executable vector is generated, which consists of the boundary reference value, the dynamically expanded tolerance boundary value, and the stability index in sequence.

[0027] After the evaluation rule generation module completes the generation of the executable vector, the real-time evaluation module immediately takes over the task and compares the data packets bit by bit. The real-time evaluation module is the execution link of data quality evaluation, responsible for evaluating the data packets in real time according to the executable vector generated by the evaluation rule generation module. The real-time evaluation module first compares the field values ​​of the data packets with the boundary reference value of the first element of the executable vector. Taking the systolic blood pressure of an electronic blood pressure monitor as an example, the systolic blood pressure value corresponding to the boundary reference value is calculated as: nominal upper limit 130mmHg + (130-110)×0.425 = 130 + 21.25 = 138.5mmHg. The measured value of 140mmHg exceeds the boundary reference value of 138.5mmHg, so the second-level evaluation is initiated, comparing the measured value with the second element. The systolic blood pressure value (130 + (130 - 110) × 0.665 = 143.3 mmHg) corresponding to the dynamically extended tolerance boundary value is compared in real time. The measured value of 140 mmHg is less than the tolerance boundary value of 143.3 mmHg, triggering the third-level evaluation: extract the stability index (0.4) of the third element of the executable vector and perform a real-time numerical comparison with the emergency fluctuation threshold (assumed to be 0.5) dynamically set in the device registry. Since the stability index of 0.4 is lower than the emergency fluctuation threshold of 0.5, the systolic blood pressure field value is determined to be unqualified and a data report is triggered, recording the location of the systolic blood pressure field, the measured value of 140 mmHg, and details of the violation of the boundary level, among which the value exceeds the boundary reference value but does not exceed the tolerance boundary value, and the stability index is lower than the emergency fluctuation threshold.

[0028] Upon receiving the data report generated by the real-time assessment module, the data governance module takes swift action. As a crucial component of data quality governance, the data governance module handles anomalous data to ensure its accuracy and reliability. Based on the location of the anomalous field and the type of boundary violation recorded in the report, the data governance module extracts the systolic blood pressure value of 151.25 mmHg corresponding to the boundary baseline value of the relevant field from the executable vector currently used by the real-time assessment module as a coverage value. This coverage value replaces the measured value of the anomalous field, 140 mmHg, and a replacement bit is marked in the data packet header, for example, marked as "systolic blood pressure field replaced, boundary type is boundary baseline value". The replaced data packet is then sent back to the downstream links of the system to ensure data flow continuity, enabling subsequent medical business processes to operate normally based on accurate data.

[0029] After the data governance module completes the data replacement, the feedback optimization module immediately begins its work. This module is a crucial part of the system's self-optimization process, responsible for adjusting and optimizing the evaluation rules based on the data governance results. In the next clock cycle after the replacement operation, the feedback optimization module parses the replacement bit marker in the data packet header, extracts the location code of the abnormal field, and identifies the boundary type. Based on the boundary type identifier, such as the boundary baseline value, it dynamically generates a low-weight adjustment amount; for example, setting the weight adjustment amount for the baseline boundary to 0.2 and the weight adjustment amount for the tolerance boundary to 0.4. This weight adjustment amount of 0.2 is then sent back to the evaluation rule generation module as a real-time correction factor for the stability index. The evaluation rule generation module multiplies the weight adjustment amount with the current stability index of 0.4, updating the stability index value required for the next time window's executable vector calculation to 0.4 × 0.2 = 0.08. In subsequent time windows, the evaluation rule generation module recalculates the executable vector based on the updated stability index, thereby achieving self-optimization iteration of the evaluation rules and enabling the system to better adapt to changes in data quality in the community healthcare environment.

[0030] The system's data monitoring module can capture data in a timely and accurate manner, ensuring data integrity and real-time performance. The evaluation rule generation module generates precise evaluation rules based on historical and real-time data, providing a reliable basis for the real-time evaluation module. The real-time evaluation module can quickly identify anomalies in the data and generate detailed data reports. The data governance module processes abnormal data in a timely manner, ensuring data accuracy and reliability. The feedback optimization module continuously adjusts the evaluation rules, improving the system's adaptability and evaluation accuracy.

[0031] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A data governance system based on data quality assessment, characterized by, The application comprises: a data monitoring module, which, at the moment of data generation, continuously intercepts data packets carrying field length and sampling frequency information in synchronization with the device clock, and then sends the data packets to the evaluation rule generation module after time window division; the evaluation rule generation module, which calculates the instantaneous offset percentage of each field measured value relative to the nominal range in the device registry within the same window, maps it to a trusted marker, and calculates the stability index by counting the frequency of the trusted marker flipping in the continuous window, then adds the offset percentage and the stability index to obtain a dynamic threshold, and finally adds the stability index and the dynamic threshold to obtain an executable vector containing three elements as the final evaluation rule, and each element of the executable vector defines the field qualification boundary in real time; the real-time evaluation module, which compares the qualification boundary of the current executable vector with the field value of the data packet when the data packet passes through the evaluation rule generation module, and if any field value exceeds the corresponding boundary, the data report of the data evaluation is output in the same clock cycle; the data governance module, which, upon receiving the data report, immediately performs operations on the abnormal field value in the data report, i.e., replaces the original value with the boundary value and marks the replacement bit, and then sends the replaced data packet back to the downstream link of the system; the feedback optimization module, which, in the next clock cycle after each replacement action is completed, sends the replacement bit information back to the evaluation rule generation module, triggers the replacement bit information as a weight adjustment amount of the stability index, multiplies the weight adjustment amount and the stability index to participate in the calculation of the executable vector again.

2. The data governance system based on data quality assessment according to claim 1, characterized in that: The data monitoring module starts the interception action based on the local clock synchronized with the device clock source at the physical signal transition moment when the data source generates data, which includes: real-time analysis of the physical layer frame structure in the data stream, identification and extraction of complete data packets carrying original field length identifiers and device inherent sampling frequency codes; according to the physical time reference corresponding to the device inherent sampling frequency code, the continuous intercepted data packet sequence is divided into discrete time window units with time alignment relationship, each time window unit contains an integer number of complete data fields determined by the original field length identifier; the output end of the time window unit is connected to the input buffer queue of the evaluation rule generation module to make the divided data packets sent to the evaluation rule generation module.

3. The data governance system based on data quality assessment of claim 1, wherein: The evaluation rule generation module calculates the instantaneous offset percentage of each data packet field in the time window unit relative to the nominal range of the field in the device registry in real time; the nominal range is dynamically generated by the device registry based on the statistical distribution characteristics of the field in the historical operation data of the device, specifically: extracting all valid measured values of the field in a certain historical period, calculating the probability density distribution curve, taking the first quartile of the distribution curve as the lower limit of the dynamic range and the third quartile as the upper limit of the dynamic range; the instantaneous offset percentage is mapped to a trusted marker.

4. The data governance system based on data quality assessment according to claim 3, characterized in that the instantaneous The mapping rule of the offset percentage to the reliable mark is that when the instantaneous offset percentage falls into an asymmetric qualified interval dynamically set based on the data type of the field, the mark is reliable, otherwise, the mark is unreliable; the asymmetric qualified interval is independently set for each field, the lower limit is a negative floating threshold of the lower limit of the nominal range, the upper limit is a positive floating threshold of the upper limit of the nominal range, and the positive and negative floating thresholds are not equal, wherein for a numerical field, the positive floating threshold is set to be greater than the negative floating threshold, and for an enumerated field, the negative floating threshold is set to be zero and the positive floating threshold is bound with the increment of the enumerated item.

5. The data governance system based on data quality assessment according to claim 3, characterized in that, The evaluation rule generation module sets the evaluation rule based on the reliable mark, specifically: the state flip frequency of the reliable mark of the same field in a continuous time window unit is counted, and the state flip frequency value is quantized as a stability index representing the current system running stability; the instantaneous offset percentage calculated in the current window and the stability index are dynamically associated, specifically: the contribution weight of the instantaneous offset percentage in generating the boundary reference value is dynamically adjusted according to the size of the stability index, the higher the stability index, the lower the weight of the instantaneous offset percentage, and the lower the stability index, the higher the weight, and a boundary reference value constrained by the real-time stability index state is output; the boundary reference value and the inherent fluctuation degree of the device represented by the stability index are fused, a dynamically expanded tolerance boundary value is generated by amplifying the stability index by a preset gain coefficient and adding it to the boundary reference value; finally, an executable vector composed of the boundary reference value, the dynamically expanded tolerance boundary value and the stability index in sequence is generated; wherein the boundary reference value is the first element, the dynamically expanded tolerance boundary value is the second element, and the stability index is the third element.

6. The data governance system based on data quality assessment according to claim 5, characterized in that: The executable vector as the final evaluation rule specifically is: when the real-time evaluation module performs bit-by-bit comparison, the data packet field value is compared with the first element of the executable vector, that is, the boundary reference value, if it does not exceed, it is determined to be qualified; if it exceeds the boundary reference value but does not exceed the second element, that is, the dynamically expanded tolerance boundary value, a secondary determination is made in combination with the third element, the stability index: when the stability index is lower than a preset emergency fluctuation threshold, it is determined to be unqualified and triggers data reporting, and when the stability index is higher than the threshold, the field value is temporarily qualified to accommodate the inherent fluctuation of the device, and the emergency fluctuation threshold is dynamically set based on the historical stability index distribution of the device registry; if the field value directly exceeds the dynamically expanded tolerance boundary value, it is immediately determined to be unqualified; wherein the secondary determination logic is realized by comparing the stability index with the emergency fluctuation threshold in real time, the threshold is dynamically set based on the historical stability index distribution characteristics of the field in the device registry, and the determination result is fed back to the evaluation rule generation module in real time, which is used to adjust the contribution weight of the instantaneous offset percentage in the next time window dynamic association processing step.

7. The data governance system based on data quality assessment according to claim 6, characterized in that: The real-time evaluation module performs a three-stage progressive evaluation on each field of the data packet within the same clock cycle after receiving the executable vector. First, the measured value of the field is compared with the first element boundary reference value of the executable vector. If it does not exceed, it is determined to be qualified. If it exceeds the boundary reference value, the second-stage evaluation is started: the measured value is compared with the second element dynamic expansion tolerance boundary value in real time. If it exceeds this value, a data report is immediately generated. If the measured value is between the boundary reference value and the dynamic expansion tolerance boundary value, the third-stage evaluation is triggered: the third element stability index of the executable vector is extracted and compared with the emergency fluctuation threshold dynamically set by the device registry in real time. When the stability index is lower than the emergency fluctuation threshold, a data report marking a real anomaly is generated. Otherwise, a temporary qualified mark is output.

8. The data governance system based on data quality assessment of claim 1, wherein: The data governance module extracts the first element boundary reference value or the second element dynamic expansion tolerance boundary value of the corresponding field in the executable vector currently used by the real-time evaluation module as an override value according to the abnormal field position and the boundary violation type recorded in the report, replaces the measured value of the abnormal field with the override value, marks the replacement position in the data packet header, and sends the replaced data packet to the downstream link. The mechanism of directly reusing the existing boundary value of the executable vector avoids secondary calculation delay, and the technical details of the replacement position marking synchronous recording of the boundary type provide double positioning information of the field position and the boundary level for the feedback optimization module.

9. The data governance system based on data quality assessment as claimed in claim 1 wherein, The feedback optimization module analyzes the replacement position mark in the data packet header to extract the abnormal field position code and the boundary type identifier. According to the boundary type identifier, a differentiated weight adjustment amount is dynamically generated. When the boundary type identifier is a tolerance boundary, a high-weight adjustment amount is generated. When it is a reference boundary, a low-weight adjustment amount is generated. The weight adjustment amount is returned to the evaluation rule generation module as a real-time correction factor of the stability index. By performing multiplication operation between the weight adjustment amount and the current stability index, the stability index value required for the executable vector calculation in the next time window is updated.

Images