Drug compliance detection method and system for drug supervision

Abstract

The invention relates to the technical field of data integrity verification, in particular to a drug compliance detection method and system for drug supervision, and the method comprises the following steps: extracting a process change data field, eliminating a deviation value and setting a reference, marking and classifying a current batch difference field, recognizing a deviation expansion item, constructing a field mapping path and recognizing an interruption point. And judging whether the interrupt field is missing or not, and generating various result lists. According to the method, through medicine process parameter extraction and abnormal value elimination, a stable reference standard is constructed, representativeness of a comparison standard is guaranteed, classification is performed based on field differences, parameter offset is identified, difference identification precision is improved, continuous offset fields are judged in combination with trend changes, risk early warning timeliness is enhanced, and data inheritance continuity is expressed; according to the method, breakpoint fields are filled and verified, missing information is accurately captured, a closed-loop process of offset identification, trend research and judgment, path tracing and missing verification is integrally formed, and the integrity and sensitivity of data compliance inspection are improved.

Description

Technical Field

[0001] This invention relates to the field of data integrity verification technology, and in particular to a drug compliance testing method and system for drug regulation. Background Technology

[0002] The field of data integrity verification technology involves the detection and control of whether electronic data remains in its original, unaltered state during generation, transmission, storage, and access. Its core aspects include data tampering detection, information consistency verification, electronic evidence preservation, and record timestamp verification. The overall technical methods typically employ hash value comparison, digital signature verification, irreversible encryption verification, and blockchain evidence storage to ensure a trustworthy chain of data from source to end. Traditionally, drug compliance testing methods used in pharmaceutical regulation involve regulatory authorities or third-party institutions comparing the consistency between original records submitted by companies in the production, distribution, and sales stages of drugs and standard specifications to determine whether drugs meet national quality and distribution standards. This type of method targets the verification of the authenticity and consistency of drug information records. Traditional methods typically involve manually sampling company-submitted data against paper ledgers, reviewing electronic archive system logs, and accessing key data files, comparing them using checksums or file digest values ​​to complete data compliance reviews.

[0003] Current data integrity verification methods primarily rely on manual verification of paper records, log systems, or summary checks, resulting in low automation in data processing and low identification efficiency in large-scale data scenarios. For process parameter data, existing methods lack quantitative extraction of the fluctuation range and baseline patterns of field values, only allowing for rough comparisons through sampling, making it difficult to capture the trend risks implied by field anomalies. Furthermore, traditional methods cannot track and identify the continuous transmission of data across batches; interruptions or missing data often go undetected due to their subtlety, posing potential risks of process changes. Verification of missing data typically relies on manual review and field checks, which is inefficient and susceptible to subjective judgment, leading to the omission of key fields and affecting the comprehensive understanding of the authenticity of declared information. For example, if a key process parameter in a drug batch shows a continuous deviation trend but has not reached the sampling threshold, traditional methods may fail to detect it in time, potentially allowing non-compliant processes to pass the declaration and increasing regulatory risks. These shortcomings are particularly prominent in high-frequency, multi-batch drug regulatory tasks, restricting the systematic and in-depth nature of data compliance review. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies by proposing a drug compliance testing method for drug regulation.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a drug compliance testing method for drug regulation, comprising the following steps:

[0006] S1: Extract the quality control fields from the drug manufacturing process change data, classify them according to product attributes and specifications, remove records that deviate from the median value by more than twice the interquartile range, group the retained fields, extract the median value and range of variation, and generate a regulatory indicator reference benchmark value table.

[0007] S2: Based on the field standards set in the regulatory indicator reference benchmark value table, compare the process parameter values ​​in the current batch of drug applications, mark the fields with differences and classify and summarize them to generate a list of application field offset comparisons;

[0008] S3: Based on the difference fields in the declaration field offset comparison list, combined with the classification of key process parameters, retrieve the change trend of the corresponding records, identify field items with consistent direction and continuously expanding offset, and generate a key field boundary judgment table.

[0009] S4: Based on the field batch information in the key field boundary judgment table, extract the matching parameters of upstream and downstream batches, construct the field mapping path, identify the location where value transmission is interrupted, and generate a cross-batch field mapping chain integrity structure diagram.

[0010] S5: Based on the interruption nodes in the cross-batch field mapping chain integrity structure diagram, and in conjunction with the field filling rules, determine whether the corresponding field is missing, and generate a list of missing declaration field breakpoints.

[0011] As a further embodiment of the present invention, the regulatory indicator reference benchmark value table includes representative median values, normal variation ranges, and field grouping structures; the declaration field offset comparison list includes segment comparison results, offset field list, and difference field markers; the key field boundary judgment table includes offset trend fields, change direction continuity fields, and key process difference fields; the cross-batch field mapping chain integrity structure diagram includes field mapping paths, interruption node positions, and batch matching parameter groups; and the declaration field missing breakpoint list includes interruption node fields, filling rule items, and specification compliance fields.

[0012] As a further aspect of the present invention, step S1 is as follows:

[0013] S101: Extract the quality control field content from the drug manufacturing process change data, classify the fields according to the product attributes and specifications, remove all records in all fields whose values ​​deviate from the median position by more than twice the interquartile range within the field distribution, obtain the numerical sequence of each field after removing anomalies, and generate the standardized numerical sequence of the fields.

[0014] S102: Based on the field standard numerical sequence, perform numerical grouping processing on fields under the same category, call the variety attributes, specifications, field values ​​and field names corresponding to the fields, sort the median position of each group of field values, and calculate the upper and lower fluctuation ranges in combination with the interquartile range to generate a set of field numerical intervals.

[0015] S103: Based on the set of field value intervals, the median position values ​​of the same group of fields are numerically summarized, normalized, and incrementally evaluated. By calculating the distribution difference, absolute difference, and normalization ratio difference between fields, the representative difference value of the fields is obtained. Based on the representative difference value of the fields, the indicator benchmark table is constructed to obtain the regulatory indicator reference benchmark value table.

[0016] As a further aspect of the present invention, step S2 is as follows:

[0017] S201: Based on the setting standard of each field in the regulatory indicator reference benchmark value table, compare the process parameter values ​​in the current batch of drug application, determine whether there is a deviation difference in the value of the corresponding field, record the field items with deviation differences, and obtain the number of difference field identifications.

[0018] S202: Invoke the number of differences in the field identification, classify the field according to the index category, parameter attribute and offset direction of the differences in the field item, mark the classification type, and generate a field classification mark set;

[0019] S203: Based on the field classification tag set, determine the offset magnitude and direction of each classification field item, and filter according to the change threshold in the field setting standard to obtain field items whose offset degree meets the identification conditions, and establish a declaration field offset comparison list.

[0020] As a further aspect of the present invention, step S3 is as follows:

[0021] S301: Based on the difference fields marked in the declaration field offset comparison list, identify the key process parameter classification information bound to the record corresponding to each field, and map the classification information to the difference fields to construct the process parameter set corresponding to the offset fields and generate a field parameter matching set.

[0022] S302: Call the field items in the field parameter matching set, retrieve the time series records bound in the corresponding records, obtain the change range of the field value, the unit time offset value and the number of changes in the field direction, obtain the time change offset intensity value corresponding to the field, filter out the fields whose change direction is continuous in the same direction and whose offset intensity shows an increasing trend, and obtain the enhanced offset signal.

[0023] S303: Based on the identified fields in the enhanced offset signal, summarize the classification labels in the process classification, extract the field items whose signal strength is greater than the signal strength benchmark value, establish a three-item combination table of field name, classification label and enhanced offset signal value, and obtain the key field out-of-bounds judgment table.

[0024] As a further aspect of the present invention, step S4 is as follows:

[0025] S401: Based on the batch information of the fields listed in the key field boundary judgment table, extract the upstream batch number and downstream batch number corresponding to each field, and by matching the matching parameter group of the field in the different batches, call the matching parameter number and parameter attribute value of the field to identify the matching structure matching items between the different batches and generate the matching structure mapping matrix between batches.

[0026] S402: Call the batch-to-batch matching structure mapping matrix to locate the transmission path of the field between upstream and downstream batches. By using the matching parameter values ​​of the nodes in the field mapping path, determine whether there is an interruption position in the transmission path of the field, identify the field jump segment and empty segment in the continuous path, and obtain the field transmission continuity breakpoint index.

[0027] S403: Based on the field transmission continuity breakpoint index, extract the field mapping path segment where the interrupted node is located, calculate the path span, number of transmission parameter groups, field state variability and mean offset of the transmission value sequence within the path segment, and obtain the cross-batch integrity of the field mapping chain through joint analysis of the indexes, and establish a cross-batch field mapping chain integrity structure diagram.

[0028] As a further aspect of the present invention, step S5 is as follows:

[0029] S501: Based on the interruption node in the cross-batch field mapping chain integrity structure diagram, extract the field number and corresponding batch, call the field distribution sequence table and parameter structure, determine whether there is a missing state, and generate a list of missing field identifiers.

[0030] S502: Call the field missing identifier list, combine the field filling rules and the status of upstream and downstream batch data, determine the compliance of the missing fields with the constraint requirements, and obtain the field missing breakpoint index sequence;

[0031] S503: Based on the sequence of missing field breakpoint indicators, filter out abnormal fields, extract field numbers, missing locations and attribute contents, organize the structure records by batch, and establish a list of missing field breakpoints for the application.

[0032] Drug compliance testing systems used for drug regulation include:

[0033] The benchmark indicator construction module obtains the quality control field values, groups the fields by generic name and dosage form parameters, calculates the median, first quartile, third quartile and interquartile range of the field group, filters out records that deviate from the median value by more than twice the interquartile range, calls the data after filtering out to calculate the median, maximum and minimum values, establishes the field fluctuation range, and generates a regulatory indicator reference benchmark value table.

[0034] The field offset identification module, based on the regulatory indicator reference benchmark value table, calls the process parameter field values ​​of the current batch of declarations, compares the field values ​​with the median fluctuation range, filters fields that exceed the range and marks the offset direction, classifies them according to the offset magnitude and field type, and generates a declaration field offset comparison list.

[0035] The trend boundary judgment module, based on the declaration field offset comparison list, calls the key process parameter category information of the offset field, extracts the numerical sequence of the field in consecutive batches, calculates the consistency of change direction and offset increase trend coefficient, filters trend enhancement fields, and generates a key field boundary judgment table.

[0036] Based on the key field out-of-bounds judgment table, the field mapping chain identification module calls the corresponding field parameter group of upstream and downstream of the corresponding batch, analyzes the numerical correspondence between upstream and downstream fields, determines whether the numerical transmission is interrupted, extracts the interruption position node, and generates a cross-batch field mapping chain integrity structure diagram.

[0037] The missing breakpoint extraction module, based on the cross-batch field mapping chain integrity structure diagram, calls the field filling rule list, compares the correspondence between the interrupted node fields and the required fields, filters the missing fields, and generates a list of missing breakpoints for the declared fields.

[0038] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0039] In this invention, a reliable reference benchmark is constructed by structurally extracting and removing outliers from drug process parameters, ensuring the representativeness and stability of the comparison standard. By classifying and marking the difference fields, the deviation of the declared parameters is accurately identified, enhancing the precision of difference identification. Combined with parameter trend changes, key fields with continuous deviations are identified, improving the timeliness of risk warning. By constructing the inter-batch field value transmission path and identifying breakpoints, the integrity of data inheritance is enhanced. The filling and verification of breakpoint fields realizes the targeted capture and feedback of missing information. The linkage of each link forms a closed-loop detection process of difference identification, trend analysis, path tracing, and missing verification, which systematically improves the sensitivity, integrity, and penetration of data compliance review. Attached Figure Description

[0040] Figure 1 This is a flowchart of the main steps of the present invention;

[0041] Figure 2 This is a flowchart of step S1 of the present invention;

[0042] Figure 3 This is a flowchart of step S2 of the present invention;

[0043] Figure 4 This is a flowchart of step S3 of the present invention;

[0044] Figure 5 This is a flowchart of step S4 of the present invention;

[0045] Figure 6 This is a flowchart of step S5 of the present invention. Detailed Implementation

[0046] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0047] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0048] Please see Figure 1 A drug compliance testing method for drug regulation includes the following steps:

[0049] S1: Extract the quality control field content from the drug manufacturing process change data, classify it according to product attributes and specifications, remove records whose deviation from the median position value exceeds twice the interquartile range of the field distribution, retain the remaining values ​​and perform field grouping processing, determine the representative median value and the normal range of variation of the field, and generate a regulatory indicator reference benchmark value table.

[0050] S2: Based on the setting standards of each field in the regulatory indicator reference benchmark value table, compare the process parameter values ​​in the current batch of drug applications, identify the field items with differences and classify them, establish field comparison results, and obtain the application field offset comparison list;

[0051] S3: Based on the difference fields marked in the declaration field offset comparison list, combined with the classification information of key process parameters, retrieve the change trends in the corresponding records for comparison, identify the fields with continuous change direction and continuously aggravating offset trend, and generate a key field boundary judgment table.

[0052] S4: Based on the batch information of the fields listed in the key field boundary judgment table, extract the matching parameter groups between upstream and downstream batches, construct the field mapping path, identify the location of the interruption node in the value transmission in the path, and generate a cross-batch field mapping chain integrity structure diagram.

[0053] S5: Based on the identified interrupted nodes in the cross-batch field mapping chain integrity structure diagram, and in conjunction with the field filling rules, determine whether there are any missing fields corresponding to the nodes, establish a corresponding field list, and generate a list of missing breakpoints for declared fields.

[0054] The regulatory indicator reference benchmark table includes representative median values, normal range of variation, and field grouping structure; the declaration field offset comparison list includes segment comparison results, offset field list, and difference field markers; the key field boundary judgment table includes offset trend fields, change direction continuity fields, and key process difference fields; the cross-batch field mapping chain integrity structure diagram includes field mapping paths, interruption node positions, and batch matching parameter groups; and the declaration field missing breakpoint list includes interruption node fields, filling rule items, and specification compliance fields.

[0055] Please see Figure 2 Step S1 is as follows:

[0056] S101: Extract the quality control field content from the drug manufacturing process change data, classify the fields according to the product attributes and specifications, remove all records in all fields whose values ​​deviate from the median position by more than twice the interquartile range within the field distribution, obtain the numerical sequence of each field after removing anomalies, and generate the standardized numerical sequence of the fields.

[0057] To extract the quality control fields from the pharmaceutical manufacturing process change data, first obtain the process records and corresponding quality inspection data for all products in the changed batch. Then, extract the corresponding fields for products such as injections and oral liquids using a structured approach. For example, for 10ml injections, common fields include "filling volume," "clarity," "pH value," "visible foreign matter," and "sealing." Taking the "filling volume" field as an example, extract 30 batches of samples, as follows: [10.1, 10.2, 10.0, 9.9, 10.3, 10.4, 10.1, 10.0, 10.2, 9.8, 9.7, 10.3, 10.5, 10.6, 9.6, 10.0, 10.1, 10.2, 10.3, 1 [0.5, 10.7, 10.6, 9.8, 9.9, 10.2, 10.3, 10.4, 10.5, 10.1, 10.2], with a median of 10.2, Q1 of 10.0, and Q3 of 10.4. The IQR is calculated to be 0.4, and the upper and lower bounds are calculated to be 10.2 ± 2 × 0.4, i.e., [9.4, 11.0]. All outliers outside this range are removed. This can screen out filling batches that exceed the upper limit of the floating range. Similar operations are performed to independently filter other fields, ultimately forming a set of valid values ​​for each field. Each record comes from the actual batch's corresponding test file or database record system. The table below lists the original extracted and processed samples of the filling capacity field.

[0058] Table 1 Initial Records and Processed Sample Tables for Filling Capacity Field

[0059] Batch number Original value (ml) Should it be removed? B001 10.1 no B005 10.5 no B015 9.6 no B021 10.7 no B024 9.8 no B030 10.2 no …… …… ……

[0060] As shown in Table 1, the sample distribution did not exceed the IQR limit, and no rejection occurred, forming a standardized numerical sequence of fields.

[0061] S102: Based on the field standard numerical sequence, perform numerical grouping processing on fields under the same category, call the variety attributes, specifications, field values ​​and field names corresponding to the fields, sort the position values ​​by the median position of each group of field values, and calculate the upper and lower fluctuation ranges by combining the interquartile range, and generate a set of field numerical intervals.

[0062] Based on the standardized numerical sequences of fields, and according to the classification rules of drug varieties and specifications, the numerical values ​​of each field are grouped. Taking a 10ml injection as an example, the fields "pH value", "clarity", and "filling volume" are grouped together. First, the median value is calculated for the data within each field. For example, the records in the "pH value" group are: [6.8, 7.0, 7.1, 6.9, 7.0], with a median of 7.0; the records for "clarity" are [95, 94, 96, 94, 97], with a median of 95; and the records for "filling volume" are [10.2, 10.1, 10.2, 10]. The median value is 10.2. The median values ​​of this type of field are sorted in ascending order to obtain [7.0, 10.2, 95]. Then, the interquartile range of each field is extracted, and its numerical fluctuation range is established. For example, if the Q1 value of the filling capacity field is 10.1 and the Q3 value is 10.3, then its fluctuation range is [10.1, 10.3]. Similarly, if the Q1 value of the "pH value" field is 6.9 and the Q3 value is 7.1, then its fluctuation range is [6.9, 7.1]. ​​This type of information, along with its corresponding field and specification information, forms a structured set of field numerical fluctuation ranges, as shown below.

[0063] Table 2: Definition of Field Numerical Ranges

[0064] Field Name Q1 value Q3 value Median Interval range Filling capacity 10.1 10.3 10.2 [10.1，10.3] pH value 6.9 7.1 7.0 [6.9，7.1] Clarity 94.0 96.5 95.0 [94.0，96.5]

[0065] As shown in Table 2, different fields were grouped and assigned their distribution ranges, which were subsequently used to calculate the difference in representative values.

[0066] S103: Based on the set of field numerical intervals, perform numerical summarization, normalization, and incremental evaluation of the median position values ​​of fields within the same group. This is done by calculating the distributional differences, absolute differences, and normalization ratio differences between fields, using the following formula:

[0067] ;

[0068] The representative difference values ​​of the fields are obtained through calculation. Based on the representative difference values ​​of the fields, the indicator benchmark table items are constructed, and the regulatory indicator reference benchmark value table is obtained.

[0069] in, The representative difference value represents the field's numerical dispersion within the categorization group; This indicates the median position value of the field. This indicates the median mean of the group to which this field belongs. This indicates the monitoring period for the corresponding field, in days. This represents the upper quartile value of the field. This represents the lower quartile value of the field. The length unit of the specification to which the field belongs; This indicates the normalized value of the field. This indicates the minimum normalized value in the grouping of this field. This represents the span of the normalized distribution corresponding to the normalized value. Number of grouping fields;

[0070] Based on the set of field value ranges, retrieve the midpoint value of the field. Group median mean Monitoring cycle Upper quartile lower quartile value Specifications and length Normalized value Minimum Normalized Value Normalized span Substitute into the following formula:

[0071] ;

[0072] in, This represents the summary of all field numbers i, j, and k in the group. The absolute value part reflects the symmetrical offset rate of the field differences, the square root term expresses the level of fluctuation under the specification limit, and the fractional term reflects the field offset under the normalization degree, reflecting multiple reduction processing.

[0073] The actual calculation is performed using 3 fields:

[0074] F1: , , , , , , , , ;

[0075] F2: =23.8, =22.6, T2=30, =24.0, =20.3, L2=1.5, =0.91, min(Vg)=0.82, R2=0.15;

[0076] F3: =21.9, =22.6, T3=30, =23.6, =20.5, L3=1.5, =0.82, min(Vg)=0.82, R3=0.15;

[0077] Calculate the numerator:

[0078] ;

[0079] Calculate the first term in the denominator:

[0080] ;

[0081] ;

[0082] The second term in the denominator is grouped into a portion:

[0083] ;

[0084] Substituting the final formula:

[0085] ;

[0086] The results indicate that the field's representative difference value is 0.00207, which is far below the classification difference threshold of 0.01. This value can be used as an assessment standard for the grouping stability of the field. If Mr exceeds twice this value in future batches of changes, such as above 0.004, it is determined that the field has a structural deviation and the regulatory indicator reference benchmark value table item needs to be updated.

[0087] The advantage of the formula is that by introducing the normalized median difference of the monitoring period, combined with the standard fluctuation of the specification dimension and the degree of normalization deviation, a composite difference identification mechanism is constructed, which makes the field grouping and representative quantitative assessment consistent and controllable.

[0088] The overall operational logic of this formula aims to comprehensively assess the representative differences of drug quality control fields through a multi-layered structure. The numerator is calculated by summing the absolute values ​​of the ratios of median differences to monitoring periods, representing the rate of deviation between each field and the mean of its category per unit time, thus reflecting the stability of the field over time. The first part of the denominator is calculated by multiplying the upper and lower quartile values ​​of a field, dividing by the specification length, summing the results, and then taking the square root. This square root operation reflects the magnitude of fluctuations in different fields under the same specification, and is used to balance the cumulative fluctuations of a large order of magnitude. The first part amplifies the overall evaluation value, thus reducing the volatility of the fields. The second part of the denominator is the sum of the deviation ratios of the field normalized values ​​relative to the minimum normalized value. This is divided by the number of fields and averaged to reflect the relative deviation of each field during the normalization process. The sum of the two parts constitutes the total denominator value, forming a "normalized volatility adjustment factor" in the overall structure. This normalizes the deviation behavior between fields to a structurally stable framework. Finally, the ratio of the numerator to the denominator is used to construct the "structural deviation rate," which measures the representativeness and consistency of the fields within the classification structure using numerical results.

[0089] The representativeness difference value of a field is used to measure the degree of structural deviation of a quality control field within its category (i.e., the same variety attributes and specification combination). It is an important quantitative indicator reflecting the representativeness and numerical stability of the field within the group. The lower the value, the closer the median value of the field is to the overall level of the group, and the stronger its attribution, making it suitable as a reference benchmark for the category. Conversely, the higher the value, the more significant the deviation of the field, which may have lost its representativeness in structure due to batch fluctuations, process differences, or abnormal sampling periods. Therefore, this indicator not only integrates the central trend and degree of variation of the field itself, but also comprehensively considers the influence of normalized distribution and time dimension, ultimately forming the core judgment basis for regulatory determination, field screening, and benchmark value revision.

[0090] Please see Figure 3 Step S2 is as follows:

[0091] S201: Based on the setting standards of each field in the regulatory indicator reference benchmark value table, compare the process parameter values ​​in the current batch of drug application to determine whether there are any deviation differences in the values ​​of the corresponding fields, record the fields with deviation differences, and obtain the number of difference fields identified.

[0092] First, extract all fields listed in the table. This table typically includes field number, field name, standard value range, and upper and lower limits. Taking the field "Concentration of effective ingredient in injection" as an example, its standard value range might be [95%, 105%]. Then, retrieve all process parameter data from the current batch of the drug application and match them according to the corresponding fields. For example, if the declared value for this batch is 93.6%, then the declared value of this field should be extracted and compared with the standard range to determine if there is any deviation. The determination method is that if the field value is less than the lower limit or greater than the upper limit, it is considered a deviation field. For such determinations, the comparison operation needs to be performed on each field to complete the process. In the initial difference determination, the total number of fields called during execution is set to n. After comparing each field, the total number of offset fields m is counted to obtain the final number of identified fields. At the same time, to enhance the reliability of the judgment, the trend comparison can be performed on the sampled values ​​at multiple time points in the batch. For example, if the collected values ​​at three time points t1, t2, and t3 are 94.1%, 93.6%, and 93.4% respectively, the moving average is calculated to be 93.7%. By comparing this with the lower limit of the standard, it can be determined that the field is continuously offset and further marked as a continuously offset field. In order to achieve statistical summary, all offset fields are encoded and the field code, field name, offset direction, offset value, etc. are recorded. Finally, the number of difference fields identified is obtained.

[0093] S202: Call the difference field identification quantity, classify the field belonging according to the indicator category, parameter attribute and offset direction of the difference field item, mark the classification type, and generate a field classification tag set;

[0094] The system retrieves the number of differentially identified fields obtained in the previous step, acquires the codes and names of all fields with offsets, and further extracts the corresponding indicator category and parameter attribute content. For example, if a field is "sterilization temperature," its associated indicator is "critical process parameter," and its parameter attribute is "time-sensitive." Then, the offset direction of the field item is determined. For instance, if the field's set value is 121℃ and the offset value is 123.8℃, the offset direction is positive. For each field, a "field-category-attribute-direction" quadruple is constructed, and classification is performed to assign it to the corresponding category. The classification criteria are as follows: If the field belongs to... If the field is in the "Key Control" category and the offset value is continuously increasing, it will be classified into the "Strong Monitoring Positive Bias" category. If it is in the "Auxiliary Indicator" category and the offset is occasional, it will be classified into the "Tolerance-Controllable Negative Bias" category, and so on. The classification logic is completed by dynamically combining the role attributes and values ​​of the fields. After execution, a classification label table is established for all fields. The label table contains structured fields such as field code, category number, direction identifier, and classification type code. For example, field number P02 (pH value), category is "Quality Control Item", offset direction is negative, and it is classified into the "Abnormal Fluctuation Negative Bias" group. After executing this operation, a field classification label set can be generated.

[0095] S203: Based on the field classification tag set, determine the offset magnitude and direction of each classified field item, and filter according to the change threshold in the field setting standard to obtain field items whose offset degree meets the identification conditions, and establish a declaration field offset comparison list.

[0096] Based on the established field classification tag set, the marked offset direction and classification type are called for each classified field item. At the same time, the offset degree of the field item is evaluated by combining the original set value of the field and the regulatory reference threshold. The magnitude is determined by the absolute value of the difference between the current offset value and the reference standard. For example, if "sterilization time" is set to 30 minutes and the declared batch is 33.2 minutes, the offset magnitude is 3.2 minutes. The reference threshold is set to ±2.5 minutes. The offset value of this field has exceeded the limit. Such fields that exceed the limit are further filtered and grouped by classification type. An offset field list is generated for fields of the same type. Different judgment ranges are set for different types of field items. For example, the offset threshold is set to ±3% for "time-sensitive" fields and ±5% for "content control" fields. After the offset judgment of each type of field, its offset range, field name, set value, current value and offset magnitude are summarized and recorded in the list. Finally, all field lists are integrated to form the declaration field offset comparison list.

[0097] Table 3 Example Data Table of Field Offset

[0098] Field Number Field Name Setting value Current value offset amplitude Offset threshold Offset direction Classification type F001 Injection concentration 100.0% 93.6% -6.4% ±5.0% negative bias Strong monitoring negative bias class F002 sterilization temperature 121℃ 123.8℃ +2.8℃ ±2.5℃ Positive and negative Strong monitoring of positive and negative classes F003 pH value 7.0 6.65 -0.35 ±0.3 negative bias Abnormal fluctuations with negative bias F004 sterilization time 30min 33.2min +3.2min ±3.0min Positive and negative Strong monitoring of positive and negative classes

[0099] Table 3 lists the numerical comparisons, judgment intervals, and classification types of several typical offset fields for actual comparison and filtering.

[0100] Please see Figure 4 Step S3 is as follows:

[0101] S301: Based on the difference fields marked in the declaration field offset comparison list, identify the key process parameter classification information bound to the record corresponding to each field, and map the classification information to the difference fields to construct the process parameter set corresponding to the offset fields and generate a field parameter matching set.

[0102] Based on the discrepancies marked in the declaration field offset comparison list, such as temperature control accuracy, pressure offset, and flow response, the original time-series data is extracted from the system equipment sensor acquisition records. Field values, sampling timestamps, and record numbers are extracted sequentially. Using the binding rules between field names and system process modules, the process module identifier to which each field belongs is extracted. For example, temperature control accuracy belongs to the thermal control subsystem, pressure offset to the sealing module, and flow response to the fluid subsystem. A mapping pair between field names and process tags is established. Subsequently, the value of each field within a continuous recording period is extracted from historical data records. For example, the continuous recorded values ​​for temperature control accuracy within one minute are 23.5℃, 23.95℃, 24.2℃, 24.55℃, and 24.90℃. The unit offset values ​​for each time period are 0.45, 0.25, 0.35, and 0.35℃, respectively. Comparing this to the pressure offset field, the recorded values ​​within a five-minute period are 2.1MPa, 1.38MPa, 2.0MPa, 2.3MPa, and 1.6MPa, with corresponding unit offset values ​​of 0.72, 0.62, 0.3, and 0.7MPa. The number of times the offset direction changes within each time period is also recorded. Each reversal of direction within the same field sample is counted as one change; for example, a change from rising to falling in the pressure offset field is counted as one change. By iterating through all the difference fields, statistical parameters such as field name, unit time offset value, offset amplitude value, and number of offset direction changes are established, and the original data table is generated as follows:

[0103] Table 4. Monitoring Data for Key Fields

[0104] Field Name Average unit time offset Number of directional changes Offset magnitude value (average) Directional continuity identifier Temperature control accuracy 0.35℃ / min 0 0.35℃ 1 Pressure offset 0.58 MPa / min 3 0.675MPa 0 Flow response 0.6L / min 1 0.42L / min 1

[0105] As shown in Table 4, the temperature control accuracy field maintains a stable upward trend during the monitoring period, with 0 directional changes and relatively stable offset values, with an average offset amplitude of 0.35℃. In contrast, the pressure offset direction changes 3 times, with the largest amplitude, corresponding to a continuity of 0. By mapping these field parameters to their bound process tags, and generating classification information of fields and key process parameters, the corresponding combination of field names, offset parameters, and process tags is finally output to obtain the field parameter matching set.

[0106] S302: Retrieves the field items from the matching set of field parameters, retrieves the time series record bound to the corresponding record, and obtains the range of field value changes, unit time offset, and number of field direction changes using the following formula:

[0107] ;

[0108] The calculation obtains the time change offset intensity value corresponding to the field, filters out fields whose change direction is continuous in the same direction and whose offset intensity shows an increasing trend, and obtains the enhanced offset signal.

[0109] in, This represents the enhanced offset signal value of field t under process category θ. This represents the unit time offset value of the field in the i-th record. This indicates the direction of field change for the i-th record, with 1 for upward and -1 for downward. Let be the offset value of the i-th record. This is the direction continuity indicator for the i-th record; 1 indicates consistent direction, and 0 indicates inconsistent direction. Let i be the number of times the field direction changes in the i-th record. This is a small constant to prevent division by zero (recommended value range: 0.001~0.01). The total number of records for the current field;

[0110] Retrieve the field monitoring data from the records in Table 4 and set its unit time offset value. Directional signs Offset range directional continuity Number of directional changes Substitute them into the calculation formula for the enhanced offset signal respectively:

[0111] This is the unit time offset value, obtained by dividing the difference between two consecutive data points by the time interval. For example, it takes 1 minute for the temperature control accuracy to increase from 23.5 to 23.95. ;

[0112] If the direction is upward, then Offset range The absolute value of the offset, for example ;

[0113] Directional continuity This indicates that the direction has not changed;

[0114] Number of directional changes This represents the number of times the field changes direction within the sequence; for example, temperature control accuracy is 0 and pressure offset is 3. Used to prevent division by zero in the denominator, and to ensure uniformity of units;

[0115] Taking the "Temperature Control Accuracy" field as an example, the first four sets of records are as follows:

[0116] ;

[0117] ;

[0118] ;

[0119] ;

[0120] ;

[0121] Substituting into the formula, we get:

[0122] molecular ;

[0123] The sum of squares term is: , take the square root ;

[0124] The denominator is: ;

[0125] Since the denominator is 0, This indicates that the offset direction of this field is absolutely stable and the intensity accumulation is significant, and it is marked as a strongly enhanced offset signal; compare with the "Pressure Offset" field:

[0126] ;

[0127] ;

[0128] ;

[0129] ;

[0130] ;

[0131] Then the molecule: ;

[0132] Sum of squares part: because ,but ;

[0133] Denominator: ;

[0134] final:

[0135] ;

[0136] The result is far below the system signal reference value of 0.85, and does not constitute enhanced offset behavior. Finally, fields with unidirectional trend and intensity superposition effect are selected to generate enhanced offset signals.

[0137] The overall operational logic of this formula can be understood as a composite evaluation mechanism that integrates directional offset intensity, amplitude cumulative effect, and change instability. First, the numerator consists of two parts: This represents the sum of the absolute values ​​of the unit time offset value multiplied by its direction indicator. The purpose is to unify and standardize the offset direction, accumulating all offset intensities through absolute value operations, reflecting the total offset intensity regardless of direction. This part constitutes the "trend intensity"; the subsequent term... Introducing the superposition of squares of the offset amplitude reflects the cumulative effect of the offset amplitude and relates it to directional continuity. Binding accumulates only the offset magnitude where the direction remains unchanged, making this term more sensitive to trend stability. Using the square root adjusts the order of magnitude of the offset value, maintaining a relative balance with the trend strength component and avoiding the amplification effect of extreme values; the denominator... Construct a direction-change weighted term; if the offset of a certain field changes frequently, i.e. If the value is large, its overall offset trend is unstable, and it is necessary to use... As a scaling factor, it can achieve a greater suppression effect on signals that change frequently but have small amplitudes. It constitutes a penalty factor for the trend instability of this field. The entire fractional expression has a larger value in the scenario where the offset is stable and the amplitude is enhanced, while it will be greatly reduced when the offset changes frequently and there is no continuous directional trend, forming a basis for distinguishing and judging the enhanced offset signal.

[0138] The time-varying offset strength value is a comprehensive numerical indicator used to quantify the stability of the offset trend and the cumulative degree of offset amplitude of a field in a continuous time series. It not only reflects the strength of the field's value change per unit time within the monitoring period, but also integrates factors such as whether the direction of change is consistent and the number of reversals during the change process. Specifically, when the value of a field maintains a single direction of change (such as continuous rise or continuous fall) in multiple consecutive time points, and the amplitude of each change gradually accumulates, its time-varying offset strength value will be significantly improved. Conversely, if the direction of change of the field is frequently reversed, even if the offset amplitude is large, the indicator will be reduced due to the discontinuity of direction and the penalty mechanism for the number of reversals. Therefore, this indicator comprehensively expresses the cumulative rate of offset value, the continuity of trend direction, and the stability of change pattern, and is a key judgment quantity for identifying whether a field has structural offset anomalies.

[0139] S303: Based on the identified fields in the enhanced offset signal, summarize the classification labels in the process classification, extract the field items whose signal strength is greater than the signal strength benchmark value, establish a three-item combination table of field name, classification label and enhanced offset signal value, and obtain the key field out-of-bounds judgment table.

[0140] Based on the calculation results of the enhanced offset signal, the following is calculated for each field: The value is compared with the system signal baseline value of 0.85. The system baseline value is obtained by statistically analyzing normal offset data under multiple operating conditions and is set as the offset intensity boundary of the 95% confidence interval of normal changes. The statistical method is to sort the offset signal values ​​of each field in 100 samples and take the 95th value as the threshold. The temperature control accuracy field has a calculated result of infinity, which is far beyond the 0.85 threshold and is judged as an out-of-bounds field. The pressure offset field has a value of 0.308, which is below the threshold and is not included as an anomaly. Finally, the field name, its process classification label (such as "thermal control system"), and the calculated enhanced offset signal value are extracted to form a triple. These three items are summarized according to the field order and output as the final key field out-of-bounds judgment table. This judgment table serves as the output basis of the system detection module and is used by the subsequent risk decision module for identification and control strategies.

[0141] Please see Figure 5 Step S4 is as follows:

[0142] S401: Based on the batch information of the fields listed in the key field boundary judgment table, extract the upstream batch number and downstream batch number corresponding to each field, and by matching the matching parameter group of the field in the different batches, call the matching parameter number and parameter attribute value of the field to identify the matching structure matching items between the different batches and generate the matching structure mapping matrix between batches.

[0143] Based on the batch information of the fields listed in the key field boundary judgment table, the corresponding parameter group numbers of each field in each batch are parsed to clarify the upstream and downstream batch numbers corresponding to each field. The parameter groups within each batch are retrieved by combining the field number information. Using the field number as the index, all parameter items configured for that field in each batch are called from the batch parameter record table. For example, field F001 corresponds to parameter group PGA-01 in batch A, with parameter values ​​{voltage=5.0V, current=1.2A, temperature=60℃}, and corresponds to parameter group PGB-03 in batch B, with parameter values ​​{voltage=5.2V, current=1.2A, temperature=60℃}. If the current is 1.2A and the temperature is 61℃, then this parameter group is compared with the set tolerance thresholds, where the voltage tolerance is ±0.3V, the current tolerance is ±0.1A, and the temperature tolerance is ±2℃. The differences between each parameter are calculated: the voltage deviation is 0.2V, the current deviation is 0A, and the temperature deviation is 1℃. If all are within the threshold range, then the field is determined to be a matching item. If the parameter group value of field F003 in batches A and B exceeds the above tolerance standards, then it is determined to be a mismatched item. Subsequently, the field number, upstream and downstream batch numbers, parameter group numbers of each batch, parameter value groups, and matching determination results are constructed into a structured table, as shown in the table below:

[0144] Table 5 Comparison of Corresponding Parameter Values ​​for Fields Across Batches

[0145] Field Number Upstream batch Parameter group number A Parameter value group A Downstream batches Parameter group number B Parameter value group B Is it compatible? F001 A PGA-01 5.0V, 1.2A, 60℃ B PGB-03 5.2V, 1.2A, 61℃ yes F002 A PGA-04 3.3V, 0.8A, 55℃ B PGB-08 3.1V, 0.9A, 53℃ yes F003 A PGA-10 2.5V, 0.5A, 40℃ B PGB-12 3.0V, 0.6A, 43℃ no

[0146] As shown in Table 5, the deviations of the corresponding parameter items in fields F001 and F002 between different batches do not exceed the set tolerance threshold, thus meeting the matching judgment conditions. The voltage parameter deviation in field F003 reaches 0.5V, which exceeds the set tolerance standard, so it is judged as an unmatched item. This table structure is output as the processing result, generating a matching structure mapping matrix between batches.

[0147] S402: Call the batch-to-batch matching structure mapping matrix to locate the transmission path of the field between upstream and downstream batches. By using the matching parameter values ​​of the nodes in the field mapping path, determine whether there is an interruption position in the transmission path of the field, identify the field jump segment and empty segment in the continuous path, and obtain the field transmission continuity breakpoint index.

[0148] The batch-to-batch matching structure mapping matrix is ​​invoked to continuously track the existence status of fields in each batch flow path, identify whether each field maintains a stable parameter matching relationship between adjacent batches, and construct a path identification linked list structure. Each field number corresponds to a batch number sequence, and the parameter matching status of the field in that batch is recorded using a Boolean label of "matching status". For example, field F002 is bound to parameter groups PGA-04, PGB-08, PGC-09, and PGD-06 in batches A, B, C, and D respectively, and its matching determination results are {yes, yes, no, yes}. Based on this linked list structure, the fields at the path breakpoints are extracted. The starting and ending positions of consecutive "No" labels are used to determine the interruption interval and calculate the number of batches and breaks in the path. Further analysis of field F003 shows that its corresponding labels in batches A, B, C, D, and E are {Yes, No, No, Yes, No}. Therefore, the interruption nodes appear in batches B, C, and E, with a field break frequency of 3 and a total of 5 batches, resulting in an interruption rate of 60%. The breakpoint sequence {B, C, E} is recorded. A set of breakpoint indicators is established by combining information such as field number, path sequence, interruption rate, and breakpoint distribution. Finally, the continuity status of all field paths is summarized and output in a structured manner to form a field transmission continuity breakpoint indicator.

[0149] S403: Based on the field transmission continuity breakpoint index, extract the field mapping path segment where the interruption node is located, calculate the path span, number of transmission parameter groups, field state variability and mean offset of the transmission value sequence within the path segment, and obtain the cross-batch integrity of the field mapping chain through joint analysis of the indexes, and establish a cross-batch field mapping chain integrity structure diagram.

[0150] Based on the field transmission continuity breakpoint index, fields with broken paths are screened, and their corresponding transmission path segments are extracted. The path span, number of transmission parameter groups, field state variability, and transmission value offset are calculated for each segment. The path span is obtained by subtracting the start and end batch index numbers of the path segment. For example, field F003 has an interrupted path from batch B to E, with a span of 4 and 3 transmission parameter groups. By extracting the field values ​​for each batch, we obtain {3.0V, —, 2.9V, 2.8V}. After removing missing values, the mean μ is calculated to be 2.9V, the maximum value is 3.0V, and the minimum value is 2. The average offset is 0.07V, calculated by the offset between the field parameter values ​​within the path segment and the mean of that segment. After summarizing the above multiple indicators, the structure is identified according to the field number and expanded horizontally to each batch sequence. The path integrity level of each field is marked vertically to construct a structure diagram representation. Each field in the diagram is labeled with a number, the batch sequence is used as the coordinate axis, the breakpoint position is marked in red, and the integrity level is divided by the level score. This comprehensively shows the degree of interruption and stability of the field in the batch transmission process, thereby establishing a cross-batch field mapping chain integrity structure diagram.

[0151] Please see Figure 6 The S5 steps are as follows:

[0152] S501: Based on the interrupted nodes in the cross-batch field mapping chain integrity structure diagram, extract the field number and corresponding batch, call the field distribution sequence table and parameter structure, determine whether there is a missing state, and generate a list of missing field identifiers.

[0153] Based on the identified interrupted nodes in the cross-batch field mapping chain integrity structure diagram, field numbers and their corresponding batch indexes are extracted. The field distribution sequence table is then queried line by line using the field number to locate the field in the specified batch. Conditions for determining a missing field include: a null field value, a null value identifier, or the field not being activated in the current batch. This is identified using the status label field in the data item structure. After identification, the parameter group index value to which the field belongs is further located based on the batch number. This clarifies whether a valid parameter item exists at the field position that should be filled in within that batch. In practice… If field F_017 is an interruption node in batch B, and no corresponding data or a null status value is found in the parameter records of this field in batch B, indicating a missing record, the result is a missing field. Field number F_017 is combined with batch number B to form a structural index item, constructing a list of basic information for the missing field. If multiple fields exhibit the same interruption status, each record is identified and compared against its missing field structural status. Records with missing fields but not interrupted status are removed to ensure the consistency of the field list structure and batch validity. The table structure is generated by combining the structural statuses of multiple field items, as shown below:

[0154] Table 6. Example Table of Missing Field Identification

[0155] Field Number Batch number Field status Parameter group number Missing markers F_017 B null PG_B12 yes F_028 C null PG_C05 yes F_011 D efficient PG_D03 no

[0156] As shown in Table 6, fields F_017 and F_028 have a clear missing status in their respective batches and are both recorded as missing fields. Although field F_011 is an interrupted node, the data exists, so it is not included in the missing list. Finally, a list of missing field identifiers is generated.

[0157] S502: Call the list of missing field identifiers, combine the field filling rules with the status of upstream and downstream batch data, determine the compliance of missing fields with constraint requirements, and obtain the sequence of missing field breakpoint indicators;

[0158] The system retrieves the list of missing field identifiers, reads the field number and corresponding batch of each field, and searches the field entry rule configuration table by field number to identify the constraint level and type of each field. Field constraint types are categorized into three types: required, format-dependent, and logically dependent. Required fields must have valid entries in all batches; format-dependent fields must meet format conventions; and logically dependent fields must be valid in conjunction with other field values. If field F_017 is a required field, its absence immediately constitutes a breakpoint exception. The system further links the field number to upstream and downstream batches to determine if the field is present. If the data has been submitted but is missing in the current batch, the interruption level of the record field transmission chain is high. By comparing the current state of the field with the state of the upstream and downstream, the risk level of the interruption of the declaration chain caused by the missing field is assessed, and the field constraint strength is matched and judged. If field F_028 is a logically dependent field, although it is missing, it is not submitted in the upstream and downstream batches, so its breakpoint impact level is low. Finally, the matching relationship between the field submission rule strength and the batch transmission chain breakpoint degree is compared to form a field missing impact level sequence, and a field missing breakpoint index sequence is generated.

[0159] S503: Based on the sequence of missing field breakpoint indicators, filter out abnormal fields, extract field numbers, missing locations and attribute contents, organize structural records by batch, and establish a list of missing field breakpoints in the declaration.

[0160] Based on the sequence of missing field breakpoint indicators, extract the field numbers whose breakpoint indicator values ​​exceed the field reporting baseline strength. Organize the field numbers with their corresponding batches, missing statuses, and constraint types. Establish a field index mapping structure through field numbers and batch information, construct field missing breakpoint record units, classify all records by batch number, sort by field number, and write them into the missing field summary structure record table. Mark the field missing level and matching status in the records. At the same time, establish the total number of missing field records and the list of key breakpoint fields for each batch. Construct a unique field breakpoint mapping through the combination key of field and batch, and output the structure record list. The field missing level is divided into three levels: high, medium, and low. Mark the current impact level of the field. Finally, summarize the field breakpoint data for all batches and construct the declaration field missing breakpoint list.

[0161] Drug compliance testing systems used for drug regulation include:

[0162] The benchmark indicator construction module obtains the quality control field values, groups the fields by generic name and dosage form parameters, calculates the median, first quartile, third quartile and interquartile range of the field group, filters out records that deviate from the median value by more than twice the interquartile range, calls the data after filtering out to calculate the median, maximum and minimum values, establishes the field fluctuation range, and generates a regulatory indicator reference benchmark value table.

[0163] The field offset identification module is based on the regulatory indicator reference benchmark value table. It calls the process parameter field values ​​of the current batch of declarations, compares the field values ​​with the median fluctuation range, filters fields that exceed the range and marks the offset direction, classifies them according to the offset magnitude and field type, and generates a declaration field offset comparison list.

[0164] The trend boundary judgment module is based on the declaration field offset comparison list, calls the key process parameter category information of the offset field, extracts the numerical sequence of the field in consecutive batches, calculates the consistency of change direction and offset increase trend coefficient, filters trend enhancement fields, and generates a key field boundary judgment table.

[0165] The field mapping chain identification module is based on the key field out-of-bounds judgment table. It calls the corresponding field parameter groups of upstream and downstream of the corresponding batch, analyzes the numerical correspondence between upstream and downstream fields, determines whether the numerical transmission is interrupted, extracts the interruption position node, and generates a cross-batch field mapping chain integrity structure diagram.

[0166] The missing breakpoint extraction module is based on the cross-batch field mapping chain integrity structure diagram. It calls the field filling rule list, compares the correspondence between the interrupted node fields and the required fields, filters the missing fields, and generates a list of missing breakpoints for the declared fields.

[0167] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A drug compliance testing method for drug regulation, characterized in that, Includes the following steps: S1: Extract the quality control fields from the drug manufacturing process change data, classify them according to product attributes and specifications, remove records that deviate from the median value by more than twice the interquartile range, group the retained fields, extract the median value and range of variation, and generate a regulatory indicator reference benchmark value table. S2: Based on the field standards set in the regulatory indicator reference benchmark value table, compare the process parameter values ​​in the current batch of drug applications, mark the fields with differences and classify and summarize them to generate a list of application field offset comparisons; S3: Based on the difference fields in the declaration field offset comparison list, combined with the classification of key process parameters, retrieve the change trend of the corresponding records, identify field items with consistent direction and continuously expanding offset, and generate a key field boundary judgment table. S4: Based on the field batch information in the key field boundary judgment table, extract the matching parameters of upstream and downstream batches, construct the field mapping path, identify the location where value transmission is interrupted, and generate a cross-batch field mapping chain integrity structure diagram. S5: Based on the interruption nodes in the cross-batch field mapping chain integrity structure diagram, and in conjunction with the field filling rules, determine whether the corresponding field is missing, and generate a list of missing declaration field breakpoints.

2. The drug compliance testing method for drug regulation according to claim 1, characterized in that, The regulatory indicator reference benchmark value table includes representative median values, normal variation ranges, and field grouping structures. The declaration field offset comparison list includes segment comparison results, offset field list, and difference field markers. The key field boundary judgment table includes offset trend fields, change direction continuity fields, and key process difference fields. The cross-batch field mapping chain integrity structure diagram includes field mapping paths, interruption node positions, and batch matching parameter groups. The declaration field missing breakpoint list includes interruption node fields, filling rule items, and specification compliance fields.

3. The drug compliance testing method for drug regulation according to claim 1, characterized in that, The S1 step is as follows: S101: Extract the quality control field content from the drug manufacturing process change data, classify the fields according to the product attributes and specifications, remove all records in all fields whose values ​​deviate from the median position by more than twice the interquartile range within the field distribution, obtain the numerical sequence of each field after removing anomalies, and generate the standardized numerical sequence of the fields. S102: Based on the field standard numerical sequence, perform numerical grouping processing on fields under the same category, call the variety attributes, specifications, field values ​​and field names corresponding to the fields, sort the median position of each group of field values, and calculate the upper and lower fluctuation ranges in combination with the interquartile range to generate a set of field numerical intervals. S103: Based on the set of field value intervals, the median position values ​​of the same group of fields are numerically summarized, normalized, and incrementally evaluated. By calculating the distribution difference, absolute difference, and normalization ratio difference between fields, the representative difference value of the fields is obtained. Based on the representative difference value of the fields, the indicator benchmark table is constructed to obtain the regulatory indicator reference benchmark value table.

4. The drug compliance testing method for drug regulation according to claim 1, characterized in that, The S2 step is as follows: S201: Based on the setting standard of each field in the regulatory indicator reference benchmark value table, compare the process parameter values ​​in the current batch of drug application, determine whether there is a deviation difference in the value of the corresponding field, record the field items with deviation differences, and obtain the number of difference field identifications. S202: Invoke the number of differences in the field identification, classify the field according to the index category, parameter attribute and offset direction of the differences in the field item, mark the classification type, and generate a field classification mark set; S203: Based on the field classification tag set, determine the offset magnitude and direction of each classification field item, and filter according to the change threshold in the field setting standard to obtain field items whose offset degree meets the identification conditions, and establish a declaration field offset comparison list.

5. The drug compliance testing method for drug regulation according to claim 1, characterized in that, The S3 step is as follows: S301: Based on the difference fields marked in the declaration field offset comparison list, identify the key process parameter classification information bound to the record corresponding to each field, and map the classification information to the difference fields to construct the process parameter set corresponding to the offset fields and generate a field parameter matching set. S302: Call the field items in the field parameter matching set, retrieve the time series records bound in the corresponding records, obtain the change range of the field value, the unit time offset value and the number of changes in the field direction, obtain the time change offset intensity value corresponding to the field, filter out the fields whose change direction is continuous in the same direction and whose offset intensity shows an increasing trend, and obtain the enhanced offset signal. S303: Based on the identified fields in the enhanced offset signal, summarize the classification labels in the process classification, extract the field items whose signal strength is greater than the signal strength benchmark value, establish a three-item combination table of field name, classification label and enhanced offset signal value, and obtain the key field out-of-bounds judgment table.

6. The drug compliance testing method for drug regulation according to claim 1, characterized in that, The S4 step is as follows: S401: Based on the batch information of the fields listed in the key field boundary judgment table, extract the upstream batch number and downstream batch number corresponding to each field, and by matching the matching parameter group of the field in the different batches, call the matching parameter number and parameter attribute value of the field to identify the matching structure matching items between the different batches and generate the matching structure mapping matrix between batches. S402: Call the batch-to-batch matching structure mapping matrix to locate the transmission path of the field between upstream and downstream batches. By using the matching parameter values ​​of the nodes in the field mapping path, determine whether there is an interruption position in the transmission path of the field, identify the field jump segment and empty segment in the continuous path, and obtain the field transmission continuity breakpoint index. S403: Based on the field transmission continuity breakpoint index, extract the field mapping path segment where the interrupted node is located, calculate the path span, number of transmission parameter groups, field state variability and mean offset of the transmission value sequence within the path segment, and obtain the cross-batch integrity of the field mapping chain through joint analysis of the indexes, and establish a cross-batch field mapping chain integrity structure diagram.

7. The drug compliance testing method for drug regulation according to claim 1, characterized in that, The S5 step is as follows: S501: Based on the interruption node in the cross-batch field mapping chain integrity structure diagram, extract the field number and corresponding batch, call the field distribution sequence table and parameter structure, determine whether there is a missing state, and generate a list of missing field identifiers. S502: Call the field missing identifier list, combine the field filling rules and the status of upstream and downstream batch data, determine the compliance of the missing fields with the constraint requirements, and obtain the field missing breakpoint index sequence; S503: Based on the sequence of missing field breakpoint indicators, filter out abnormal fields, extract field numbers, missing locations and attribute contents, organize the structure records by batch, and establish a list of missing field breakpoints for the application.

8. A drug compliance testing system for drug regulation, characterized in that, The system is used to perform the method according to any one of claims 1-7, comprising: The benchmark indicator construction module obtains the quality control field values, groups the fields by generic name and dosage form parameters, calculates the median, first quartile, third quartile and interquartile range of the field group, filters out records that deviate from the median value by more than twice the interquartile range, calls the data after filtering out to calculate the median, maximum and minimum values, establishes the field fluctuation range, and generates a regulatory indicator reference benchmark value table. The field offset identification module, based on the regulatory indicator reference benchmark value table, calls the process parameter field values ​​of the current batch of declarations, compares the field values ​​with the median fluctuation range, filters fields that exceed the range and marks the offset direction, classifies them according to the offset magnitude and field type, and generates a declaration field offset comparison list. The trend boundary judgment module, based on the declaration field offset comparison list, calls the key process parameter category information of the offset field, extracts the numerical sequence of the field in consecutive batches, calculates the consistency of change direction and offset increase trend coefficient, filters trend enhancement fields, and generates a key field boundary judgment table. Based on the key field out-of-bounds judgment table, the field mapping chain identification module calls the corresponding field parameter group of upstream and downstream of the corresponding batch, analyzes the numerical correspondence between upstream and downstream fields, determines whether the numerical transmission is interrupted, extracts the interruption position node, and generates a cross-batch field mapping chain integrity structure diagram. The missing breakpoint extraction module, based on the cross-batch field mapping chain integrity structure diagram, calls the field filling rule list, compares the correspondence between the interrupted node fields and the required fields, filters the missing fields, and generates a list of missing breakpoints for the declared fields.

Images