A system for monitoring operational and operational risks in the entry of sample information

Abstract

The application relates to the technical field of inspection information system management, in particular to an inspection sample information input, operation and operation risk monitoring system, which comprises a personnel information collection device used for collecting the unique identity codes of all staff in an inspection system; and an inspection operation definition device used for defining the information codes of all inspection items in the inspection system, wherein the information codes are generated based on the paths of execution operations of the inspection items, and the paths of execution operations include the sample acquisition mode, sample transfer mode, sample processing mode and sample inspection mode; the system collects the inspection information in real time, fuses the standardized inspection operation path codes and unique identity codes, constructs multi-dimensional inspection features, extracts time perception features for revealing the process timeliness and operation mode from the multi-dimensional inspection features, and performs deep analysis in combination with the personnel information knowledge graph; and finally, the two types of key risks are actively identified and warned based on the features.

Description

Technical Field

[0001] This application relates to the field of inspection information system management technology, and more specifically, to an inspection sample information input, operation and operation risk monitoring system. Background Technology

[0002] The content in this section provides only background information related to this application and may not constitute prior art.

[0003] Medical test results are crucial for disease diagnosis, and the initial assessment of many diseases relies on the analysis of various test indicators. Existing medical testing systems typically include four core stages: sample acquisition, sample testing, sample verification, and result generation. Among these, the sample verification stage primarily relies on manual review of test results, focusing on verifying sample source information and testing methods to prevent operational errors.

[0004] However, as the final step in the entire testing process, verification work essentially only checks the compliance of the previous step (sample testing). For example, for a blood lead content test report, the verifier can usually only confirm whether the testing method meets the requirements. The design and positioning of this step makes it difficult to examine and identify systemic risks throughout the entire testing process from a holistic perspective.

[0005] Such systemic risks are rarely isolated events, but rather stem from interconnected errors across multiple stages. For example, insufficient centrifugation time before testing might lead to inaccurate results in a batch of blood samples, or large quantities of samples might not be handled with strict temperature control during transport and storage. Once such problems occur at a critical stage due to negligence on the part of the responsible party, they can easily trigger widespread, interconnected oversights, causing systemic deviations in the test results of the entire batch or even the entire system. This not only directly affects the accuracy of individual tests but also severely disrupts the effective conduct of subsequent medical diagnostic work. Summary of the Invention

[0006] The summary section of this application is intended to provide a brief overview of the concepts, which will be described in detail in the detailed description section below. This summary section is not intended to identify key or essential features of the claimed technical solutions, nor is it intended to limit the scope of the claimed technical solutions.

[0007] Some embodiments of this application propose a system for monitoring the entry, operation, and risk of testing sample information to address the technical problems mentioned in the background section above.

[0008] As a first aspect of this application, some embodiments of this application provide a system for monitoring the entry, processing, and operation of test sample information, including:

[0009] Personnel information collection device, used to collect the unique identification code of all staff in the inspection system;

[0010] The inspection operation definition device is used to define the information code of all inspection items in the inspection system. The information code is generated based on the execution operation path of the inspection item. The execution operation path includes the sample acquisition method, sample transportation method, sample processing method, and sample inspection method.

[0011] The inspection information acquisition device acquires inspection information generated by the inspection system in real time, and generates inspection features for each piece of inspection information based on information coding and identity tag coding.

[0012] A feature extraction device acquires the inspection features of the same inspection items, and extracts the time-aware features of each inspection item from all inspection features. The time-aware features include the time statistical features and time structure features of each inspection item.

[0013] Personnel information extraction device, which establishes an information knowledge graph for all staff members;

[0014] The risk clustering device generates risk warning information about systematic inspection errors based on the time-aware characteristics of each inspection item.

[0015] Based on the information knowledge graph and all inspection features, risk warning information is generated to prevent staff from coordinating and assigning tasks incorrectly.

[0016] This system collects inspection information in real time and integrates standardized inspection operation path codes and unique identification codes to construct multi-dimensional inspection features. It then extracts time-aware features that reveal process timeliness and operational patterns, and conducts in-depth analysis using a personnel information knowledge graph. Based on these features, it proactively identifies and warns of two key risks: 1) Systemic inspection errors (such as batch sample processing deviations) indicated by abnormal time-aware features (e.g., timeouts, sequence disorder); 2) Staff coordination and allocation errors determined by the knowledge graph and inspection features (e.g., unqualified personnel, task overload). Simultaneously, it integrates multi-dimensional features to generate a comprehensive risk level, enabling intelligent assessment of the reliability of individual inspection information. This transforms passive verification into proactive early warning, significantly improving the ability to detect systemic risks throughout the entire inspection process and enhancing the reliability of results.

[0017] In existing inspection systems, personnel identification is usually limited to simple IDs or names, which cannot effectively embed and represent their inspection skills and corresponding skill levels. Storing or associating complex skill data additionally would lead to data redundancy and a loose structure, making it difficult to quickly and efficiently link personnel qualifications and specific operational tasks in subsequent process monitoring and risk assessment. This increases the system's processing burden and reduces the timeliness of risk identification.

[0018] Furthermore, the personnel information collection device includes:

[0019] The identity information acquisition module is used to obtain the identity IDs of all inspection personnel;

[0020] The skill tag acquisition module retrieves the corresponding skill tags for each employee. The skill tags include the testing skills they have mastered and the corresponding skill levels.

[0021] The identity coding generation module generates a unique identity identifier code for each employee based on their identity ID and skill tag.

[0022] This solution innovatively integrates an employee's identity ID with their skill tags (including inspection skills and levels), forming a compact and self-contained unique code. This design inherently embeds key information about an employee's work capabilities, enabling the system to instantly identify the type of inspection skills and proficiency level an employee can perform based on the code itself, without requiring additional queries or complex data associations. This significantly improves the efficiency of personnel capability matching and provides an efficient and structured data foundation for subsequent risk assessment.

[0023] Furthermore, the identification code is generated as follows: a unique and sequentially increasing number range is generated for each inspection skill level; the range of the number range is equal to the number of inspectors corresponding to the skill level of that inspection skill, and all number ranges are obtained to generate a number sequence;

[0024] Each staff member is arranged according to their identity ID, and then each element in the numbering sequence is matched one-to-one with the skill level corresponding to the aforementioned inspection skills. The staff member's serial number in the numbering sequence is the identity identification code.

[0025] The identity identifier generated by this scheme is a compact one-dimensional sequence, which significantly reduces the data dimensionality and greatly reduces the computational overhead of subsequent feature extraction and risk calculation. At the same time, the code cleverly embeds the inspection skills and level information possessed by the personnel. Their skill qualifications can be intuitively mapped by their position in the number sequence, providing an efficient and structured data foundation for subsequent accurate association of personnel factors and operational risks, balancing computational efficiency and information representation capabilities.

[0026] Furthermore, the methods for obtaining samples include: venous blood collection, throat swab collection, urine collection, and tissue biopsy;

[0027] The methods of sample transport include: pneumatic pipeline transport, cold chain transport box, biosafety container, and manual timed delivery for testing;

[0028] Sample processing methods include: centrifugation, anticoagulation, frozen sectioning, and Gram staining;

[0029] The methods for testing samples include: mass spectrometry, microscopic slide reading, rapid detection, and biochemical and immunoassay analyzer detection.

[0030] Existing inspection systems typically record operational steps, personnel, and time information in natural language or discrete data, resulting in unstructured or weakly structured inspection information. This format makes efficient computational analysis difficult, and traditional methods often rely on complex model training to convert it into machine-processable vector representations. This process is not only cumbersome but also struggles to simultaneously and effectively preserve the complete correlation between the three key types of information: operational sequence logic, personnel identity and skills, and temporal dynamics. This hinders the rapid extraction of implicit features and risk analysis.

[0031] Furthermore, the inspection information acquisition device includes:

[0032] The inspection information collection module is used to collect inspection information uploaded by the inspection system;

[0033] The inspection information extraction module extracts the steps involved in generating the inspection information, the personnel involved in each step, and the duration of each step. The steps involved in generating the inspection information include, in sequence: the sample acquisition method, the sample transportation method, the sample processing method, and the sample inspection method.

[0034] The information matrix generation module replaces the steps of generating verification information with the corresponding information codes to generate the first sequence; it then replaces the staff members corresponding to the verification information with their identification codes to generate the second sequence.

[0035] Using the first sequence as the x-axis and the second sequence as the y-axis, and the duration corresponding to each step as the height value, a feature polyline is generated in the three-dimensional coordinate system. The feature polyline is then used as the verification feature.

[0036] This solution achieves an efficient and intuitive vectorized representation (inspection features) of inspection information by structurally mapping the operational steps (as information encoding sequences), the personnel performing the inspections (as identity identification encoding sequences), and the duration of each step to a three-dimensional coordinate space and generating feature polylines. This vectorization method does not rely on model training and can directly and synchronously encode and retain the three core dimensions of information: inspection project logic, personnel capability identification, and time dynamic features.

[0037] Existing inspection systems, when analyzing process timeliness, often focus only on single or aggregated time indicators (such as total time), lacking in-depth analysis of the time distribution patterns within inspection steps and the time correlation patterns between steps. This coarse-grained analysis makes it difficult to accurately capture potential risks caused by delays in specific steps, abnormal operating rhythms, or problems in the connection between steps, resulting in a one-sided and insensitive assessment of the timeliness risks of the inspection process.

[0038] The feature extraction device includes:

[0039] The inspection feature acquisition and classification module is used to acquire inspection features and classify them according to the inspection items they belong to, resulting in several inspection item groups;

[0040] The time statistical feature extraction module is used to extract time statistical features that characterize the inspection time of each inspection step from the inspection features of the inspection project group.

[0041] The time structure feature extraction module is used to extract time structure features from the inspection features of the inspection project group to characterize the changes in inspection time of each inspection step.

[0042] The feature fusion module is used to fuse time statistical features and time structure features to generate time-aware features that characterize each test item group.

[0043] This solution groups the characteristics of similar inspection items and extracts time statistical features (step-level time distribution) and time structure features (time correlation and change patterns between steps) in parallel, ultimately fusing them to generate a multi-dimensional "time-aware feature." This feature comprehensively depicts the actual time characteristics of the inspection process from both local (individual step) and global (inter-step structure) levels. It can keenly identify deep-seated time sequence anomalies such as systematic timeouts in specific steps, disordered operation rhythms, and poor step connections, providing richer and more discriminative quantitative evidence for subsequent accurate assessment of systemic timeliness risks.

[0044] Traditional inspection systems typically calculate the average time taken over a single time span (such as the whole or a fixed period) when assessing process timeliness. This method cannot effectively capture the dynamic trends of inspection efficiency across different time scales, and it is particularly difficult to identify systemic timeliness risks that accumulate slowly or fluctuate periodically, resulting in delayed and insensitive early warnings of potential timeliness problems.

[0045] Furthermore, the time statistical characteristics F h.f The extraction method is as follows:

[0046] ;

[0047] in, It represents the total average time of all test features in the same test item group within the first time span; This represents the total average time for all test features within the same test item group during the second time span. This represents the total average time of all test features within the same test item group over the third time span. This represents the total average time of all test features in the same test item group within the fourth time span; the first, second, third, and fourth time spans gradually increase, f represents the index of the test item group, and h represents the index of the time statistical feature.

[0048] This scheme calculates the average time taken for the same inspection item across multiple increasing time spans and multiplies these averages element-wise to generate a comprehensive time statistical feature. This feature innovatively integrates short-, medium-, and long-term time efficiency information, reflecting not only immediate efficiency but also amplifying and highlighting efficiency change patterns and abnormal trends (such as continuous declines and periodic fluctuations) at different time scales. This significantly enhances the early detection capability for gradual or periodic timeliness risks, providing a more comprehensive and forward-looking quantitative indicator for early warning of systemic timeliness deviations.

[0049] Existing inspection systems, when assessing process risks, often limit their analysis of time-dimensional anomalies to static statistics (such as anomaly counts, mean / variance), making it difficult to quantify the dynamic trends of anomaly changes (such as whether anomalies are accelerating or fluctuating randomly). This static perspective cannot effectively capture the key inflection points (risk thresholds) in the evolution of the system state, resulting in delayed and insensitive early warnings of potential systemic collapse risks, and failing to provide sufficient time for intervention before the risks actually materialize.

[0050] Furthermore, the temporal structural feature F u.f The extraction method is as follows:

[0051] ;

[0052] Indicates first-order structural features, Indicates second-order structure characteristics;

[0053] ;

[0054] ;

[0055] Where u represents the time structure feature F u.f The corresponding time interval, g represents the time period, u > 2g; i represents the index of the time period, and q(i) represents the number of test features that show anomalies in the i-th time period;

[0056] The test characteristics are determined based on the activity range of historical normal test characteristics, where f represents the index of the test item group.

[0057] This innovative approach calculates the first-order difference ("velocity," reflecting the rate of anomaly growth / decline) and the second-order difference ("acceleration," reflecting the trend of rate change) of the number of anomaly test features, and then multiplies them element-wise. This feature dynamically transforms the time dimension into a quantitative indicator characterizing changes in "risk potential," enabling it to keenly capture the critical state of accelerated accumulation or abrupt trend changes in the number of anomalies within a specific time interval. Compared to static statistics, it reveals the dynamic signals of systemic test errors brewing or erupting earlier and more accurately, providing crucial evidence for proactive risk intervention.

[0058] Traditional laboratory risk monitoring systems often employ rigid, single-dimensional threshold rules (such as fixed time / personnel restrictions) when identifying operational anomalies. This approach struggles to adapt to the flexibility of actual hospital operations (such as different transport / pretreatment methods, multi-positional staff, and reasonable time fluctuations), and is highly susceptible to numerous false positives due to overly stringent rules, interfering with effective risk identification. Conversely, excessively relaxing the rules to improve adaptability significantly reduces the ability to detect actual operational deviations, resulting in a dilemma where sensitivity and applicability are difficult to balance.

[0059] Furthermore, the method for judging abnormal characteristics includes the following steps:

[0060] S1: Obtain the x-coordinate and y-coordinate of each data point in the test feature, and set the corresponding x-coordinate range and y-coordinate range for the test features generated for each type of test item.

[0061] The x-coordinate of each data point is compared with the x-coordinate range in turn. If the x-coordinate of at least one data point is not within the preset x-coordinate range, then the test feature is set as an anomaly test feature.

[0062] The ordinate of each data point is compared with the range of ordinates in turn. If the ordinate of at least one data point is not within the preset range of ordinates, then the test feature is set as an anomaly test feature.

[0063] S2: Set the corresponding feature space for the inspection features generated for each type of inspection item;

[0064] If at least one data point exists outside the feature space, then this test feature is set as an anomaly test feature.

[0065] This solution combines a multi-dimensional flexible verification mechanism (an adjustable horizontal axis range adapts to different testing procedures, an adjustable vertical axis range accommodates overlapping personnel functions, and an expandable feature space to accommodate reasonable time fluctuations) with spatial distribution verification (whether data points exceed preset ranges / spaces), achieving high adaptability and low false alarm detection for abnormal testing characteristics. This design effectively balances the flexibility required for actual hospital operations with the precision required for risk monitoring. While significantly reducing invalid alarms triggered by reasonable operational differences, it ensures reliable capture of substantial operational deviations. Subsequent analysis of these "flexibly screened" anomalies, combined with temporal structure feature analysis, allows for more focused and reliable identification of potential systemic testing risks.

[0066] Existing inspection systems typically manage personnel skill information in a limited way, limited to discrete, isolated records (such as simple lists or database entries), failing to clearly and structurally express the multidimensional relationships between staff and required inspection attributes (skills and levels). This fragmented storage method hinders the system from quickly and accurately analyzing the match between personnel capabilities and inspection task requirements.

[0067] Furthermore, the information knowledge graph G = (V, E);

[0068] V=P∪S;

[0069] P = {p1, p2, ... p} n}, S = {s1, s2, ... s} m}, where P represents the set of identification codes for all staff members, n represents the total number of staff members, and p n Let s represent the identification code of the nth employee, S represent the total number of all inspection attributes, and s m This represents the m-th test attribute, where m represents the total number of test attributes. Test attributes include test skills and skill levels.

[0070] ;

[0071] R is a set of relations, where (p, s) represents a pair of staff and test attributes, (p... i s j This indicates that the i-th employee has mastered the j-th inspection attribute, (p i s j )∈R, Represents subset notation, E represents all conditions that satisfy the condition. The composition of (p, s).

[0072] This solution innovatively constructs a knowledge graph that uses staff and inspection attributes as nodes, and precisely models the "mastery" relationships through relation sets, forming a structured and computable knowledge network of personnel capabilities. This graph intuitively and completely depicts the inspection attributes mastered by each staff member and their proficiency level, enabling the system to efficiently query personnel skills, accurately assess the matching degree between tasks and personnel qualifications, and deeply analyze the personnel skill distribution map. This proactively identifies systemic work risks (such as process bottlenecks and quality hazards) caused by potential factors such as skill gaps, level mismatches, or unreasonable task allocation.

[0073] Risk clustering devices include:

[0074] The systemic risk extraction module uses the support vector machine algorithm to perform cluster analysis on time-aware features, generate time-aware vectors, and map the time-aware vectors to generate risk warning information.

[0075] The personnel allocation extraction module obtains all inspection features within a certain period, extracts the relationship mapping information from the inspection features, compares the relationship mapping information with the mapping information in the information knowledge graph, and generates risk warning information for personnel coordination and allocation errors.

[0076] Furthermore, the risk warning information is generated in the following ways:

[0077] Z1: Prepare datasets in advance. Each dataset includes several samples and their labels. The samples are time-aware features t. i , tag y i The risk warning level is {-1, 0, 1}, which includes 3 level labels, where i represents the index of the sample.

[0078] Z2: Construct a support vector machine model to find the optimal decision function;

[0079] ;

[0080] Where w and b are the parameters to be determined, w represents the weight vector, and b represents the bias term. T represents the feature space; T represents the matrix bias sign.

[0081] Construct the primal optimization problem: ;

[0082] The constraints are: ;

[0083] e i This represents the prediction error for the i-th sample. represents the regularization parameter, N represents the number of samples, and i represents the sample index;

[0084] Lagrange function :

[0085] ;

[0086] Represents the Lagrange multipliers;

[0087] Find the KKT conditions;

[0088] Condition 1: , ;

[0089] Condition 2: , ;

[0090] Condition 3: , ;

[0091] Condition 4: , T represents the matrix bias symbol; Indicates the sign of the partial derivative;

[0092] The constraint equations introducing KKT conditions are as follows: ;

[0093] The kernel function is: ;

[0094] , j represents the index of the sample;

[0095] Transform the constraint equations into matrix form:

[0096] 1 represents a vector of all 1s, I represents the identity matrix, and K represents the kernel function;

[0097] ;

[0098] ;

[0099] ;

[0100] ;

[0101] Z3: Decision generation function:

[0102] By using decision functions to predict subsequent time-aware features, the corresponding risk warning level can be obtained.

[0103] The steps for generating risk warning information regarding errors in member coordination and allocation in the personnel arrangement extraction module are as follows:

[0104] Q1: Preset the risk control time length, obtain all inspection features within the risk control time length, and extract the relationship mapping information from the inspection features; the relationship mapping information includes several information groups, and each information group includes identity identification code and information code.

[0105] For example, within a preset risk control timeframe, there are 10 inspection features, each of which includes 4 steps, resulting in a total of 40 information groups. Each information group corresponds to the identification code of the staff member performing that step, as well as the information code for that step.

[0106] Q2: Compare the relationship mapping information with the mapping information in the information knowledge graph to generate a risk warning information H for personnel coordination and allocation errors. The mapping information in the information knowledge graph is... .

[0107] , ;

[0108] Where d represents the number of information groups whose mapping relationships belong to records in the information knowledge graph, D represents the total number of knowledge graphs, c represents the index of skill levels, C represents the total number of skill levels, and s represents the total number of skill levels. c The weighting coefficient representing the skill level, u c This represents the number of information groups whose mapping relationships belong to the c-th skill level of the mapping relationships recorded in the information knowledge graph.

[0109] In the technical solution provided in this application, the risk warning information for systematic inspection errors is related to the time-aware characteristics of anomalies, and support vector machines are used to classify the time-aware features. Therefore, it can accurately analyze and plan the scene environment according to the real environment, and accurately increase the risk identification level. For the risk warning information of personnel coordination and allocation errors, a collaborative evaluation is conducted based on the degree of overlap between the current personnel coordination and allocation method and the information knowledge graph, as well as the skill allocation situation, increasing the accuracy of personnel arrangement evaluation.

[0110] The technical solution of this application embodiment has at least the following advantages and beneficial effects:

[0111] Achieve dynamic monitoring and risk visualization throughout the entire process: By constructing standardized inspection features in real time, including operation paths, execution times (or non-execution indicators) and operators, the system can completely and objectively record and present the status of each key link from sample acquisition to result generation. This breaks through the limitations of traditional verification, which can only view static information, and allows potential risks (such as process interruption and link timeout) to be exposed intuitively.

[0112] Accurately identify timeliness risks: The time-aware features extracted by the system (including time statistics such as process time, total duration, and time fluctuation, as well as time structure features such as operation sequence patterns and critical path dependencies) can keenly capture the timeliness decline caused by the stagnation and delay of samples during the process of circulation, and warn of the risk of deviation in test results that may be caused by this.

[0113] Deeply linking personnel factors and operational risks: By constructing a personnel information knowledge graph and integrating it with inspection features to extract personnel perception features, the system can analyze the potential risk tendencies of specific operators (based on their identity, qualifications, experience, and historical behavior patterns) when performing specific inspection items (such as easy omission of specific steps or habitual overtime in operation). Personnel factors are incorporated into the risk assessment system to achieve quantitative analysis of "people" as the key risk source in "people-machine-material-method-environment".

[0114] Multi-dimensional Risk Feature Fusion and Intelligent Assessment: The system integrates time-aware and human-aware features to generate a comprehensive risk feature vector. Based on this, using clustering or classification algorithms, the system can automatically and objectively assess the risk level (e.g., low risk, medium risk, high risk) of each inspection record, identifying samples with questionable reliability or significant flaws in the process, thus providing precise targets for manual review.

[0115] Improving the reliability of test results and the level of process quality control: This system, through automated and intelligent risk monitoring and level assessment, can significantly reduce the risk of test results leaking out of the system due to errors in process execution (omitted steps), loss of sample timeliness, or improper human operation. It identifies high-risk stages before or during result generation, providing a basis for timely intervention and correction. Attached Figure Description

[0116] Figure 1 A schematic diagram of the structure of the sample information entry, transfer, and operational risk monitoring system for verification;

[0117] Figure 2 A schematic diagram of the encoding of identity identifiers.

[0118] Figure 3 This is a schematic diagram for verifying the broken line. Detailed Implementation

[0119] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments. The same reference numerals in the accompanying drawings represent the same components. It should be noted that the described embodiments are only some, not all, of the embodiments of this application. All other embodiments obtained by those skilled in the art based on the described embodiments of this application without creative effort are within the scope of protection of this application.

[0120] Compared to the embodiments shown in the accompanying drawings, feasible embodiments within the scope of this application may have fewer components, other components not shown in the drawings, different components, differently arranged components, or components with different connections, etc. Furthermore, two or more components in the drawings may be implemented in a single component, or a single component shown in the drawings may be implemented as multiple separate components.

[0121] Unless otherwise defined, the technical or scientific terms used herein shall have the ordinary meaning as understood by one of ordinary skill in the art to which this application pertains. The terms “first,” “second,” and similar terms used in this specification and claims do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Similarly, the terms “an” or “a” and similar terms do not necessarily indicate a quantity limitation. Terms such as “upper” and “lower” are used only to indicate relative positional relationships, and these relative positional relationships may change accordingly when the absolute position of the described object changes.

[0122] refer to Figure 1Example 1: The sample information entry, operation, and risk monitoring system includes: a personnel information collection device, an inspection operation definition device, an inspection information acquisition device, a feature extraction device, a personnel information extraction device, and a risk clustering device. The personnel information collection device collects unique identification codes for all staff members in the inspection system. The inspection operation definition device defines information codes for all inspection items in the inspection system. These information codes are generated based on the execution path of the inspection item, which includes the sample acquisition method, sample transportation method, sample processing method, and sample inspection method. The inspection information acquisition device acquires inspection information generated by the inspection system in real time and generates inspection features for each piece of inspection information based on the information code and identity tag code. The feature extraction device acquires inspection features for the same inspection item and extracts the time-aware features of each inspection item from all inspection features. These time-aware features include the time statistical features and time structure features of each inspection item. The personnel information extraction device establishes an information knowledge graph for all staff members. The risk clustering device generates risk warning information for systematic inspection errors based on the time-aware features of each inspection item and generates risk warning information for staff coordination and allocation errors based on the information knowledge graph and all inspection features.

[0123] The sample information entry, processing, and operational risk monitoring system essentially connects to the hospital's HIS system to acquire label information from various stages of the hospital's testing system. It then performs cluster analysis on this label information to promptly identify hidden systemic risks and prevent large-scale testing errors. The key lies in extracting effective information from complex data, vectorizing and reducing its dimensionality, and mapping it to a feature space. Based on the classification results in the feature space, the corresponding risk level is determined.

[0124] The information collected includes the identification information of staff and the inspection information of the inspection items. Staff refers to the personnel involved in the actual sample processing throughout the entire inspection system. These staff members possess different inspection skills and are present at various stages of the inspection process, working together according to the personnel allocation of the department within the inspection system. Therefore, it is necessary to obtain the identification information of the staff to determine whether there are any risk warnings regarding errors in staff coordination and allocation during the inspection process.

[0125] The risk warning information of staff coordination and allocation error refers to the situation in the work arrangement process where staff with multiple skill attributes are assigned to work of a lower skill level than they possess, resulting in the entire verification system being in a state of high error rate.

[0126] Inspection information refers to various information generated in the inspection system. This inspection information needs to be processed according to certain steps. If a step is executed incorrectly for some reason and is not detected, the time for that step will be found to be abnormally increased or decreased.

[0127] Based on this: the collection of staff identity information involves personnel information collection devices and personnel information extraction devices.

[0128] The personnel information collection device includes: an identity information acquisition module, a skill tag acquisition module, and an identity code generation module; the identity code generation module is signal-connected to both the identity information acquisition module and the skill tag acquisition module. The identity information acquisition module is used to acquire the identity IDs of all inspection personnel; the identity ID is a unique identification code for each employee, such as an employee number.

[0129] The skill tag acquisition module obtains the corresponding skill tag for each staff member. The skill tag includes the testing skills they possess and their corresponding skill levels. Testing skills refer to the tasks a staff member can perform in the testing process. For example, some staff members can perform both sample transport and sample processing. Others can perform both sample testing and sample preprocessing. Specific testing skills are the specific work content a staff member can undertake. A staff member can master multiple testing skills. Skill levels are the evaluation indicators for each staff member's testing skills. A staff member typically has one primary job position, corresponding to one or two high-skill-level testing skills. In addition, there are several temporary adjustment positions, corresponding to one or two low-skill-level testing skills.

[0130] The identity coding generation module generates a unique identity identifier code for each employee based on their identity ID and skill tag.

[0131] The identity identifier code is generated as follows:

[0132] refer to Figure 2 S1: Generate a unique and sequentially increasing numbering interval for each inspection skill level; the range of the numbering interval is equal to the number of inspectors corresponding to the skill level of that inspection skill, and generate a numbering sequence by obtaining all the numbering intervals;

[0133] S2: Arrange each staff member according to their identity ID, and then match each element in the number sequence one by one according to the skill level corresponding to the aforementioned inspection skill. The staff member's serial number in the number sequence is the identity identification code.

[0134] An identification code is essentially a one-dimensional sequence of numbers, and the order of these codes describes the worker's job skills. For example, there are four workers skilled in venous blood collection. Therefore, by setting the interval corresponding to the venous blood collection skill to [1, 4], and then arranging the workers according to their skill level in venous blood collection, a numbering sequence can be obtained. The sequence number of each worker within this interval is their identification code.

[0135] To avoid generating multiple different identity codes for the same employee when generating identification codes, each employee only selects the inspection skill with the highest skill level for generation when calculating the number range. This is the method for generating identity codes, which includes the employee's most proficient inspection skills.

[0136] The personnel information extraction device is used to extract information knowledge graphs; information knowledge graph G = (V, E);

[0137] V=P∪S;

[0138] P = {p1, p2, ... p} n}, S = {s1, s2, ... s} m}, where P represents the set of identification codes for all staff members, n represents the total number of staff members, and p n Let s represent the identification code of the nth employee, S represent the total number of all inspection attributes, and s m This represents the m-th test attribute, where m represents the total number of test attributes. Test attributes include test skills and skill levels.

[0139] ;

[0140] R is a set of relations, where (p, s) represents a pair of staff and test attributes, (p... i s j This indicates that the i-th employee has mastered the j-th inspection attribute, (p i s j )∈R, Represents subset notation, E represents all conditions that satisfy the condition. The composition of (p, s).

[0141] The information knowledge graph further supplements the various inspection attributes mastered by each staff member, namely inspection skills and skill levels. Thus, the information knowledge graph can be used to accurately determine the skill mastery of each staff member.

[0142] The above describes the methods for collecting all personnel information. Below is the process for collecting verification information:

[0143] The testing process is divided into four steps: sample acquisition, sample transportation, sample processing, and sample testing.

[0144] Samples can be obtained through methods including: venous blood collection, throat swab collection, urine collection, and tissue biopsy.

[0145] The methods of sample transport include: pneumatic pipeline transport, cold chain transport box, biosafety container, and manual timed delivery for testing;

[0146] Sample processing methods include: centrifugation, anticoagulation, frozen sectioning, and Gram staining;

[0147] The methods for testing samples include: mass spectrometry, microscopic slide reading, rapid detection, and biochemical and immunoassay analyzer detection.

[0148] Step 4 only includes the four items mentioned above, but the specific methods for each step are not limited to the listed items. This can be reasonably divided according to the scope of the hospital's testing work and equipment items.

[0149] The laboratory information acquisition device includes: a laboratory information collection module, a laboratory information extraction module, and an information matrix generation module. The laboratory information collection module is connected to the hospital's HIS system, the laboratory information extraction module is connected to the laboratory information collection module, and the information matrix generation module is connected to the laboratory information extraction module.

[0150] The inspection information collection module is used to collect inspection information uploaded by the inspection system. Inspection information refers to all completed inspection information uploaded to the HIS system. This inspection information fully records the personnel involved in each execution step, the duration of each step, and the corresponding execution method.

[0151] The inspection information extraction module extracts the steps involved in generating the inspection information, the personnel involved in each step, and the duration of each step. The steps involved in generating the inspection information include, in sequence: the sample acquisition method, the sample transportation method, the sample processing method, and the sample inspection method.

[0152] refer to Figure 3 The information matrix generation module replaces the steps of generating the inspection information with the corresponding information codes in sequence to generate the first sequence; it replaces the staff corresponding to the inspection information with the identity code to generate the second sequence; it uses the first sequence as the horizontal axis, the second sequence as the vertical axis, and the duration corresponding to each step as the height value to generate a feature polyline in the three-dimensional coordinate system, and uses the feature polyline as the inspection feature.

[0153] The first sequence is an identification code, which corresponds to the identity of the staff member performing each step.

[0154] The second sequence is the information encoding, which is a one-dimensional sequence. The information encoding for each step is also a one-dimensional sequence. For example, the sequences for venous blood collection, throat swab collection, urine collection, and tissue biopsy are 1 / 2 / 3 / 4, while the sequences for pneumatic pipeline transport, cold chain transport boxes, biosafety containers, and manual timed delivery are 5 / 6 / 7 / 8.

[0155] Each step has an execution time, which is the start and end time of the step. This time is normalized and used as the height coordinate. Thus, each piece of inspection information can be transformed into an inspection feature. The x-axis of this feature indicates the identity of the worker involved in each step, and the y-axis indicates the operation method of each step. The height indicates the duration of each step.

[0156] The above describes the methods for obtaining test features. The reason for extracting test features is mainly to vectorize and digitize the test information, reduce the difficulty of extracting subsequent information, and allow for adjustment of the selection range within the three-dimensional coordinate system.

[0157] The following are the specific methods for extracting test features:

[0158] The feature extraction device includes: a test feature acquisition and classification module, a time statistical feature extraction module, a time structure feature extraction module, and a feature fusion module. The test feature acquisition and classification module is signal-connected to the information matrix generation module. The time statistical feature extraction module and the time structure feature extraction module are respectively connected to the test feature acquisition and classification module, and the feature fusion module is connected to the time statistical feature extraction module and the time structure feature extraction module.

[0159] The inspection feature acquisition and classification module is used to acquire inspection features and classify them according to the inspection items they belong to, resulting in several inspection item groups.

[0160] In practice, the correlation between different testing items for systemic risks is often weak. For example, blood routine tests and urine routine tests have completely different testing procedures. Comparing their testing characteristics together will not reveal systemic risks and will lead to overly redundant models. Therefore, risk management should be implemented for each testing item separately. To this end, testing characteristics are categorized according to the testing item they belong to.

[0161] The time statistical feature extraction module is used to extract time statistical features that characterize the inspection time of each inspection step from the inspection features of the inspection project group.

[0162] Time statistical characteristics F h.f The extraction method is as follows:

[0163] ;

[0164] in, It represents the total average time of all test features in the same test item group within the first time span; This represents the total average time for all test features within the same test item group during the second time span. This represents the total average time of all test features within the same test item group over the third time span. This represents the total average time of all test features in the same test item group within the fourth time span; the first, second, third, and fourth time spans gradually increase, f represents the index of the test item group, and h represents the index of the time statistical feature.

[0165] The first time span is 1 hour, the second is 12 hours, the third is 24 hours, and the fourth is 48 hours. Different time spans introduce different test features within each time period. This approach is designed to ensure that the time statistical features encompass changes in information across different time scales, guaranteeing that the time statistical features accurately describe changes over the total time.

[0166] The time structure feature extraction module is used to extract time structure features from the inspection features of the inspection project group to characterize the changes in inspection time of each inspection step.

[0167] Time structure feature F u.f The extraction method is as follows:

[0168] F u.f The extraction method is as follows:

[0169] ;

[0170] Indicates first-order structural features, Indicates second-order structure characteristics;

[0171] ;

[0172] ;

[0173] Where u represents the time structure feature F u.f The corresponding time interval, g represents the time period, u > 2g; i represents the index of the time period, and q(i) represents the number of test features that show anomalies in the i-th time period;

[0174] The test characteristics are determined based on the activity range of historical normal test characteristics, where f represents the index of the test item group.

[0175] The time structure feature primarily examines whether there is an abnormal increase in the number of abnormal test features within a standard time interval. This method of judging abnormal test features is relatively lenient; it cannot differentiate between individual abnormalities but rather assesses the presence of systemic risk based on the changes in abnormal test features. The specific steps for judging abnormal test features include:

[0176] S1: Obtain the x-coordinate and y-coordinate of each data point in the test feature, and set the corresponding x-coordinate range and y-coordinate range for the test features generated for each type of test item.

[0177] The x-coordinate of each data point is compared with the x-coordinate range in turn. If the x-coordinate of at least one data point is not within the preset x-coordinate range, then the test feature is set as an anomaly test feature.

[0178] The ordinate of each data point is compared with the ordinate range in turn. If the ordinate of at least one data point is not within the preset ordinate range, then the test feature is set as an anomaly test feature.

[0179] S2: Set the corresponding feature space for the inspection features generated for each type of inspection item;

[0180] If at least one data point exists outside the feature space, then this test feature is set as an anomaly test feature.

[0181] For example, a routine blood test should ideally be performed by a venous blood draw, which is typically done by a nurse at the blood collection station. However, if a doctor specifically assigned to sample testing were to collect the blood, an abnormal alarm would be triggered. But this blood collection process is due to an internal personnel change and does not require additional processing.

[0182] For example, a tissue biopsy requires specialized personnel for cold chain transportation. However, in clinical practice, a doctor might urgently bring the sample directly from the operating room for testing. This would also trigger an abnormal alarm.

[0183] The two abnormal alerts mentioned above represent common testing behaviors within hospitals. When the overall allocation of the testing system is reasonable, a small number of abnormal test results are acceptable. However, as workload increases, the number of abnormal test results also increases, inevitably leading to a greater systemic risk. Therefore, in the event of systemic risk, personnel can be assigned to re-examine recent testing work, while simultaneously increasing the number of staff and rescheduling shifts to mitigate the risk.

[0184] The feature space is essentially the upper and lower limits of the operation time set for each step. For example, in a specific test, the blood stratification time cannot exceed 2 hours, so an upper and lower limit can be set for this operation. In practice, a feature space can be generated for each test feature. If a step of a test feature exceeds the set upper limit, it can be identified as an abnormal test feature.

[0185] The feature fusion module is used to fuse time statistical features and time structure features to generate time-aware features representing each test item group. The fusion method mainly involves concatenating features to generate a new feature vector.

[0186] The above describes the time feature extraction method for the inspection project. The practical features include not only the duration of the inspection steps, but also the risk factors of whether each specific operation step is abnormal.

[0187] The risk clustering device includes a systematic risk extraction module and a personnel arrangement extraction module. The systematic risk extraction module is signal-connected to the feature fusion module and is used to acquire event perception features. The personnel arrangement extraction module is signal-connected to both the personnel information extraction device and the information matrix generation module.

[0188] The systemic risk extraction module uses the support vector machine algorithm to perform cluster analysis on time-aware features, generate time-aware vectors, and map these vectors to generate risk warning information. The personnel arrangement extraction module obtains all inspection features within a certain period, extracts the relationship mapping information from the inspection features, compares the relationship mapping information with the mapping information in the information knowledge graph, and generates risk warning information for personnel coordination and allocation errors.

[0189] Furthermore, the risk warning information is generated in the following ways:

[0190] Z1: Prepare datasets in advance. Each dataset includes several samples and their labels. The samples are time-aware features t. i , tag y i The risk warning level is {-1, 0, 1}, which includes 3 level labels, where i represents the index of the sample.

[0191] Z2: Construct a support vector machine model to find the optimal decision function;

[0192] ;

[0193] Where w and b are the parameters to be determined, w represents the weight vector, and b represents the bias term. Represent the feature space;

[0194] Construct the primal optimization problem:

[0195] ;

[0196] represents the regularization parameter, N represents the number of samples, and i represents the sample index;

[0197] The constraints are:

[0198] ;

[0199] e i This represents the prediction error for the i-th sample;

[0200] Lagrange function:

[0201] ;

[0202] Represents the Lagrange multipliers;

[0203] Find the KKT conditions;

[0204] Condition 1: , ;

[0205] Condition 2: , ;

[0206] Condition 3: , ;

[0207] Condition 4: , T represents the matrix bias symbol; Indicates the sign of the partial derivative;

[0208] Introduce kernel functions into the constraint equations;

[0209] The constraint equations introducing KKT conditions are as follows: ;

[0210] The kernel function is: ;

[0211] , j represents the index of the sample;

[0212] Transform the constraint equations into matrix form:

[0213] 1 represents a vector of all 1s, I represents the identity matrix, and K represents the kernel function;

[0214] ;

[0215] ;

[0216] ;

[0217] ;

[0218] Z3: Decision generation function:

[0219] By using decision functions to predict subsequent time-aware features, the corresponding risk warning level can be obtained.

[0220] The above describes the construction process of a support vector machine. The training process is actually about continuously feeding the support vector machine model with a large number of samples to obtain the optimal decision function.

[0221] In this approach, the core value of Least Squares Support Vector Machine (LS-SVM) lies in its equality constraint design and closed-form solution characteristics, which enable it to efficiently capture the dynamic evolution patterns of time series. Although this design sacrifices prediction speed (requiring the computation of the kernel function for the entire sample), its dual advantages in time series modeling accuracy and model interpretability significantly surpass those of traditional SVM. In practice, this interpretability translates into enhanced trust in systemic risk management—by clearly demonstrating feature weights and decision boundaries, it avoids the limitations of traditional neural network models that are unable to play a decisive role in key risk prediction due to the black box nature of their internal weight information.

[0222] The steps for generating risk warning information regarding errors in member coordination and allocation in the personnel arrangement extraction module are as follows:

[0223] Q1: Preset the risk control time length, obtain all inspection features within the risk control time length, and extract the relationship mapping information from the inspection features; the relationship mapping information includes several information groups, and each information group includes identity identification code and information code.

[0224] For example, within a preset risk control timeframe, there are 10 inspection features, each of which includes 4 steps, resulting in a total of 40 information groups. Each information group corresponds to the identification code of the staff member performing that step, as well as the information code for that step.

[0225] Q2: Compare the relationship mapping information with the mapping information in the information knowledge graph to generate a risk warning information H for personnel coordination and allocation errors. The mapping information in the information knowledge graph is... .

[0226] , ;

[0227] Where d represents the number of information groups whose mapping relationships belong to records in the information knowledge graph, D represents the total number of knowledge graphs, c represents the index of skill levels, C represents the total number of skill levels, and s represents the total number of skill levels. c The weighting coefficient representing the skill level, u c This represents the number of information groups whose mapping relationships belong to the c-th skill level of the mapping relationships recorded in the information knowledge graph.

[0228] The risk warning information H for personnel allocation errors in this solution essentially reflects whether personnel allocation is based on relationships recorded in the information knowledge graph. If allocation is based on skill attributes recorded in the information knowledge graph, the corresponding risk warning information H will be low; conversely, it will be high. If the risk warning information H is too high, it indicates a systemic risk in the personnel arrangement of the entire system, requiring timely adjustments.

[0229] The above are merely preferred embodiments of this application and are not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A system for monitoring the input, transfer, and operational risks of test sample information, characterized in that, include: Personnel information collection device, used to collect the unique identification code of all staff in the inspection system; The inspection operation definition device is used to define the information code of all inspection items in the inspection system. The information code is generated based on the execution operation path of the inspection item. The execution operation path includes the sample acquisition method, sample transportation method, sample processing method, and sample inspection method. The inspection information acquisition device acquires inspection information generated by the inspection system in real time, and generates inspection features for each piece of inspection information based on information coding and identity tag coding. A feature extraction device acquires the inspection features of the same inspection items, and extracts the time-aware features of each inspection item from all inspection features. The time-aware features include the time statistical features and time structure features of each inspection item. Personnel information extraction device, which establishes an information knowledge graph for all staff members; The risk clustering device generates risk warning information about systematic inspection errors based on the time-aware characteristics of each inspection item. Based on the information knowledge graph and all inspection features, generate risk warning information for staff coordination and allocation errors; Information knowledge graph G = (V, E); V=P∪S; P = {p1, p2, ... p} n }, S = {s1, s2, ... s} m }, where P represents the set of identification codes for all staff members, n represents the total number of staff members, and p n Let s represent the identification code of the nth employee, S represent the total number of all inspection attributes, and s m This represents the m-th test attribute, where m represents the total number of test attributes. Test attributes include test skills and skill levels. ； R is a set of relations, where (p, s) represents a pair of staff and test attributes, (p i , s j This indicates that the i-th employee has mastered the j-th inspection attribute, (p i , s j )∈R, Represents subset notation, E represents all conditions that satisfy the condition. The composition of (p, s); Risk clustering devices include: The systemic risk extraction module uses the support vector machine algorithm to perform cluster analysis on time-aware features, generate time-aware vectors, and map the time-aware vectors to generate risk warning information. The personnel arrangement extraction module obtains all inspection features within a certain period of time, extracts the relationship mapping information from the inspection features, compares the relationship mapping information with the mapping information in the information knowledge graph, and generates risk warning information for personnel coordination and allocation errors. The risk warning information for systematic inspection errors is generated as follows: Z1: Prepare datasets in advance. Each dataset includes several samples and their labels. The samples are time-aware features t. i , tag y i The risk warning level is {-1, 0, 1}, which includes 3 level labels, where i represents the index of the sample; Z2: Construct a support vector machine model to find the optimal decision function; ； Where w and b are the parameters to be determined, w represents the weight vector, and b represents the bias term. T represents the feature space; T represents the matrix bias sign. Construct the primal optimization problem: ； The constraints are: ; e i This represents the prediction error for the i-th sample. represents the regularization parameter, N represents the number of samples, and i represents the sample index; Lagrange function : ； Represents the Lagrange multipliers; Find the KKT conditions; Condition 1: , ; Condition 2: , ; Condition 3: , ; Condition 4: , T represents the matrix bias symbol; Indicates the partial derivative sign; The constraint equations introducing KKT conditions are as follows: ; The kernel function is: ; , j represents the index of the sample; Transform the constraint equations into matrix form: 1 represents a vector of all 1s, I represents the identity matrix, and K represents the kernel function; ； ； ； ； Z3: Decision generation function: By using decision functions to predict subsequent time-aware features, the corresponding risk warning level can be obtained. The steps for the personnel arrangement extraction module to generate risk warning information regarding staff coordination and allocation errors are as follows: Q1: Preset the risk control time length, obtain all the inspection features within the risk control time length, and extract the relationship mapping information from the inspection features; the relationship mapping information includes several information groups, each information group includes an identity code and an information code; Q2: Compare the relationship mapping information with the mapping information in the information knowledge graph to generate risk warning information H for personnel coordination and allocation errors; The mapping information in the information knowledge graph is ; ， ； Where d represents the number of mapping relationship information groups in all information groups that belong to the mapping relationship recorded in the information knowledge graph, D represents the total number of knowledge graphs, c represents the index of skill level, C represents the total number of skill levels, and s represents the total number of skill levels. c The weighting coefficient representing the skill level, u c This represents the number of information groups whose mapping relationships belong to the c-th skill level of the mapping relationships recorded in the information knowledge graph.

2. The test sample information entry, transfer, and operational risk monitoring system according to claim 1, characterized in that, Personnel information collection devices include: The identity information acquisition module is used to obtain the identity IDs of all inspection personnel; The skill tag acquisition module retrieves the corresponding skill tags for each employee. The skill tags include the testing skills they have mastered and the corresponding skill levels. The identity coding generation module generates a unique identity identifier code for each employee based on their identity ID and skill tag.

3. The test sample information entry, transfer, and operational risk monitoring system according to claim 1, characterized in that, The identity identifier code is generated as follows: For each inspection skill, a unique and sequentially increasing numbering interval is generated for each corresponding skill level. The range of the numbering interval is equal to the number of inspectors corresponding to the skill level of that inspection skill. All numbering intervals are then used to generate a numbering sequence. Each staff member is arranged according to their identity ID, and then each element in the numbering sequence is matched one-to-one with the skill level corresponding to the aforementioned inspection skills. The staff member's serial number in the numbering sequence is the identity identification code.

4. The test sample information entry, transfer, and operational risk monitoring system according to claim 1, characterized in that, Samples can be obtained through methods including: venous blood collection, throat swab collection, urine collection, and tissue biopsy. The methods of sample transport include: pneumatic pipeline transport, cold chain transport box, biosafety container, and manual timed delivery for testing; Sample processing methods include: centrifugation, anticoagulation, frozen sectioning, and Gram staining; The methods for testing samples include: mass spectrometry, microscopic slide reading, rapid detection, and biochemical and immunoassay analyzer detection.

5. The test sample information entry, transfer, and operational risk monitoring system according to claim 1, characterized in that, The inspection information acquisition device includes: The inspection information collection module is used to collect inspection information uploaded by the inspection system; The inspection information extraction module extracts the steps involved in generating the inspection information, the personnel involved in each step, and the duration of each step. The steps involved in generating the inspection information include, in sequence: the sample acquisition method, the sample transportation method, the sample processing method, and the sample inspection method. The information matrix generation module replaces the steps of generating verification information with the corresponding information codes to generate the first sequence; it then replaces the staff members corresponding to the verification information with their identification codes to generate the second sequence. Using the first sequence as the x-axis and the second sequence as the y-axis, and the duration corresponding to each step as the height value, a feature polyline is generated in the three-dimensional coordinate system. The feature polyline is then used as the verification feature.

6. The test sample information entry, transfer, and operational risk monitoring system according to claim 5, characterized in that, The feature extraction device includes: The inspection feature acquisition and classification module is used to acquire inspection features and classify them according to the inspection items they belong to, resulting in several inspection item groups; The time statistical feature extraction module is used to extract time statistical features that characterize the inspection time of each inspection step from the inspection features of the inspection project group. The time structure feature extraction module is used to extract time structure features from the inspection features of the inspection project group to characterize the changes in inspection time of each inspection step. The feature fusion module is used to fuse time statistical features and time structure features to generate time-aware features that characterize each test item group.

7. The test sample information entry, transfer, and operation risk monitoring system according to claim 6, characterized in that, Time statistical characteristics F h.f The extraction method is as follows: in, It represents the total average time of all test features in the same test item group within the first time span; This represents the total average time for all test features within the same test item group during the second time span. This represents the total average time of all test features within the same test item group over the third time span. This represents the total average time of all test features in the same test item group within the fourth time span; the first, second, third, and fourth time spans gradually increase, f represents the index of the test item group, and h represents the index of the time statistical feature.

8. The test sample information entry, transfer, and operational risk monitoring system according to claim 1, characterized in that, Time structure feature F u.f The extraction method is as follows: ； Indicates first-order structural features, Indicates second-order structure characteristics; ； ； Where u represents the time structure feature F u.f The corresponding time interval, g represents the time period, u > 2g; i represents the index of the time period, and q(i) represents the number of test features that show anomalies in the i-th time period; The test characteristics are determined based on the activity range of historical normal test characteristics, where f represents the index of the test item group.

9. The test sample information entry, transfer, and operational risk monitoring system according to claim 8, characterized in that, The method for identifying abnormal characteristics includes the following steps: S1: Obtain the x-coordinate and y-coordinate of each data point in the test feature, and set the corresponding x-coordinate range and y-coordinate range for the test features generated for each type of test item. The x-coordinate of each data point is compared with the x-coordinate range in turn. If the x-coordinate of at least one data point is not within the preset x-coordinate range, then the test feature is set as an anomaly test feature. The ordinate of each data point is compared with the range of ordinates in turn. If the ordinate of at least one data point is not within the preset range of ordinates, then the test feature is set as an anomaly test feature. S2: Set the corresponding feature space for the inspection features generated for each type of inspection item; If at least one data point exists outside the feature space, then this test feature is set as an anomaly test feature.

Images