Multi-level suspected cheating risk calculation method and device and computer readable storage medium

Abstract

The invention discloses a multi-level suspected cheating risk calculation method and device and a computer readable storage medium. The multi-level suspected cheating risk calculation method comprises the steps of determining an examinee suspected cheating rate corresponding to each examinee in a target examination room; determining an examination room suspected cheating rate of the target examination room according to the examinee suspected cheating rate; according to the examination room suspected cheating rates corresponding to the subordinate target examination rooms of the examination room, determining an examination room suspected cheating rate; determining the suspected cheating rate of the examination area according to the examination site suspected cheating rates corresponding to the examination sites in the examination area; and determining a region suspected cheating rate according to the examination area suspected cheating rates corresponding to the examination areas administered by the region management center. Therefore, by constructing a multi-level risk aggregation mechanism of examinees, examination rooms, examination sites, examination areas and area management centers, identified individual abnormal behaviors are fused into group risk indexes level by level, and risk level division, weighted aggregation and dynamic calibration are introduced into each level. Therefore, systematic risk calculation from local detection to global evaluation is realized.

Description

Technical Field

[0001] This application relates to the technical field of educational examination supervision, and more specifically, to a multi-level method, apparatus, and computer-readable storage medium for calculating suspected cheating risks. Background Technology

[0002] With the continuous expansion of the scale of educational examinations, the number of candidates involved in various standardized tests is increasing and the coverage is constantly expanding. In particular, large-scale national examinations such as the National College Entrance Examination (Gaokao) and the Civil Service Examination involve a large number of candidates and cover a wide area. The traditional supervision method that relies on manual inspection is no longer able to meet the supervision requirements of "fairness, impartiality and efficiency".

[0003] While AI-based intelligent patrol systems based on behavior recognition have emerged in existing technologies, they are mostly limited to detecting abnormal behavior in a single examination unit (e.g., examinee or examination room). These systems suffer from a single risk assessment dimension, judging risk based on only a single behavior or frequency, resulting in a high rate of misjudgment. Summary of the Invention

[0004] In view of the above problems, this application proposes a multi-level suspected cheating risk calculation method, device and computer-readable storage medium, which can improve the risk assessment dimensions and thus reduce the false judgment rate.

[0005] In a first aspect, embodiments of this application provide a multi-level suspected cheating risk calculation method. The method includes: performing hierarchical weighting, dynamic correction, and normalization processing on the alarm behavior data of each examinee in the target examination room to determine the suspected cheating rate for each examinee; classifying examinees in the target examination room into multiple examinee risk levels based on the suspected cheating rate, and performing weighted aggregation and dynamic correction of the group risk based on the examinee distribution of each risk level to generate the examination room suspected cheating rate; classifying the target examination rooms under the examination center into multiple examination room risk levels based on the suspected cheating rate of each examination room under the examination center, and performing weighted aggregation and dynamic correction of the concentration perception of the group risk based on the examinee distribution of each risk level to generate the examination room suspected cheating rate; and classifying the target examination rooms under the examination center into multiple examination room risk levels based on the suspected cheating rate of each examination room under the examination center, and performing weighted aggregation and dynamic correction of the concentration perception of the group risk based on the suspected cheating rate of each examination room under the examination center. The examination room risk level is generated by weighting and aggregating group risks and dynamically correcting concentration perception to generate the suspected cheating rate of each examination site. Based on the suspected cheating rates of each examination site within an examination area, the examination sites within that area are divided into multiple examination site risk levels. Weighted aggregation is then performed based on the number of examination sites at each risk level, combined with a regional calibration coefficient, to generate the suspected cheating rate for the entire examination area. Similarly, based on the suspected cheating rates of each examination area within a regional management center, the examination areas within that center are divided into multiple examination area risk levels. Weighted aggregation is then performed based on the number of examination sites at each risk level, combined with a regional management center calibration coefficient, to generate the suspected cheating rate for the entire region.

[0006] Secondly, embodiments of this application also provide a multi-level suspected cheating risk calculation device, which includes: a first module, used to perform hierarchical weighting, dynamic correction, and normalization processing on the alarm behavior data of each candidate in the target examination room to determine the suspected cheating rate of each candidate; a second module, used to divide the candidates in the target examination room into multiple candidate risk levels according to the suspected cheating rate, and to perform weighted aggregation and dynamic correction of the group risk based on the candidate distribution of each candidate risk level to generate the examination room suspected cheating rate; and a third module, used to divide the target examination rooms under the examination center into multiple examination room risk levels according to the suspected cheating rate of each target examination room under the examination center. The system generates a suspected cheating rate for each examination site by weighting and dynamically adjusting the risk level of each examination site based on the risk level of each examination site and the concentration perception. The fourth module is used to divide the examination sites under the jurisdiction of the examination area into multiple risk levels based on the suspected cheating rates of the examination sites under each examination area, and to generate the suspected cheating rate of the examination area by weighting and aggregating the number of examination sites under each risk level and combining it with the regional calibration coefficient. The fifth module is used to divide the examination areas under the jurisdiction of the regional management center into multiple risk levels based on the suspected cheating rates of the examination areas under each examination area, and to generate the suspected cheating rate of the region by weighting and aggregating the number of examination sites under each risk level and combining it with the regional management center calibration coefficient.

[0007] Thirdly, embodiments of this application also provide a computer-readable storage medium storing program code, wherein the above-mentioned multi-level suspected cheating risk calculation method is executed when the program code is run by a processor.

[0008] The technical solution provided in this application includes the following method: performing hierarchical weighting, dynamic correction, and normalization processing on the alarm behavior data of each candidate in the target examination room to determine the suspected cheating rate for each candidate; classifying candidates in the target examination room into multiple candidate risk levels based on the suspected cheating rate, and performing weighted aggregation and dynamic correction of group risk based on the candidate distribution of each risk level to generate the examination room suspected cheating rate; and classifying the target examination rooms under the examination center into multiple examination room risk levels based on the suspected cheating rate of each examination room under the examination center, and then performing weighted aggregation and dynamic correction of group risk based on the candidate distribution of each risk level to generate the examination room suspected cheating rate. This system employs a multi-level risk aggregation mechanism—"examinee—exam room—exam site—exam area—regional management center—regional management center—to generate a systemic risk calculation from local detection to global assessment. Based on the suspected cheating rates of individual test centers within an exam area, test centers are categorized into multiple risk levels. Weighted aggregation based on the number of test centers at each risk level, combined with a regional calibration coefficient, generates the exam area's suspected cheating rate. Similarly, based on the suspected cheating rates of test areas within a regional management center, exam areas are categorized into multiple risk levels. Weighted aggregation based on the number of test centers at each risk level, combined with a regional management center calibration coefficient, generates the regional suspected cheating rate. Thus, by constructing a multi-level risk aggregation mechanism—"examinee—exam room—exam area—regional management center"—identified individual abnormal behaviors are progressively integrated into group risk indicators. Risk level classification, weighted aggregation, and dynamic calibration are introduced at each level. Attached Figure Description

[0009] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments and drawings obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0010] Figure 1 The illustration shows a flowchart of a multi-level suspected cheating risk calculation method provided in an embodiment of this application.

[0011] Figure 2 A schematic diagram of a multi-level suspected cheating risk calculation device provided in an embodiment of this application is shown.

[0012] Figure 3 This illustration shows a schematic diagram of the structure of a computer-readable storage medium provided in an embodiment of this application. Detailed Implementation

[0013] This application provides a multi-level suspected cheating risk calculation method, device, and computer-readable storage medium. The method includes: performing hierarchical weighting, dynamic correction, and normalization processing on alarm behavior data of each candidate in the target examination room to determine the suspected cheating rate for each candidate; classifying candidates in the target examination room into multiple candidate risk levels based on the suspected cheating rate, and performing weighted aggregation and dynamic correction of group risk based on the candidate distribution of each risk level to generate the examination room suspected cheating rate; and classifying the target examination rooms under the examination center into multiple examination room risk levels based on the suspected cheating rate of each examination room under the examination center. Based on the examination room composition according to the risk level of each examination room, a weighted aggregation of group risks and dynamic correction of concentration perception are performed to generate the suspected cheating rate of each examination site. According to the suspected cheating rate of each examination site under the jurisdiction of the examination area, the examination sites under the jurisdiction of the examination area are divided into multiple examination site risk levels, and a weighted aggregation is performed based on the number of examination sites under each risk level, combined with the regional calibration coefficient, to generate the suspected cheating rate of the examination area. According to the suspected cheating rate of each examination area under the jurisdiction of the regional management center, the examination areas under the jurisdiction of the regional management center are divided into multiple examination area risk levels, and a weighted aggregation is performed based on the number of examination sites under each risk level, combined with the regional management center calibration coefficient, to generate the regional suspected cheating rate.

[0014] Therefore, by constructing a multi-level risk aggregation mechanism of "examinee - examination room - examination site - examination area - regional management center", the identified individual abnormal behaviors are gradually integrated into group risk indicators. At each level, risk level classification, weighted aggregation and dynamic calibration are introduced to achieve systematic risk calculation from local detection to global assessment.

[0015] This invention provides a multi-level suspected cheating risk calculation method. The executing entity of this method includes, but is not limited to, at least one of the following electronic devices that can be configured to execute the method provided in this application: a server, a terminal, etc. In other words, this multi-level suspected cheating risk calculation method can be executed by software or hardware installed on a terminal device or a server device. The software can be a blockchain platform. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster. The server can be an independent server or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDN), and big data and artificial intelligence platforms.

[0016] Please see Figure 1 , Figure 1This illustration shows a flowchart of a multi-level suspected cheating risk calculation method provided in an embodiment of this application. Figure 1 As shown, the method may include steps 110 to 150.

[0017] In step 110, the alarm behavior data of each candidate in the target examination room are subjected to hierarchical weighting, dynamic correction and normalization to determine the suspected cheating rate of each candidate.

[0018] The target examination room can be a specific examination room currently undergoing a cheating risk assessment, i.e., the examination room being processed by the method of this invention. In a large-scale examination, an examination center typically contains multiple examination rooms (e.g., different classrooms in a teaching building). The system performs candidate risk analysis and group risk aggregation on each of these examination rooms sequentially or in parallel. Any examination room processed in this case is the "target examination room".

[0019] Alarm behavior data can be a set of candidate behavior events that are automatically detected and recorded by the intelligent proctoring system during the examination process based on audio and video analysis, behavior recognition algorithms or sensor monitoring, and judged as abnormal or suspicious.

[0020] In some implementations, alarm behavior data includes, but is not limited to, alarm type, behavior trigger frequency, and behavior combination. Alarm type represents the specific category of abnormal behavior. Examples include "frequent head turning," "hand obscuring the answer area," "using electronic devices," "prolonged head-down posture," and "communicating with a neighbor," with each type corresponding to a preset behavior recognition rule or AI model output label. Trigger frequency refers to the number of times the same alarm type is triggered in a single exam, reflecting the repetitiveness and persistence of the behavior. Behavior combination refers to a combination pattern of multiple different alarm types that overlap, occur consecutively, or have a specific sequential relationship in time (e.g., "head-down posture + hand under the table + remaining still for more than 30 seconds"), used to identify complex suspicious behaviors with collaborative intent.

[0021] Alarm behavior data can come from the behavior logs of cameras, audio acquisition devices, desktop sensing devices, or online examination clients deployed in the examination room. After being preprocessed by the system, the data is stored in a structured manner and used as the raw input for calculating the suspected cheating rate of candidates.

[0022] The suspected cheating rate is a quantitative indicator ranging from 0% to 100%, used to characterize the likelihood that a particular candidate has cheated during the current exam. This indicator is not a legal determination of cheating, but rather a relative risk score derived from multi-dimensional risk modeling based on alert behavior data. The suspected cheating rate allows for horizontal comparison of the risk levels of different candidates within a target examination room.

[0023] Specifically, in some implementations, the step of "determining the suspected cheating rate of each candidate by performing hierarchical weighting, dynamic correction, and normalization processing on the alarm behavior data of each candidate in the target examination room" may include the following steps: (1) Obtain alarm behavior data for each candidate in the target examination room. The alarm behavior data shall include at least the alarm type, frequency of behavior triggering, and combination of behaviors. (2) In the preset alarm behavior basic information table, query the basic score and behavior weight corresponding to the alarm type; (3) Determine the frequency coefficient based on the frequency of behavior triggering; (4) In the preset behavior combination basic information table, query the correlation coefficient corresponding to the behavior combination; (5) Based on the basic score, behavior weight, frequency coefficient, correlation coefficient, preset full score threshold for candidates and candidate multidimensional risk scoring formula, determine the suspected cheating rate of each candidate.

[0024] Alarm behavior data can be generated by local edge computing devices in the examination room, or it can be centrally processed and distributed by the cloud-based proctoring platform. Finally, it is organized in a structured form (such as JSON or database records) by candidate ID, serving as the raw input for calculating the suspected cheating rate of candidates.

[0025] To quantify the risks of different alarm behaviors, the system pre-builds a basic information table for preset alarm behaviors. This table is a structured data table used to store risk assessment parameters for various alarm behaviors in standardized examination scenarios, serving as the basis for calculating the suspected cheating rate of examinees.

[0026] The basic information table for preset alarm behaviors should include at least: alarm type, examinee risk level, base score, and behavior weight. The base score represents the inherent risk level of that alarm type in the default exam scenario, and is typically set by invigilators based on historical cheating cases and behavioral hazard assessments. The behavior weight reflects the relative importance of that alarm type within the current exam type (e.g., closed-book written exam, open-book exam, computer-based exam, oral test, etc.) or exam level (e.g., national exam, school-level test).

[0027] During the calculation process, the system retrieves matching items from the preset alarm behavior basic information table based on the alarm types extracted from the proctoring data, and obtains the corresponding basic score and behavior weight. In a specific implementation, the preset alarm behavior basic information table can be Table 1. It is worth noting that the preset alarm behavior basic information table is not fixed, but supports dynamic updates and scenario adaptation.

[0028] Table 1 The frequency coefficient is an adjustment factor greater than or equal to 1.0, determined based on the total trigger frequency of various alarm behaviors or the repetition frequency of a single type of alarm in a single exam session. By introducing the frequency coefficient, the system can effectively distinguish between "occasional anomalies" and "persistent suspicious behavior," improving the accuracy and robustness of the assessment of the suspected cheating rate.

[0029] The frequency coefficient is set based on the following invigilation experience: occasional abnormal behavior (such as turning one's head once) is mostly caused by distraction or physiological habits and has a low risk; while high-frequency repeated alarms of the same or different types are more likely to point to premeditated cheating behavior and have a significantly higher risk.

[0030] Specifically, frequency coefficients can be generated through a preset nonlinear mapping function. For example, piecewise linear functions, logarithmic functions, or saturated exponential functions can be used to avoid scoring distortion due to extremely high frequencies, while retaining sensitivity to medium- and high-risk behaviors.

[0031] In one specific embodiment, when the total alarm frequency is ≤2 times, the frequency coefficient is 1.0; when 3 ≤ total alarm frequency is ≤5 times, the frequency coefficient is 1.2; when 6 ≤ total alarm frequency is ≤8 times, the frequency coefficient is 1.5; and when the total alarm frequency is >8 times, the frequency coefficient is 1.8.

[0032] To identify complex suspicious behaviors with collaborative intent, the system further introduces a behavior combination analysis mechanism and enhances and corrects the risk score through correlation coefficients. A behavior combination can refer to a higher-order behavior pattern formed by two or more different alarm types that meet preset correlation conditions in the time dimension (such as time overlap, consecutive occurrence, or co-occurrence within a set time window).

[0033] For example, "looking down + covering the answer area with hands + remaining still for more than 20 seconds" may indicate hiding and checking cheat sheets; "turning the head multiple times + focusing the gaze on the neighboring seat + answering questions in sync" may indicate plagiarism.

[0034] To quantify the risk-enhancing effect of such combinations, the system pre-constructs a basic information table of preset behavioral combinations. The correlation coefficient is set by invigilators based on historical cheating cases, behavioral psychology analysis, or machine learning model training results.

[0035] For example, the correlation coefficient for a common combination (such as "looking down + turning a page") can be set to 1.1, while that for a high-risk combination (such as "using a mobile phone + obstructing the camera"), it can be set to 2.0. This table supports dynamic updates, allowing exam administrators to promptly expand or adjust the combination rules and corresponding coefficients based on the emergence of new cheating methods, thereby improving the system's adaptability and foresight.

[0036] In one specific implementation, the basic information table of preset behavior combinations can be Table 2. It is worth noting that the basic information table of preset behavior combinations is not fixed, but supports dynamic updates and scenario adaptation.

[0037] Table 2 The preset maximum score threshold for a candidate can be the highest risk score that a candidate could theoretically trigger under the current exam configuration. This preset maximum score threshold is not a fixed constant, but rather dynamically calculated or pre-configured based on the exam type, the set of allowed alarm types, the maximum possible frequency of each alarm type, and the strongest combination of behaviors.

[0038] In one specific implementation, a maximum score threshold for candidates is preset to select the combination of the highest risk behaviors that candidates may trigger. This threshold is set by multiplying the base score by the highest behavior weight by the highest frequency coefficient by the highest correlation coefficient by the number of behavior types, ensuring coverage of extreme risk scenarios.

[0039] In one specific implementation, the preset full score threshold for candidates is calculated as follows: (highest single-category basic score × highest behavior weight) × highest frequency coefficient × highest correlation coefficient × number of behavior types.

[0040] In one specific implementation, the expression for the candidate's multidimensional risk scoring formula can be: Wherein, represents the suspected cheating rate of candidates, represents the basic score, represents the behavior weight, represents the frequency coefficient, represents the correlation coefficient, and represents the preset full score threshold for candidates.

[0041] Thus, through the aforementioned steps for calculating the suspected cheating rate of examinees, a refined and standardized quantitative assessment of the cheating risk of each examinee in the target examination room can be conducted based on multi-dimensional alarm behavior data, outputting an individual risk score ranging from 0% to 100%. This score not only integrates key factors such as the nature of the behavior, frequency of occurrence, and combined correlations, but also achieves horizontal comparability across examinees and examination rooms through normalization processing, effectively supporting the accurate identification and verification of high-risk individuals.

[0042] However, in large-scale educational examination monitoring scenarios, relying solely on individual risk assessments is insufficient to fully reflect the overall security situation of the examination venue. For example, the concentrated presence of multiple low- to medium-risk candidates may indicate systemic cheating, while a single high-risk candidate may be diluted in a large examination venue. Therefore, it is necessary to further construct an examination venue-level group risk aggregation mechanism based on individual assessments to capture risk distribution characteristics and clustering effects.

[0043] In step 120, based on the suspected cheating rate of candidates, candidates in the target examination room are divided into multiple candidate risk levels, and the group risk is dynamically corrected by weighted aggregation and concentration perception based on the candidate distribution of each candidate risk level to generate the suspected cheating rate of the examination room.

[0044] In some implementations, multiple candidate risk levels include high-risk candidates, medium-risk candidates, and low-risk candidates.

[0045] The suspected cheating rate in an examination room can be a standardized scoring indicator used to quantify the degree of abnormality in the overall examination of a single examination room. Its value ranges from 0% to 100%, and it is calculated by integrating the individual risk characteristics of all candidates in the examination room.

[0046] On the one hand, the suspected cheating rate in the examination room transforms discrete abnormal behaviors of candidates into a structured risk representation at the examination room level, avoiding misjudgments caused by relying solely on a few extreme cases. On the other hand, the standardized output (in percentage form) of the suspected cheating rate in the examination room makes it comparable across examination rooms with different capacities and question types, facilitating the setting of unified warning thresholds (e.g., ≥50% is considered a high-risk examination room).

[0047] Based on the distribution of candidates at each level, the overall risk of the examination room is weighted and aggregated, and further dynamically adjusted by incorporating the concentration of high-risk candidates, thereby generating a normalized suspected cheating rate for the examination room. This indicator not only reflects the absolute number of high-risk candidates but also captures their clustering characteristics, effectively identifying examination rooms with systemic anomalies or suspicious group behavior.

[0048] Specifically, in some implementations, the step "based on the suspected cheating rate of candidates, dividing candidates in the target examination room into multiple candidate risk levels, and dynamically correcting the group risk based on the distribution of candidates at each risk level and concentration perception to generate the suspected cheating rate of the examination room" may include the following steps: (1) Based on the suspected cheating rate of candidates and the preset classification threshold, candidates in the target examination room are classified into high-risk candidates, medium-risk candidates and low-risk candidates; (2) Determine the target number of examinees per examination room based on the sum of the number of high-risk examinees and medium-risk examinees; (3) Query the first basic contribution value and first weighting coefficient corresponding to high-risk candidates, medium-risk candidates and low-risk candidates respectively in the preset level assessment parameter table; (4) Based on the ratio between the target number of examinees in the examination room and the number of examinees in the target examination room, determine the risk distribution coefficient of the target examination room in the preset risk distribution coefficient configuration table; (5) Determine the suspected cheating rate of the examination room based on the number of candidates with each risk level, the first basic contribution value, the first weighting coefficient, the risk distribution coefficient, the preset full score threshold of the examination room and the examination room multidimensional risk scoring formula.

[0049] The preset threshold can be a pre-configured set of numerical boundaries used to define the range of risk levels for different candidates. It is worth noting that the preset threshold is not fixed, but can be dynamically adjusted based on the type of exam (e.g., provincial exam vs. school test), security level requirements, or historical cheating data statistics.

[0050] In one specific implementation, the preset classification thresholds may include 60% and 30%. When a candidate's suspected cheating rate is ≥60%, the candidate is classified as a high-risk candidate; when a candidate's suspected cheating rate is <60%, the candidate is classified as a medium-risk candidate; and when a candidate's suspected cheating rate is <30%, the candidate is classified as a low-risk candidate.

[0051] In order to model the differentiated contributions of candidates with different risk levels to the overall risk of the examination room, the system is pre-configured with a preset level assessment parameter table. The preset level assessment parameter table is a structured data table used to store the first basic contribution value and the first weighting coefficient for high-risk, medium-risk and low-risk candidates.

[0052] In one specific implementation, the contents of the preset level evaluation parameter table can be shown in Table 3.

[0053] Table 3 The first basic contribution value in the preset level assessment parameter table can represent the baseline impact of a candidate's risk level on the overall risk of the examination room, reflecting the degree of their inherent suspiciousness. For example, high-risk candidates are given a higher first basic contribution value (e.g., 10 points) because they are more likely to cheat individually, while low-risk candidates are only considered as background noise and have a lower contribution value (e.g., 1 point).

[0054] The first weighting coefficient in the preset grade assessment parameter table is used to adjust the risk margin effect when the number of candidates of the same grade increases.

[0055] To reflect the varying impacts of high-risk candidate concentration levels on overall risk assessment across examination rooms of different sizes, the system introduces a risk distribution coefficient as a dynamic correction factor. The specific value of this coefficient is determined through a preset risk distribution coefficient configuration table. This table is a structured data table used to store the risk distribution coefficients based on the examination room size (i.e., the ratio of the total number of examinees to the actual number of candidates).

[0056] In one specific implementation, the contents of the preset risk distribution coefficient configuration table can be shown in Table 4.

[0057] Table 4 The risk distribution coefficients in the preset risk distribution coefficient configuration table can be used to adjust the overall risk score when high-risk and medium-risk candidates are concentrated in the target examination room. The higher the value of the risk distribution coefficient, the higher the overall risk of the target examination room.

[0058] The preset full score threshold for the exam room can be the maximum group risk score that the target exam room can theoretically achieve under the current exam configuration. It is used to normalize the actual calculated total risk value, thereby outputting a standardized exam room suspected cheating rate in the range of 0% to 100%.

[0059] In one specific implementation, the preset full score threshold for the examination room is calculated as: total number of candidates in the target examination room × basic contribution value corresponding to high-risk candidates × first weighting coefficient corresponding to high-risk candidates.

[0060] In one specific implementation, the formula for the examination room multidimensional risk scoring can be expressed as: Wherein, is the suspected cheating rate of the target examination room, is the number of candidates at each risk level, is the first basic contribution value, is the first weighting coefficient, is the risk distribution coefficient, and is the preset full score threshold for the examination room.

[0061] The multidimensional risk scoring formula for examination rooms enables a refined, structured, and standardized assessment of the overall risk of the target examination room. By introducing the first basic contribution value and the first weighting coefficient, the influence of high-risk and medium-risk candidates is reasonably amplified, while the influence of low-risk candidates is moderately suppressed, thus more accurately reflecting the composition of group risk. Furthermore, by introducing a risk distribution coefficient to perform a secondary correction on the aggregation results, the essential difference between "a few high-risk groups clustering together" and "uniformly dispersed risk" can be identified.

[0062] The above-mentioned technical solution for calculating the suspected cheating rate in examination rooms integrates the risk level weights of candidates, the effect of the number of candidates, and the risk concentration correction, and generates a standardized suspected cheating rate in examination rooms through normalization. This effectively solves the problem of assessment bias caused by the difference in the size of examination rooms and achieves accurate quantification of the risk of group cheating.

[0063] However, in actual exam management, a test center usually contains multiple target exam rooms, and its overall security level depends not only on individual high-risk exam rooms, but also on the risk distribution pattern among the exam rooms. For example, if multiple exam rooms within a test center simultaneously show medium to high risk, it may indicate organized cheating or invigilation loopholes, which is far more serious than isolated exam room anomalies.

[0064] In step 130, based on the suspected cheating rates of the target examination rooms under the test center, the target examination rooms under the test center are divided into multiple examination room risk levels. Based on the examination room composition of each examination room risk level, the group risk is weighted and aggregated and dynamically corrected by concentration perception to generate the suspected cheating rate of the test center.

[0065] In some implementations, multiple examination room risk levels include high-risk examination rooms, medium-risk examination rooms, and low-risk examination rooms.

[0066] The suspected cheating rate at a test center can be a standardized indicator used to quantify the overall examination security risk level of a test center (e.g., a middle school or examination center), with a value ranging from 0% to 100%. Essentially, the suspected cheating rate at a test center is a comprehensive representation of the internal risk structure, scale characteristics, and concentration degree of the test center.

[0067] By introducing the standardized indicator of suspected cheating rate at test centers, a higher-level perception of risk has been achieved, moving from individual anomalies to group risks: by structurally aggregating the risks of subordinate test centers, scattered abnormal signals are integrated into systemic risk judgments, significantly improving the ability to identify organized and gang-related cheating; at the same time, to address the inherent assessment bias in the proportion of high-risk test centers of different sizes, a scale coefficient is introduced and combined with normalization design to ensure that the scoring results are horizontally comparable.

[0068] The suspected cheating rate at test centers is output as a percentage from 0% to 100%, which makes it easy to set a unified warning threshold (e.g., ≥60% is considered a high-risk test center), thereby automatically triggering precise supervision measures such as video review and additional invigilators, which greatly improves the intelligence level and response efficiency of test administration.

[0069] Based on the composition of examination rooms according to their risk levels, the overall risk of each examination site is weighted and aggregated, and further dynamically adjusted by considering the concentration of high-risk examination rooms, ultimately generating a standardized suspected cheating rate for each examination site. This approach not only considers the absolute number of high-risk examination rooms but also identifies systemic anomalies implied by the clustering of multiple medium-to-high-risk examination rooms through a "concentration perception" mechanism, thereby effectively improving the ability to identify organized cheating behavior.

[0070] Specifically, in some implementations, the step of "dividing the target examination rooms under the test center into multiple examination room risk levels according to the suspected cheating rate of each examination room under the test center, and dynamically correcting the group risk based on the examination room composition of each examination room risk level and concentration perception to generate the suspected cheating rate of the test center" may include the following steps: (1) Based on the suspected cheating rate and preset grading threshold of the test center's target test rooms, the test center's target test rooms are divided into high-risk test rooms, medium-risk test rooms and low-risk test rooms; (2) Query the second basic contribution value and second weighting coefficient corresponding to high-risk, medium-risk and low-risk examination rooms in the preset examination site risk aggregation parameter table; (3) Determine the scale coefficient in the preset scale coefficient configuration table based on the number of target examination rooms under the examination center; (4) Based on the ratio between the number of high-risk examination rooms and the number of target examination rooms under the examination site, determine the group risk coefficient of the examination site in the pre-set group risk coefficient configuration table; (5) Based on the number of examination rooms for each candidate’s risk level, the second basic contribution value, the second weighting coefficient, the scale coefficient, the group risk coefficient, the preset full score threshold for examination sites, and the examination site multidimensional risk scoring formula, generate the examination site suspected cheating rate.

[0071] Preset grading thresholds can be a pre-configured set of numerical boundaries used to map continuous suspected cheating rates in exam rooms to discrete exam room risk levels (i.e., high-risk, medium-risk, and low-risk exam rooms). Preset grading thresholds typically include a lower threshold for high-risk exam rooms and a lower threshold for medium-risk exam rooms, forming a three-level division interval. It is important to note that preset grading thresholds are not fixed but can be dynamically adjusted based on exam type, security policy requirements, or historical risk data statistics.

[0072] In one specific implementation, when the suspected cheating rate of a target examination room is ≥60%, the target examination room is classified as a high-risk examination room; when the suspected cheating rate of a target examination room is ≤30% and <60%, the target examination room is classified as a medium-risk examination room; when the suspected cheating rate of a target examination room is <30%, the target examination room is classified as a low-risk examination room.

[0073] To model the differentiated contributions of examination rooms with different risk levels to the overall risk of the examination site, the system pre-configures a preset examination site risk aggregation parameter table. This table is a structured data table used to store the second basic contribution value and second weighting coefficient for high-risk, medium-risk, and low-risk examination rooms.

[0074] In one specific implementation, the specific contents of the preset test point risk aggregation parameter table can be shown in Table 5.

[0075] Table 5 The second basic contribution value in the pre-set test site risk aggregation parameter table can represent the baseline impact of the test room of that risk level on the overall risk of the test site, reflecting its inherent level of suspicion. For example, a high-risk test room is assigned a higher basic contribution value (e.g., 20 points) because there are more abnormal behaviors inside, while a low-risk test room is only considered as background noise and has a lower contribution value (e.g., 3 points).

[0076] The second weighting coefficient in the pre-set test site risk aggregation parameter table can be used to adjust the risk marginal effect when the number of test rooms of the same level increases. For example, when there are many high-risk test rooms, the possibility of them colluding to cheat increases, so a non-linear amplification is performed by a weighting coefficient greater than 1.0 (e.g., 1.8); while even if there are many low-risk test rooms, the possibility of them colluding to cheat is still low, so a weighting coefficient less than 1.0 (e.g., 0.6) is used to suppress it.

[0077] To reflect the differences in risk assessment among test centers of varying sizes, this invention introduces a size coefficient as a dynamic adjustment factor, and determines its specific value through a preset size coefficient configuration table. The preset size coefficient configuration table is a structured data table that stores the mapping relationship between the range of the number of test rooms under a test center and the corresponding size coefficient.

[0078] In one specific implementation, the contents of the preset scale coefficient configuration table can be shown in Table 6.

[0079] Table 6 The scale coefficients in the preset scale coefficient configuration table can be used to adjust the overall risk score of the test center to reflect the dilution or amplification effect of the number of test rooms on the concentration of risk.

[0080] To quantify the concentration of high-risk examination rooms within examination centers and the implied likelihood of group cheating, this invention introduces a group risk coefficient as a key correction factor, and determines its value through a preset group risk coefficient configuration table. The preset group risk coefficient configuration table is a structured data table used to store the mapping relationship between the proportion range of high-risk examination rooms and the corresponding group risk coefficients.

[0081] In one specific implementation, the contents of the preset group risk coefficient configuration table can be shown in Table 7.

[0082] Table 7 The group risk coefficient in the preset group risk coefficient configuration table can be used to dynamically amplify the overall risk score of the test center to reflect the systemic anomalies caused by the clustering of high-risk test rooms. When more than half of the test rooms in a test center are high-risk, it is highly likely to involve invigilator negligence, test question leakage, or group cheating, and the risk weight should be significantly increased (e.g., coefficient 2.8). If only a few test rooms are abnormal (e.g., the proportion is between 1% and 10%), it is more likely to be an isolated event, and the baseline coefficient (e.g., coefficient 1.3) can be used.

[0083] In one specific implementation, the preset full score threshold for test centers = the maximum number of test rooms at the test center × the second basic contribution value corresponding to high risk × the second weighting coefficient corresponding to high risk.

[0084] In one specific implementation, the expression for the multidimensional risk scoring formula for test points can be: Among them, is the suspected cheating rate of the test center, is the number of test rooms at each risk level, is the second basic contribution value, is the second weighting coefficient, is the scale coefficient, is the group risk coefficient, and is the preset full score threshold for the test center.

[0085] By setting up a multi-dimensional risk scoring formula for test centers, and introducing a second basic contribution value and a second weighting coefficient corresponding to the risk level of each candidate, the impact of high-risk test centers is reasonably amplified, while the impact of low-risk test centers is appropriately suppressed, thus more accurately reflecting the risk composition of test centers. Furthermore, a scale coefficient is used to correct the risk perception bias caused by the different sizes of test centers—a small number of high-risk test centers in small test centers are more significant as a warning, and are therefore amplified; large test centers avoid the "quantity trap" and prevent misjudgment. A group risk coefficient is also introduced to capture the clustering effect of high-risk test centers—when multiple high-risk test centers appear together, the possibility of systematic cheating increases dramatically, and this coefficient automatically increases the score, effectively identifying organized anomalies.

[0086] In other words, the multi-dimensional risk scoring formula for test sites not only integrates the three dimensions of test site risk structure, scale effect and risk concentration, but also supports a multi-level risk aggregation system from test site to test center to test area through a normalization mechanism, providing a scientific and reliable technical foundation for intelligent invigilation and resource allocation for large-scale educational examinations.

[0087] In step 140, based on the suspected cheating rates of the test sites under the jurisdiction of the test area, the test sites under the jurisdiction of the test area are divided into multiple test site risk levels, and the test sites are weighted and aggregated based on the number of test sites at each risk level, and combined with the regional calibration coefficient to generate the suspected cheating rate of the test area.

[0088] In some implementations, the risk levels of multiple test sites may include high-risk test sites, medium-risk test sites, and low-risk test sites.

[0089] The suspected cheating rate in an examination area can be a standardized assessment indicator used to quantify the overall examination security risk level of an examination area (e.g., a city or administrative region), with a value range of 0% to 100%.

[0090] The suspected cheating rate for an examination area is not simply an average of the risk scores of its subordinate examination sites, but is dynamically generated based on a multi-dimensional fusion model. On the one hand, the suspected cheating rate for an examination area breaks through the traditional crude assessment model based on the number of examination sites or average scores, and can accurately identify the regional systemic cheating risk implied by the clustering of multiple high-risk examination sites. On the other hand, through a regional calibration mechanism, it effectively balances the differences in scale, examination importance, and historical performance among different examination areas, avoiding assessment biases such as "large examination areas naturally have high scores" or "small examination areas being dominated by single-point anomalies."

[0091] Specifically, in some implementations, the step "based on the suspected cheating rates of the test centers under the jurisdiction of the test area, dividing the test centers under the jurisdiction of the test area into multiple test center risk levels, and weighting and aggregating the test centers based on the number of test centers at each risk level, and combining this with the regional calibration coefficient to generate the suspected cheating rate of the test area" may include the following steps: (1) Based on the suspected cheating rate and preset classification threshold of the test sites under the jurisdiction of the test area, the test sites under the jurisdiction of the test area are divided into high-risk test sites, medium-risk test sites and low-risk test sites respectively; (2) Query the third basic contribution value and third weighting coefficient corresponding to high-risk test sites, medium-risk test sites and low-risk test sites in the preset test area risk aggregation parameter table; (3) Based on the third basic contribution value, the third weighting coefficient, the regional calibration coefficient, the preset full score threshold of the examination area and the examination area multidimensional risk scoring formula, the examination area suspected cheating rate is generated.

[0092] Preset classification thresholds can be a pre-configured set of numerical boundaries used to map the continuous suspected cheating rate of test sites to discrete test site risk levels (high-risk test sites, medium-risk test sites, and low-risk test sites). Preset classification thresholds typically include a high-risk lower limit threshold and a medium-risk lower limit threshold, thus dividing the test sites into three risk ranges.

[0093] In one specific implementation, if the suspected cheating rate of a test center is ≥60%, the test center is classified as a high-risk test center; if the suspected cheating rate of a test center is ≤30%, the test center is classified as a medium-risk test center; and if the suspected cheating rate of a test center is <30%, the test center is classified as a low-risk test center.

[0094] It is worth noting that the preset classification threshold is not fixed, but can be dynamically adjusted according to the type of exam (e.g., college entrance examination, academic level test), regional regulatory strategies, or historical risk distribution statistics.

[0095] To model the differentiated contributions of test sites with different risk levels to the overall risk of the test area, the system pre-configures a preset test area risk aggregation parameter table. This table is a structured data table used to store the third basic contribution value and third weighting coefficient for high-risk, medium-risk, and low-risk test sites.

[0096] In one specific implementation, the specific contents of the preset examination area risk aggregation parameter table can be shown in Table 8.

[0097] Table 8 The third basic contribution value in the pre-set test area risk aggregation parameter table can represent the baseline impact of the test point of that risk level on the overall risk of the test area, reflecting its inherent level of suspicion. For example, high-risk test points are given a higher basic contribution value (e.g., 30 points) because there are more abnormal behaviors inside, while low-risk test points are only considered as background noise and have a lower contribution value (e.g., 5 points).

[0098] The third weighting coefficient in the pre-set test area risk aggregation parameter table is used to adjust the risk marginal effect when the number of test points of the same level increases. For example, when there are many high-risk test points, the possibility of them colluding to cheat increases, so a weighting coefficient greater than 1.0 (such as 2.0) is used for non-linear amplification; while even if there are many low-risk test points, the possibility of them colluding to cheat is still low, so a weighting coefficient less than 1.0 (such as 0.7) is used to suppress it.

[0099] The regional calibration coefficient can reflect the difficulty of supervising an examination area, the importance of the examination, and the historical compliance level. For example, a large examination area that undertakes national-level examinations and has a history of frequent violations will have a significantly higher regional calibration coefficient than a small, low-risk examination area with a good compliance record, thus being reasonably amplified in risk scoring. In a specific implementation, the regional calibration coefficient can be determined by a size coefficient, a type coefficient, and a historical compliance coefficient.

[0100] Furthermore, in some embodiments, the method may further include the following steps: (1) Query the scale coefficient corresponding to the number of examination rooms included in the examination area in the preset examination area scale coefficient configuration table; (2) Query the type coefficient corresponding to the examination type undertaken by the examination area in the preset examination area type coefficient configuration table; (3) Query the historical compliance coefficient corresponding to the average cheating rate of the examination area within the preset time period in the preset examination area historical compliance coefficient configuration table; (4) Quantitatively integrate the scale coefficient, type coefficient and historical compliance coefficient to determine the regional calibration coefficient.

[0101] The preset examination area size coefficient configuration table is a structured data table used to store the mapping relationship between the total number of examination rooms under the jurisdiction of an examination area and the corresponding size coefficient. The larger the examination area, the more difficult the supervision and coordination, but the risk is also more easily diluted; conversely, a few anomalies in a small examination area may represent a systemic problem.

[0102] In one specific implementation, the contents of the preset examination area size coefficient configuration table can be shown in Table 9.

[0103] Table 9 The scale coefficient in the preset examination area scale coefficient configuration table can represent the quantification of regulatory difficulty based on the number of examination rooms in the examination area. By eliminating risk assessment bias caused by differences in examination area size through the scale coefficient, fair and comparable risk quantification can be achieved between units of different sizes.

[0104] For example, even if a small test area (such as one with only 30 test rooms) has a small number of high-risk test sites, its proportion is high and the systemic risk is prominent, but the original total risk score may be low and easily overlooked; even if a large test area (such as one with 500 test rooms) has a large number of high-risk test sites, its proportion may not be high, but the absolute number is large and the supervision is difficult. If it is calculated only by proportion, its overall risk may be underestimated.

[0105] The pre-set test area type coefficient configuration table can be a structured data table used to store the mapping relationship between different test types and their corresponding type coefficients. The design of the pre-set test area type coefficient configuration table is based on the social impact, stake, and regulatory requirements of the test, aiming to quantify the differences in the sensitivity of different test tasks to risk assessment.

[0106] In one specific implementation, the contents of the preset examination area type coefficient configuration table can be shown in Table 10.

[0107] Table 10 The type coefficients in the preset test area type coefficient configuration table can be dynamic factors used to adjust the overall risk score of a test area. Their values ​​reflect the risk weight level of the test task in the current scenario. By using type coefficients, we can avoid applying a "one-size-fits-all" approach to all tests, ensuring that high-stakes tests receive more attention; and we can improve the accuracy of risk perception, so that the same original risk score reflects different actual threats in different test contexts.

[0108] The Preset Examination Area Historical Compliance Coefficient Configuration Table is a structured data table used to store the mapping relationship between the average cheating rate range of an examination area within a preset time period (e.g., the past 3 examination cycles or 3 years) and its corresponding historical compliance coefficient. Based on the principle that "historical performance affects current risk weight," the Preset Examination Area Historical Compliance Coefficient Configuration Table aims to introduce a risk inertia mechanism to prevent repeated neglect of examination areas that "fail to correct their behavior."

[0109] In one specific implementation, the specific contents of the preset test area historical compliance coefficient configuration table can be shown in Table 11.

[0110] Table 11 The historical compliance coefficient in the preset test area historical compliance coefficient configuration table is a dynamic factor used to adjust the current risk score of the test area. Its value reflects the overall compliance level and risk inertia of the test area in historical examinations.

[0111] For example, if the average cheating rate in a test area has been consistently below 2% over the past three years, it indicates that its invigilation system is effective, and therefore a low historical compliance coefficient (e.g., 0.9) is assigned to it, representing a "compliance reward." Conversely, if the average cheating rate exceeds 10%, it is considered a serious dereliction of duty and a higher coefficient (e.g., 1.2) is assigned to it, representing a "violation penalty."

[0112] By introducing a risk inertia mechanism through historical compliance coefficients, we can avoid the situation where "one abnormality determines one's fate" and prevent "long-term violations from going unnoticed." Furthermore, we can improve risk prediction capabilities, as historical data is an important reference for future risks, and early warnings can be provided by amplifying the coefficients.

[0113] Finally, the regional calibration coefficient is generated by multiplying the determined size coefficient, type coefficient, and historical compliance coefficient.

[0114] The preset full score threshold for the test area is a benchmark constant used to normalize the original total risk score of the test area. Its physical meaning is: the theoretical maximum risk score that can be achieved if all test sites under the jurisdiction of the test area are judged to be high risk.

[0115] Standardizing risk scoring across examination areas is achieved by setting a maximum score threshold for each examination area. Due to significant differences in the number of examination centers across different areas, directly comparing raw total risk scores would result in larger examination areas naturally having "higher scores," making it impossible to fairly assess the true risk level. By dividing the actual total risk score by the examination area's own maximum score threshold, it can be mapped to a unified range of 0% to 100%, thereby generating a horizontally comparable suspected cheating rate for each examination area.

[0116] In one specific implementation, the full score threshold for an examination area = the total number of examination sites in the examination area × the basic contribution value for high-risk areas × the weighting coefficient.

[0117] In one specific implementation, the expression for the multidimensional risk scoring formula for the examination area can be: Among them, is the suspected cheating rate of the examination area, is the number of examination sites at each risk level, is the third basic contribution value, is the third weighting coefficient, is the regional calibration coefficient, and is the preset full score threshold for the examination area.

[0118] The multi-dimensional risk scoring formula for test areas abandons the traditional crude model of "simple averaging" or "counting high-risk test points" and instead adopts a graded weighted aggregation mechanism—assigning different basic contribution values ​​and risk weighting coefficients to high, medium and low-risk test points, so that the risk weight of each test point is strictly matched with its actual degree of suspicion.

[0119] By using regional calibration coefficients (generated by integrating scale, type, and historical compliance), the multidimensional risk scoring formula for examination areas can adaptively adjust the risk sensitivity of different examination areas. Furthermore, it ensures the fair comparability of results across examination areas by using the examination area's full score threshold (i.e., the theoretical maximum risk score for that examination area) as the normalized denominator, compressing the original total risk score to the range of 0% to 100%.

[0120] In step 150, based on the suspected cheating rates of the test areas under the jurisdiction of the regional management center, the test areas under the jurisdiction of the regional management center are divided into multiple test area risk levels. The test areas are then weighted and aggregated based on the number of test sites at each risk level, and combined with the calibration coefficient of the regional management center to generate the regional suspected cheating rate.

[0121] The regional management center is the next higher-level management unit above the examination area in the examination supervision system. It is responsible for coordinating the organization of examinations and risk monitoring in multiple examination areas under its jurisdiction. Its scope can be flexibly defined according to actual supervision needs (e.g., by geographical area, number of candidates, or management structure) and is not specifically defined by a certain administrative level.

[0122] The Regional Management Center Calibration Coefficient is an adjustment factor used to comprehensively correct the aggregated regional risk value. It is quantitatively fused through weighted summation, product fusion, or machine learning models of regional size coefficient, regional equilibrium coefficient, exam type coefficient, and regional historical compliance coefficient. By providing a multi-dimensional characterization and dynamic adaptation of regional characteristics, it significantly improves the objectivity and regulatory guidance value of the regional suspected cheating rate.

[0123] Furthermore, in some embodiments, the multi-level suspected cheating risk calculation method further includes the following steps: (1) Query the regional scale coefficient corresponding to the number of examination rooms included in the regional management center in the preset regional scale coefficient configuration table; (2) Query the regional balance coefficient corresponding to the regional characteristics of the regional management center in the preset regional balance coefficient configuration table; (3) Query the examination type coefficient corresponding to the examination type undertaken by the regional management center in the preset regional type coefficient configuration table; (4) Query the historical compliance coefficient of the region corresponding to the average cheating rate of the region management center in the historical time period in the preset region historical compliance coefficient configuration table; (5) The calibration coefficient of the regional management center is determined by quantitatively integrating the regional scale coefficient, regional balance coefficient, examination type coefficient and regional historical compliance coefficient.

[0124] The preset regional scale coefficient configuration table can be a pre-defined reference table used to determine the scale effect adjustment factor based on the number of examination rooms under the jurisdiction of a regional management center. This table divides the total number of examination rooms into different intervals (such as small, medium, large, and super-large) and assigns a corresponding regional scale coefficient to each interval to reflect the impact of examination scale on risk assessment results.

[0125] In one specific implementation, the contents of the preset area size coefficient configuration table can be shown in Table 12.

[0126] Table 12 The regional size coefficient in the preset regional size coefficient configuration table is a dimensionless parameter between 0 and 2, used to correct risk assessment bias caused by differences in regional size.

[0127] The pre-set regional equilibrium coefficient configuration table is a pre-defined reference table used to determine the risk distribution uniformity adjustment factor based on the spatial distribution characteristics of the examination areas under the jurisdiction of the regional management center. This table categorizes regional characteristics into different types (e.g., predominantly urban, balanced between urban and rural areas, predominantly rural) and assigns a corresponding regional equilibrium coefficient to each type.

[0128] In one specific implementation, the contents of the preset regional equilibrium coefficient configuration table can be shown in Table 13.

[0129] Table 13 The regional equilibrium coefficient in the preset regional equilibrium coefficient configuration table is a dimensionless parameter greater than or equal to 1, used to correct risk assessment bias caused by uneven spatial distribution of examination rooms.

[0130] The preset regional type coefficient configuration table is a pre-defined reference table used to determine the risk sensitivity adjustment factor based on the types of examinations currently undertaken by the regional management center. This table maps different examination types (such as national qualification examinations, provincial unified examinations, academic proficiency tests, etc.) to a corresponding examination type coefficient to reflect the inherent differences in social attention, intensity of competition, and strength of cheating motives among different examinations.

[0131] In one specific implementation, the contents of the preset region type coefficient configuration table can be shown in Table 14.

[0132] Table 14 The exam type coefficient in the preset regional type coefficient configuration table is a dimensionless parameter greater than or equal to 1, which is used to amplify the exam scenario adaptability of the regional risk assessment results.

[0133] The Preset Regional Historical Compliance Coefficient Configuration Table is a pre-defined reference table used to determine the compliance adjustment factor of a regional management center based on its average cheating rate over a historical period. This table divides the historical cheating rate into different intervals (e.g., excellent compliance, moderate compliance, minor violations, serious violations) and assigns a corresponding regional historical compliance coefficient to each interval to reflect the integrity level and regulatory effectiveness of past examination organization in that region.

[0134] In one specific implementation, the contents of the preset area historical compliance coefficient configuration table can be shown in Table 15.

[0135] Table 15 The regional historical compliance coefficient in the preset regional historical compliance coefficient configuration table is a dimensionless parameter between 0.9 and 1.2, used to correct historical credit for the current risk assessment results.

[0136] The calibration coefficient of the regional management center is determined by the integration of four sub-coefficients: regional scale coefficient, regional balance coefficient, examination type coefficient, and regional historical compliance coefficient. Each sub-coefficient is obtained by querying a preset configuration table and together supports the dynamic calibration of risk assessment.

[0137] The regional suspected cheating rate is a comprehensive risk indicator generated by aggregating risk information from all examination areas under the jurisdiction of the regional management center. It is used to characterize the probability of abnormal behavior occurring in the overall examination environment of the region. The regional suspected cheating rate is calculated by integrating the risk level distribution of each examination area, the weight of the number of examination sites, and the regional characteristic calibration coefficient, aiming to provide a quantitative basis for high-level regulatory decisions.

[0138] By using the regional suspected cheating rate as a high-level risk indicator, a quantitative mapping from individual abnormal behavior to the overall risk situation of the region is achieved. The regional suspected cheating rate comprehensively considers the risk distribution of the examination area under its jurisdiction, the weight of the examination site size, and regional characteristic calibration factors, effectively overcoming the limitation of traditional supervision that "only sees the trees and not the forest."

[0139] The regional suspected cheating rate can support tiered decision-making, providing regional management centers with a comprehensive and comparable risk view, facilitating the priority allocation of inspection resources to high-risk areas. Furthermore, by integrating the weighted average of the number of test centers with the regional calibration coefficient, it avoids the problems of "large test areas overwhelming small test areas" or "regional differences distorting the data" caused by simple averaging, thus improving the accuracy of the assessment. It also achieves a closed loop of risk transmission, connecting the entire risk aggregation path of "students - test rooms - test centers - test areas - regional management centers," supporting the upgrading of education examination supervision towards refinement and intelligence.

[0140] Specifically, in some implementations, the step of "dividing the examination areas under the jurisdiction of the regional management center into multiple examination area risk levels based on the suspected cheating rates of the examination areas under the jurisdiction of the regional management center, and weighting and aggregating the number of examination sites based on the risk level of each examination area, and combining the calibration coefficient of the regional management center to generate the regional suspected cheating rate" may include the following steps: (1) Based on the suspected cheating rate and preset discrimination threshold of the examination areas under the jurisdiction of the regional management center, the examination areas under the jurisdiction of the regional management center are divided into high-risk examination areas, medium-risk examination areas and low-risk examination areas respectively; (2) Query the fourth basic contribution value and fourth weighting coefficient corresponding to high-risk, medium-risk and low-risk examination areas in the preset regional risk aggregation parameter table; (3) Generate the suspected cheating rate of the region based on the fourth basic contribution value, the fourth weighting coefficient, the calibration coefficient of the regional management center, the preset regional full score threshold and the regional multidimensional risk scoring formula.

[0141] The preset differentiation thresholds serve as boundary reference values ​​for classifying the risk levels of examination areas, including a first threshold between low and medium risk, and a second threshold between medium and high risk. These thresholds can be preset based on historical examination data statistical analysis, regional baseline levels of cheating behavior, regulatory strategy objectives, or expert experience, and can be dynamically adjusted according to examination type, seasonal fluctuations, or regional characteristics.

[0142] By comparing the suspected cheating rate of each examination area with the preset discrimination threshold, the risk level of the examination areas under its jurisdiction can be automatically classified (high-risk examination area, medium-risk examination area, and low-risk examination area), providing structured input for subsequent weighted aggregation and resource scheduling.

[0143] The preset regional risk aggregation parameter table is a pre-configured parameter comparison table used to quantify the contribution of examination areas with different risk levels in the calculation of the regional suspected cheating rate. Based on the risk level of the examination area (high risk, medium risk, low risk), the table sets its corresponding fourth basic contribution value and fourth weighting coefficient, which serve as the input basis for subsequent weighted aggregation calculations.

[0144] In one specific implementation, the contents of the preset regional risk aggregation parameter table can be shown in Table 16.

[0145] Table 16 The fourth basic contribution value in the preset regional risk aggregation parameter table represents the baseline contribution of a certain risk level examination area to the overall regional risk without other adjustment factors, reflecting the initial intensity of its potential risk impact. For example, the basic contribution value of a high-risk examination area is 50, which is much higher than the 8 of a low-risk examination area, reflecting its higher risk weight.

[0146] The fourth weighting coefficient in the pre-defined regional risk aggregation parameter table is used as an adjustment factor to dynamically amplify or reduce the suspected cheating rate in each examination area, reflecting the sensitivity and influence of this type of examination area in the overall risk assessment. For example, the weighting coefficient for a high-risk examination area is 2.2, indicating that the impact of its abnormal behavior on regional risk is significantly amplified.

[0147] By querying a preset regional risk aggregation parameter table, the system can automatically obtain two parameters corresponding to the risk level and perform weighted aggregation based on the actual number of test areas, thereby achieving a refined calculation of the suspected cheating rate in a region. The two parameters can be adjusted based on historical data, regulatory strategies, or regional characteristics, supporting flexible adaptation to different testing scenarios.

[0148] In one specific implementation, the preset regional full score threshold = the number of all examination areas in the regional management center × the high-risk basic contribution value × the weighting coefficient.

[0149] In one specific implementation, the formula for regional multidimensional risk scoring can be expressed as: Among them, is the suspected cheating rate in the region, is the number of test areas at each risk level, is the fourth basic contribution value, is the fourth weighting coefficient, is the calibration coefficient of the regional management center, and is the preset full score threshold for the region.

[0150] Please see Figure 2 , Figure 2 This illustration shows a schematic diagram of a multi-level suspected cheating risk calculation device provided in an embodiment of this application. The multi-level suspected cheating risk calculation device 200 includes: a first module 210, a second module 220, a third module 230, a fourth module 240, and a fifth module 250. Specifically: The first module 210 is used to perform hierarchical weighting, dynamic correction and normalization processing on the alarm behavior data of each candidate in the target examination room to determine the suspected cheating rate of each candidate. The second module 220 is used to divide candidates in the target examination room into multiple candidate risk levels based on the suspected cheating rate of candidates, and to perform weighted aggregation and dynamic correction of the group risk based on the distribution of candidates at each candidate risk level and concentration perception to generate the suspected cheating rate of the examination room. The third module 230 is used to divide the target examination rooms under the test center into multiple examination room risk levels according to the suspected cheating rate of each examination room under the test center, and to generate the suspected cheating rate of the test center by weighted aggregation and dynamic correction of the group risk based on the examination room composition of each examination room risk level and concentration perception. The fourth module 240 is used to divide the test centers under the jurisdiction of the test center into multiple test center risk levels according to the suspected cheating rate of each test center, and to generate the suspected cheating rate of the test center based on the number of test centers of each risk level and combined with the regional calibration coefficient. Module 5, 250, is used to divide the examination areas under the jurisdiction of the regional management center into multiple examination area risk levels based on the suspected cheating rate of each examination area. It then performs weighted aggregation based on the number of examination sites at each risk level and combines it with the calibration coefficient of the regional management center to generate the regional suspected cheating rate.

[0151] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the above-described device and module can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0152] Please see Figure 3 , Figure 3 The diagram shows a computer-readable storage medium 300 provided in an embodiment of this application. The computer-readable storage medium 300 stores program code, which can be called by a processor to execute the multi-level suspected cheating risk calculation method described in the above method embodiment.

[0153] The computer-readable storage medium 300 may be an electronic memory such as flash memory, EEPROM (Electrically Erasable Programmable Read-Only Memory), EPROM, hard disk, or ROM. Optionally, the computer-readable storage medium 300 includes a non-transitory computer-readable storage medium. The computer-readable storage medium 300 has storage space for program code 310 that performs any of the method steps described above. This program code can be read from or written to one or more computer program devices. The program code 310 may be compressed, for example, in a suitable form.

Claims

1. A multi-level suspected cheating risk calculation method, characterized in that, The method includes: Based on the alarm behavior data of each candidate in the target examination room, the data is subjected to hierarchical weighting, dynamic correction and normalization to determine the suspected cheating rate of each candidate. Based on the suspected cheating rate of the candidates, the candidates in the target examination room are divided into multiple candidate risk levels, and the group risk is weighted and aggregated and dynamically corrected based on the distribution of candidates at each candidate risk level to generate the suspected cheating rate of the examination room. Based on the suspected cheating rate of each target examination room under the test center, the target examination rooms under the test center are divided into multiple examination room risk levels. Based on the examination room composition of each examination room risk level, the group risk is weighted and aggregated and dynamically corrected by concentration perception to generate the suspected cheating rate of the test center. Based on the suspected cheating rate of each test center under the jurisdiction of the test area, the test centers under the jurisdiction of the test area are divided into multiple test center risk levels. The test centers are then weighted and aggregated based on the number of test centers at each risk level, and combined with the regional calibration coefficient to generate the suspected cheating rate of the test area. Based on the suspected cheating rates of the examination areas under the jurisdiction of the regional management center, the examination areas under the jurisdiction of the regional management center are divided into multiple examination area risk levels. The risk level of each examination area is weighted and aggregated based on the number of examination sites, and combined with the calibration coefficient of the regional management center, to generate the regional suspected cheating rate.

2. The multi-level suspected cheating risk calculation method according to claim 1, characterized in that, The process of performing hierarchical weighting, dynamic correction, and normalization on the alarm behavior data of each candidate in the target examination room to determine the suspected cheating rate for each candidate includes: Acquire alarm behavior data for each examinee in the target examination room, wherein the alarm behavior data includes at least alarm type, behavior trigger frequency, and behavior combination; In the preset alarm behavior basic information table, query the basic score and behavior weight corresponding to the alarm type; The frequency coefficient is determined based on the frequency of the behavior triggering. In the preset behavior combination basic information table, query the correlation coefficient corresponding to the behavior combination; Based on the base score, the behavior weight, the frequency coefficient, the correlation coefficient, the preset full score threshold for candidates, and the candidate multidimensional risk scoring formula, the candidate suspected cheating rate corresponding to each candidate is determined.

3. The multi-level suspected cheating risk calculation method according to claim 1, characterized in that, The process involves classifying candidates in the target examination room into multiple risk levels based on the suspected cheating rate, and dynamically adjusting the group risk based on the distribution of candidates at each risk level, thereby generating the suspected cheating rate for the examination room. This includes: Based on the suspected cheating rate of the candidates and the preset classification threshold, the candidates in the target examination room are divided into high-risk candidates, medium-risk candidates and low-risk candidates. The target number of examinees per examination room is determined based on the sum of the number of high-risk examinees and the number of medium-risk examinees. Query the first basic contribution value and first weighting coefficient corresponding to the high-risk candidate, the medium-risk candidate and the low-risk candidate respectively in the preset level assessment parameter table; Based on the ratio between the target number of examinees in the examination room and the number of examinees in the target examination room, the risk distribution coefficient of the target examination room is determined in the preset risk distribution coefficient configuration table. The suspected cheating rate of the examination room is determined based on the number of candidates with each risk level, the first basic contribution value, the first weighting coefficient, the risk distribution coefficient, the preset full score threshold for the examination room, and the examination room multidimensional risk scoring formula.

4. The multi-level suspected cheating risk calculation method according to claim 1, characterized in that, The process involves classifying the target examination rooms under the examination center into multiple examination room risk levels based on the suspected cheating rate of each examination room, and dynamically adjusting the group risk based on the examination room composition of each risk level, thereby generating the suspected cheating rate of the examination center, including: Based on the suspected cheating rate and preset grading threshold of the target examination rooms under the test center, the target examination rooms under the test center are divided into high-risk examination rooms, medium-risk examination rooms and low-risk examination rooms. Query the second basic contribution value and second weighting coefficient corresponding to the high-risk examination room, the medium-risk examination room and the low-risk examination room respectively in the preset examination site risk aggregation parameter table; Based on the number of target examination rooms under the examination center, determine the scale coefficient in the preset scale coefficient configuration table; Based on the ratio between the number of high-risk examination rooms and the number of target examination rooms under the examination site, the group risk coefficient of the examination site is determined in the preset group risk coefficient configuration table. The suspected cheating rate of the test center is generated based on the number of test rooms for each candidate's risk level, the second basic contribution value, the second weighting coefficient, the scale coefficient, the group risk coefficient, the preset full score threshold for the test center, and the test center multidimensional risk scoring formula.

5. The multi-level suspected cheating risk calculation method according to claim 1, characterized in that, The process involves dividing the test centers within a test area into multiple risk levels based on their suspected cheating rates, weighting and aggregating the risk levels based on the number of test centers at each risk level, and combining this with a regional calibration coefficient to generate the test area's suspected cheating rate. This includes: Based on the suspected cheating rate and preset classification threshold of the test sites under the jurisdiction of the test area, the test sites under the jurisdiction of the test area are divided into high-risk test sites, medium-risk test sites and low-risk test sites. Query the third basic contribution value and the third weighting coefficient corresponding to the high-risk test point, the medium-risk test point and the low-risk test point respectively in the preset test area risk aggregation parameter table; The suspected cheating rate of the examination area is generated based on the third basic contribution value, the third weighting coefficient, the regional calibration coefficient, the preset full score threshold of the examination area, and the examination area multidimensional risk scoring formula.

6. The multi-level suspected cheating risk calculation method according to claim 5, characterized in that, The method further includes: Query the scale coefficient corresponding to the number of examination rooms included in the examination area in the preset examination area scale coefficient configuration table; Query the type coefficient corresponding to the examination type undertaken by the examination area in the preset examination area type coefficient configuration table; Query the historical compliance coefficient corresponding to the average cheating rate of the test area within the preset time period in the preset test area historical compliance coefficient configuration table; The regional calibration coefficient is determined by quantitatively fusing the scale coefficient, the type coefficient, and the historical compliance coefficient.

7. The multi-level suspected cheating risk calculation method according to claim 1, characterized in that, The process involves dividing the examination areas under the jurisdiction of the regional management center into multiple risk levels based on the suspected cheating rate of each examination area, weighting and aggregating the risk levels based on the number of examination sites in each risk level, and combining this with the regional management center's calibration coefficient to generate the regional suspected cheating rate, including: Based on the suspected cheating rate and preset discrimination threshold of the examination areas under the jurisdiction of the regional management center, the examination areas under the jurisdiction of the regional management center are divided into high-risk examination areas, medium-risk examination areas and low-risk examination areas. Query the fourth basic contribution value and fourth weighting coefficient corresponding to the high-risk examination area, the medium-risk examination area and the low-risk examination area respectively in the preset regional risk aggregation parameter table; The suspected cheating rate of the region is generated based on the fourth basic contribution value, the fourth weighting coefficient, the regional management center calibration coefficient, the preset regional full score threshold, and the regional multidimensional risk scoring formula.

8. The multi-level suspected cheating risk calculation method according to claim 7, characterized in that, The method further includes: Query the preset regional scale coefficient configuration table to find the regional scale coefficient corresponding to the number of examination rooms included in the regional management center; Query the regional balance coefficient corresponding to the regional characteristics of the regional management center in the preset regional balance coefficient configuration table; Query the examination type coefficient corresponding to the examination type undertaken by the regional management center in the preset regional type coefficient configuration table; Query the historical compliance coefficient of the region corresponding to the average cheating rate of the region management center within the historical time period in the preset region historical compliance coefficient configuration table; The calibration coefficient of the regional management center is determined by quantitatively integrating the regional scale coefficient, the regional balance coefficient, the examination type coefficient, and the regional historical compliance coefficient.

9. A multi-level suspected cheating risk calculation device, characterized in that, The device includes: The first module is used to perform hierarchical weighting, dynamic correction and normalization processing on the alarm behavior data of each candidate in the target examination room to determine the suspected cheating rate of each candidate. The second module is used to divide the candidates in the target examination room into multiple candidate risk levels based on the suspected cheating rate of the candidates, and to perform weighted aggregation and dynamic correction of the group risk based on the distribution of candidates at each candidate risk level and concentration perception to generate the suspected cheating rate of the examination room. The third module is used to divide the target examination rooms under the test center into multiple examination room risk levels according to the suspected cheating rate of each examination room under the test center, and to generate the suspected cheating rate of the test center by weighted aggregation and dynamic correction of the group risk based on the examination room composition of each examination room risk level and concentration perception. The fourth module is used to divide the test centers under the jurisdiction of the test area into multiple test center risk levels according to the suspected cheating rate of the test centers respectively, and to generate the suspected cheating rate of the test area based on the number of test centers of each risk level and combined with the regional calibration coefficient. The fifth module is used to divide the examination areas under the jurisdiction of the regional management center into multiple examination area risk levels based on the suspected cheating rate of each examination area, and to generate the regional suspected cheating rate by weighting and aggregating the number of examination sites for each examination area risk level and combining the calibration coefficient of the regional management center.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium contains program code, which can be called by a processor to execute the multi-level suspected cheating risk calculation method as described in any one of claims 1-8.

Images