Online examination cheating behavior detection method and system

Abstract

The invention provides an online examination cheating behavior detection method and system, and relates to the technical field of examination behavior detection, and the specific steps are as follows: collecting answer texts of examinees in real time, analyzing the answer texts of the examinees by using an NLP technology, and calculating the plagiarism probability of the answer texts by comparing the texts with data in a database; monitoring the network activity of the examinee, analyzing the network behavior mode of the examinee, and calculating the abnormal behavior mode index of the examinee; physiological data of examinees are collected, and psychological stress indexes of the examinees are calculated; calculating a comprehensive evaluation index by using weighted average, comparing the comprehensive evaluation index with an examination cheating threshold value, and judging that the behavior is cheating if the comprehensive evaluation index exceeds the threshold value; according to the invention, the behaviors of examinees are monitored in all directions, and a three-dimensional cheating detection system is provided; the artificial intelligence technology is applied, answer texts, network activities and psychological stress indexes are automatically analyzed, dependence on manual invigilation is reduced, and the monitoring efficiency and accuracy are improved; and cheating behaviors can be found in time.

Description

Technical Field

[0001] The present invention relates to the technical field of examination behavior detection, and in particular to an online examination cheating behavior detection method and system. Background Art

[0002] In the current online examination environment, with the rapid development of network technology, cheating in exams occurs frequently, and traditional proctoring methods can no longer effectively curb such phenomena. Many online examination platforms rely too much on manual supervision, resulting in incomplete monitoring of examinees' behavior during the exam.

[0003] Existing plagiarism detection technologies mainly focus on the surface comparison of text content, often ignoring the changes in the online behavior patterns and physiological states of candidates, which makes cheating difficult to detect in a timely manner to a certain extent. For example, candidates may obtain answers through other means, or frequently search for information online during the exam, and traditional proctoring methods cannot effectively capture these dynamic behaviors. In addition, the lack of real-time monitoring of candidates' psychological states has further challenged the fairness of the exam and brought many troubles to education management.

[0004] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not constitute the prior art that is already known to one of ordinary skill in the art. Summary of the invention

[0005] The purpose of the present invention is to provide a method and system for detecting cheating in online examinations to solve the problems raised in the above background technology.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A method and system for detecting cheating in online exams, the specific steps comprising:

[0008] Step 1: Collect the candidates’ answer texts in real time, use NLP technology to analyze the candidates’ answer texts, detect plagiarism by comparing the answer texts with the standard answer data in the database, and calculate the probability of plagiarism in the candidates’ answer texts;

[0009] Step 2: Monitor the candidate's online activities, record key information, analyze the candidate's online behavior pattern, and calculate the candidate's abnormal behavior pattern index;

[0010] Step 3: Use sensors to collect the examinee's physiological data and calculate the examinee's psychological stress index;

[0011] Step 4: The probability of plagiarism of answer text, abnormal behavior pattern index and psychological stress index are integrated, and a comprehensive evaluation index is calculated using weighted average. The index is then compared with the preset cheating threshold for the exam. If the threshold is exceeded, it is considered cheating behavior.

[0012] Step 5: Show the test results to the invigilator and take appropriate measures based on the test results, such as warning, recording or disqualification.

[0013] Furthermore, the process of collecting the examinee's answer text in real time and analyzing the examinee's answer text using NLP technology is specifically as follows:

[0014] In the examination system, a text input box is set up to receive the candidate's answer text in real time. Regular expressions are used to remove redundant blank characters in the text. For English text, it is uniformly converted to lowercase. Regular expressions are used to remove punctuation marks and other non-text characters. Jieba is used to perform word segmentation, and the word segmentation results are output to form a vocabulary list. Then, the frequency of each word in the candidate's answer is calculated to obtain the vocabulary frequency of each word in the candidate's answer.

[0015] For the test standard answer data in a given database, the word segmentation process is also performed in the above manner to obtain a vocabulary list, and the frequency of each word in the database data is calculated to obtain the vocabulary frequency of each word in the database data;

[0016] All the identical words in the examinee's answers and the database data are extracted as comparison words, and the total number of comparison words is counted.

[0017] Furthermore, the formula for detecting plagiarism by comparing the answer text with the information in the database and calculating the probability of plagiarism of the examinee's answer text is:

[0018]

[0019] Among them, PSI represents the probability of plagiarism of the candidate's answer text, F i is the lexical frequency of the ith contrasting word in the examinee’s answer, G i is the lexical frequency of the i-th contrasting vocabulary in the test standard answer data, i is the index of the contrasting vocabulary, and i=1,2,…,n, n is the total number of contrasting vocabulary.

[0020] Furthermore, the process of analyzing the network behavior pattern of the examinee and calculating the examinee's abnormal behavior pattern index is specifically as follows:

[0021] Obtain the actual network activity data of the examinee and the normal network activity data corresponding to the normal network activity;

[0022] The actual network activity data is obtained by: collecting the actual network activity data through a network monitoring tool, wherein the network activity data includes data packet size, transmission rate, connection frequency and test duration;

[0023] The standard network activity data is obtained by: determining the normal network activity data corresponding to the normal network activity based on historical data and statistical analysis;

[0024] The abnormal behavior pattern recognition formula is as follows:

[0025]

[0026] Among them, BEH is the abnormal behavior pattern index, N j is the jth type of data in the actual network activity data, P j is the j-th type of data in the normal network activity data, j is the index of the data type in the network activity data, and j=1,2,3,4.

[0027] Furthermore, the process of using sensors to collect the examinee's physiological data and calculate the examinee's psychological stress index is specifically as follows:

[0028] Using a biosensor to monitor the examinee's physiological signals in real time, the physiological signals include the examinee's heartbeat number α and breathing rate β per minute; comparing the recorded heartbeat number α with a preset heart rate threshold, if it exceeds the threshold range, calculating the difference Δα between the examinee's heart rate and the preset heart rate range; comparing the recorded breathing rate β with a preset breathing rate threshold, if it exceeds the threshold range, calculating the difference Δβ between the examinee's breathing rate and the preset breathing rate range;

[0029] The pressure index calculation formula is as follows:

[0030]

[0031] Among them, HRV is the psychological stress index, Δα is the difference between the examinee's heart rate and the preset heart rate range, Δβ is the difference between the examinee's breathing rate and the preset breathing rate range, μ is the preset proportional coefficient of Δα, ρ is the preset proportional coefficient of Δβ, μ>ρ>0, and C1 is the constant correction index.

[0032] Furthermore, the formula for comprehensive evaluation using weighted average is as follows:

[0033] F total =w1*PSI+w2*BEH+w3*HRV

[0034] Among them, F totalis the comprehensive evaluation index, PSI is the probability of answer text plagiarism, BEH is the abnormal behavior pattern index, HRV is the psychological stress index, w1, w2, and w3 are the weights of PSI, BEH, and HRV respectively;

[0035] Furthermore, the specific logic of generating the weights of each parameter in the comprehensive evaluation model through the hierarchical analysis method is as follows:

[0036] The three indicators of answer text plagiarism probability, abnormal behavior pattern index and psychological stress index are marked, and the relative importance values ​​between them are determined by the nine-scale method to construct a judgment matrix, in which the indicator of answer text plagiarism probability PSI is marked as 1, the indicator of abnormal behavior pattern index BEH is marked as 2, and the indicator of psychological stress index HRV is marked as 3. The constructed judgment matrix is:

[0037] B=[b hk ] 3×3

[0038] Among them, h, k are the labels of the indicators, b hk represents the importance of index h to index k, and b hh =1,

[0039] Each element in the judgment matrix is ​​normalized, and the specific formula of the normalization process is:

[0040]

[0041] Among them, c hk for b hk Normalized value;

[0042] Furthermore, the weights of the answer text plagiarism probability, abnormal behavior pattern index and psychological stress index are calculated based on the following formula:

[0043]

[0044] Among them, c 1k represents the importance of the probability of plagiarism of the answer text after normalization relative to the indicator k, c 2k represents the importance of the abnormal behavior pattern index after normalization relative to the index k, c 3k It indicates the importance of the normalized psychological stress index relative to the index k.

[0045] Furthermore, the comprehensive evaluation index F total With the preset threshold F cheat Compare and if the threshold is exceeded, it is determined that the examinee has cheated:

[0046] When F total≤F cheat When the examinee is judged to have no cheating behavior;

[0047] When F total >F cheat When the test is conducted, the examinee is judged to have cheated.

[0048] An online examination cheating behavior detection system, the online examination cheating behavior detection system is used to execute the above-mentioned online examination cheating behavior detection method, comprising:

[0049] The content security analysis module is used to collect the answer texts of candidates in real time, analyze the answer texts of candidates using NLP technology, detect plagiarism by comparing the answer texts with the standard answer materials in the database, and calculate the probability of plagiarism of the answer texts of candidates;

[0050] Social network analysis module, used to monitor the network activities of candidates, record key information, analyze the network behavior patterns of candidates, and calculate the abnormal behavior pattern index of candidates;

[0051] The psychological stress analysis module is used to collect the examinee's physiological data using sensors and calculate the examinee's psychological stress index;

[0052] The data fusion and decision-making module is used to fuse the probability of answer text plagiarism, abnormal behavior pattern index and psychological pressure index, use weighted average to calculate the comprehensive evaluation index, and compare it with the preset cheating threshold of the exam. If it exceeds the threshold, it is judged as cheating behavior;

[0053] The result feedback and processing module is used to display the test results to the invigilator and take corresponding measures based on the test results, such as warning, recording or disqualification from the exam.

[0054] Compared with the prior art, the present invention has the following beneficial effects:

[0055] The present invention can monitor examinee behavior in all directions, from text content, network behavior to physiological response, by analyzing examinee's answer text, monitoring network activities and evaluating psychological stress, thus providing a three-dimensional cheating detection system; applying artificial intelligence technology, answer text, network activities and psychological stress indicators can be automatically analyzed, thus reducing dependence on manual proctoring and improving monitoring efficiency and accuracy; cheating behavior can be discovered and intervened in time, thus playing a certain preventive role; this comprehensive cheating detection method provides a more refined, intelligent and comprehensive online examination monitoring system, which not only improves the accuracy of cheating detection, but also enhances the security and credibility of remote examinations, thereby maintaining the fairness of examinations; in this way, the present invention not only improves the security and reliability of online examinations, but also provides strong technical support for proctors, reduces their work pressure and improves proctoring efficiency; these beneficial effects jointly promote the healthy development of online examination systems and ensure the fairness and effectiveness of educational evaluation. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 It is a schematic diagram of the overall method flow of the present invention;

[0057] Figure 2 It is a schematic diagram of the overall system structure of the present invention. DETAILED DESCRIPTION

[0058] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with specific embodiments.

[0059] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. The words "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0060] Example:

[0061] See also Figure 1 , the present invention provides a technical solution:

[0062] A method for detecting cheating in online examinations, the specific steps comprising:

[0063] Step 1: Collect the candidates’ answer texts in real time and use NLP technology to analyze the candidates’ answer texts; detect plagiarism by comparing the answer texts with the information in the database and calculate the probability of plagiarism in the candidates’ answer texts;

[0064] In this embodiment, the process of collecting the examinee's answer text in real time and analyzing the examinee's answer text using NLP technology is specifically as follows:

[0065] In the examination system, a text input box is set up to receive the candidate's answer text in real time. Regular expressions are used to remove redundant blank characters in the text. For English text, it is uniformly converted to lowercase. Regular expressions are used to remove punctuation marks and other non-text characters. Jieba is used to perform word segmentation, and the word segmentation results are output to form a vocabulary list. Then, the frequency of each word in the candidate's answer is calculated to obtain the vocabulary frequency of each word in the candidate's answer.

[0066] For the test standard answer data in a given database, the word segmentation process is also performed in the above manner to obtain a vocabulary list, and the frequency of each word in the database data is calculated to obtain the vocabulary frequency of each word in the database data;

[0067] Extract all the same words in the examinee's answers and the database data as comparison words, and count the total number of comparison words;

[0068] The formula for detecting plagiarism by comparing the answer text with the information in the database and calculating the probability of plagiarism of the candidate's answer text is:

[0069]

[0070] Among them, PSI represents the probability of plagiarism of the candidate's answer text, F i is the lexical frequency of the ith contrasting word in the examinee’s answer, G i is the lexical frequency of the i-th contrasting vocabulary in the test standard answer data, i is the index of the contrasting vocabulary, and i=1,2,…,n, n is the total number of contrasting vocabulary.

[0071] In the above formula, the same words appearing in the candidate's answer text and the standard answer material of the exam are determined, and then the similarities and differences in the probability of the same words appearing in the two are compared to determine the probability of plagiarism in the answer text. A high degree of overlap indicates that the candidate used a large number of words that are consistent with the standard answer in the answer, especially in key concepts or terms. If the candidate's answer contains the same or similar sentence structure, wording and expression as the standard answer, this further increases the possibility of plagiarism. The same words appearing in the candidate's answer text and the standard answer material of the exam are determined, and then the similarities and differences in the probability of the same words appearing in the two are compared to determine the probability of plagiarism in the answer text.

[0072] The standard answers given in the database are usually the correct answers reviewed by experts, representing a deep understanding and mastery of the test content. If the candidate's answer and the standard answer have a high degree of overlap in vocabulary, this may mean that the candidate directly used the content in the standard answer instead of thinking independently and expressing his or her own understanding. This situation is usually regarded as plagiarism because it shows that the candidate did not complete the answer on his or her own and lacks originality.

[0073] Step 2: Monitor the candidate's online activities, record key information, analyze the candidate's online behavior pattern, and calculate the candidate's abnormal behavior pattern index;

[0074] In this embodiment, the process of analyzing the network behavior pattern of the examinee and calculating the examinee's abnormal behavior pattern index is specifically as follows:

[0075] Obtain the actual network activity data of the examinee and the normal network activity data corresponding to the normal network activity;

[0076] The actual network activity data is obtained by: collecting the actual network activity data through a network monitoring tool, wherein the network activity data includes data packet size, transmission rate, connection frequency and test duration;

[0077] The standard network activity data is obtained by: determining the normal network activity data corresponding to the normal network activity based on historical data and statistical analysis;

[0078] The abnormal behavior pattern recognition formula is as follows:

[0079]

[0080] Among them, BEH is the abnormal behavior pattern index, N j is the jth type of data in the actual network activity data, P j is the jth type of data in the normal network activity data, j is the index of the data type in the network activity data, and j = 1, 2, 3, 4;

[0081] In the above formula, by comparing the actual network activity data point N i Data point P corresponding to the normal network activity model i The difference between the abnormal behavior pattern index BEH is calculated; here |N i -P i | represents the absolute difference between the actual data point and the model-predicted data point, which can be understood as the degree of deviation between the actual behavior and the normal behavior;

[0082] When N i and P i When very close, |N i -P i |Close to 0, indicating that the actual network activity is consistent with the activity predicted by the normal model, and the abnormal behavior pattern index is low;

[0083] When N i and P i When the difference is large, |N i -P i The larger the value of |, the more significant the difference between the actual network activity and the activity predicted by the normal model, and the higher the abnormal behavior pattern index;

[0084] By calculating the average deviation of all data points, we can get the overall abnormal behavior pattern index BEH. The higher the BEH, the more likely it is that the candidate's online activity is abnormal.

[0085] Step 3: Use sensors to collect the examinee's physiological data and calculate the examinee's psychological stress index;

[0086] In this embodiment, the process of collecting the examinee's physiological data using sensors and calculating the examinee's psychological stress index includes:

[0087] Using a biosensor to monitor the examinee's physiological signals in real time, the physiological signals include the examinee's heartbeat number α and breathing rate β per minute; comparing the recorded heartbeat number α with a preset heart rate threshold, if it exceeds the threshold range, calculating the difference Δα between the examinee's heart rate and the preset heart rate range; comparing the recorded breathing rate β with a preset breathing rate threshold, if it exceeds the threshold range, calculating the difference Δβ between the examinee's breathing rate and the preset breathing rate range;

[0088] The pressure index calculation formula is as follows:

[0089]

[0090] Among them, HRV is the psychological stress index, Δα is the difference between the candidate's heart rate and the preset heart rate range, Δβ is the difference between the candidate's breathing rate and the preset breathing rate range, μ is the preset proportional coefficient of Δα, and ρ is the preset proportional coefficient of Δβ. When calculating the psychological stress index, both heart rate and breathing rate can provide important information about the individual's psychological state, but the heart rate usually has a larger weight because the heart rate is more sensitive to psychological stress and can quickly reflect the body's physiological response to stress. Specifically, the heart rate will increase rapidly when under stress, reflecting the activation of the sympathetic nervous system. Although the breathing rate will also change under stress, its changes are relatively slow. Therefore, heart rate is considered to be a more direct and reliable indicator of psychological stress, so μ>ρ>0, C1 is a constant correction index; when Δα increases, that is, the difference between the examinee's heart rate and the preset heart rate range increases, the examinee's heartbeat accelerates, and the psychological stress index HRV increases accordingly; when Δβ increases, that is, the difference between the examinee's breathing rate and the preset breathing rate range increases, the psychological stress index HRV increases accordingly, which indicates that Δα, Δβ and HRV are positively correlated.

[0091] Studies have shown that candidates who cheat in exams often face greater psychological pressure. Due to fear of being discovered and the consequences, candidates may show abnormal physiological reactions. When the psychological stress index of candidates during the exam increases abnormally, this may indicate that they are engaging in inappropriate behavior, such as plagiarism or other cheating methods. Therefore, by observing the psychological stress index, it is possible to indirectly infer whether the behavior of the candidates is abnormal. Although the psychological stress index itself cannot directly prove cheating, it can be used as one of the important factors in judging the status of candidates. When combined with the probability of plagiarism of the answer text and the network behavior monitoring data for a comprehensive evaluation, if the candidate shows obvious abnormalities in the psychological stress index and there are also suspicious signs in the text similarity and behavior patterns, it can more strongly support the judgment of cheating. This multi-dimensional data analysis method helps to improve the accuracy of detection.

[0092] Step 4: The probability of plagiarism of answer text, abnormal behavior pattern index and psychological stress index are integrated, and a comprehensive evaluation index is calculated using weighted average. The index is then compared with the preset cheating threshold for the exam. If the threshold is exceeded, it is considered cheating behavior.

[0093] In this embodiment, the formula for comprehensive evaluation using weighted average is as follows:

[0094] F total =w1*PSI+w2*BEH+w3*HRV

[0095] Among them, F totalis the comprehensive evaluation index, PSI is the probability of plagiarism of the answer text, BEH is the abnormal behavior pattern index, HRV is the psychological stress index, w1, w2, w3 are the weights of PSI, BEH, and HRV respectively; when PSI increases, that is, the probability of plagiarism of the answer text increases, at this time the comprehensive evaluation index F total When BEH increases, that is, the abnormal behavior pattern index increases, the comprehensive evaluation index F total When HRV increases, the psychological stress index increases. At this time, the comprehensive evaluation index F total Corresponding increase; that is, PSI, BEH, HRV and F total There is a positive correlation;

[0096] The specific logic of generating the weights of each parameter in the comprehensive evaluation model through the hierarchical analysis method is as follows:

[0097] The three indicators of answer text plagiarism probability, abnormal behavior pattern index and psychological stress index are marked, and the relative importance values ​​between them are determined by the nine-scale method to construct a judgment matrix, in which the indicator of answer text plagiarism probability PSI is marked as 1, the indicator of abnormal behavior pattern index BEH is marked as 2, and the indicator of psychological stress index HRV is marked as 3. The constructed judgment matrix is:

[0098] B=[b hk ] 3×3

[0099] Among them, h, k are the labels of the indicators, b kl represents the importance of index h to index k, and b hh =1,

[0100] Each element in the judgment matrix is ​​normalized, and the specific formula of the normalization process is:

[0101]

[0102] Among them, c hk for b hk The normalized value;

[0103] The calculation formulas for the weights of the answer text plagiarism probability, abnormal behavior pattern index, and psychological stress index are as follows:

[0104]

[0105] Among them, c 1k represents the importance of the probability of plagiarism of the answer text after normalization relative to the indicator k, c 2k represents the importance of the abnormal behavior pattern index after normalization relative to the index k, c3k It indicates the importance of the normalized psychological stress index relative to the index k.

[0106] The comprehensive evaluation index F total With the preset threshold F cheat Compare and if the threshold is exceeded, it is determined that the examinee has cheated:

[0107] When F total ≤F cheat When the examinee is judged to have no cheating behavior;

[0108] When F total >F cheat When the test is conducted, the examinee is judged to have cheated.

[0109] Step 5: Show the test results to the invigilator and take appropriate measures based on the test results, such as warning, recording or disqualification.

[0110] The present invention further provides an online examination cheating behavior detection system, which is used to execute the above-mentioned online examination cheating behavior detection method, including:

[0111] The content security analysis module is used to collect the answer texts of candidates in real time, analyze the answer texts of candidates using NLP technology, detect plagiarism by comparing the answer texts with the standard answer materials in the database, and calculate the probability of plagiarism of the answer texts of candidates;

[0112] Social network analysis module, used to monitor the network activities of candidates, record key information, analyze the network behavior patterns of candidates, and calculate the abnormal behavior pattern index of candidates;

[0113] The psychological stress analysis module is used to collect the examinee's physiological data using sensors and calculate the examinee's psychological stress index;

[0114] The data fusion and decision-making module is used to fuse the probability of answer text plagiarism, abnormal behavior pattern index and psychological pressure index, use weighted average to calculate the comprehensive evaluation index, and compare it with the preset cheating threshold of the exam. If it exceeds the threshold, it is judged as cheating behavior;

[0115] The result feedback and processing module is used to display the test results to the invigilator and take corresponding measures based on the test results, such as warning, recording or disqualification from the exam.

[0116] The specific values ​​of μ and ρ in the formula are generally determined by technical personnel in this field according to actual conditions. The technical personnel in this field collect multiple groups of sample data, and set corresponding preset proportional coefficients for each group of sample data. The preset proportional coefficients and the collected sample data are substituted into the formula. Through repeated experiments and parameter adjustments, the accuracy of the model output and the rationality of the results are observed, and these factor coefficients are gradually adjusted. The performance and effect of the model under different parameter settings are compared to find the optimal coefficient combination, and the calculated factor coefficients are screened and averaged to obtain the values ​​of μ and ρ.

[0117] In addition, the size of the preset factor coefficient is to quantify each parameter to obtain a specific value. In order to facilitate subsequent comparison, the size of the coefficient depends on the amount of sample data and the preset proportional coefficient initially set by technical personnel in this field for each set of sample data. It is not unique as long as it does not affect the proportional relationship between the parameter and the quantized value.

[0118] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0119] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination thereof. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product. Those skilled in the art may appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein may be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed by hardware or software methods depends on the specific application and design constraints of the technical solution.

[0120] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, and may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0121] The above description is only a specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application.

Claims

1. A method for detecting cheating in online examinations, characterized in that: The specific steps include: Step 1: Collect the candidates’ answer texts in real time, use NLP technology to analyze the candidates’ answer texts, detect plagiarism by comparing the answer texts with the standard answer data in the database, and calculate the probability of plagiarism in the candidates’ answer texts; Step 2: Monitor the candidate's online activities, record key information, analyze the candidate's online behavior pattern, and calculate the candidate's abnormal behavior pattern index; Step 3: Use sensors to collect the examinee's physiological data and calculate the examinee's psychological stress index; Step 4: The probability of plagiarism of answer text, abnormal behavior pattern index and psychological stress index are integrated, and a comprehensive evaluation index is calculated using weighted average. The index is then compared with the preset cheating threshold for the exam. If the threshold is exceeded, it is considered cheating behavior. Step 5: Show the test results to the invigilator and take appropriate measures based on the test results, such as warning, recording or disqualification.

2. The method for detecting cheating in online examinations according to claim 1, characterized in that: The process of collecting the examinee's answer text in real time and analyzing the examinee's answer text using NLP technology is specifically as follows: In the examination system, a text input box is set up to receive the candidate's answer text in real time. Regular expressions are used to remove redundant blank characters in the text. For English text, it is uniformly converted to lowercase. Regular expressions are used to remove punctuation marks and other non-text characters. Jieba is used to perform word segmentation, and the word segmentation results are output to form a vocabulary list. Then, the frequency of each word in the candidate's answer is calculated to obtain the vocabulary frequency of each word in the candidate's answer. For the test standard answer data in a given database, the word segmentation process is also performed in the above manner to obtain a vocabulary list, and the frequency of each word in the database data is calculated to obtain the vocabulary frequency of each word in the database data; All the identical words in the examinee's answers and the database data are extracted as comparison words, and the total number of comparison words is counted.

3. The method for detecting cheating in online examinations according to claim 2, characterized in that: The formula for detecting plagiarism by comparing the answer text with the information in the database and calculating the probability of plagiarism of the candidate's answer text is: Among them, PSI represents the probability of plagiarism of the candidate's answer text, F i is the lexical frequency of the ith contrasting word in the examinee’s answer, G i is the lexical frequency of the i-th contrasting vocabulary in the test standard answer data, i is the index of the contrasting vocabulary, and i=1,2,…,n, n is the total number of contrasting vocabulary.

4. The method for detecting cheating in online examinations according to claim 1, characterized in that: The process of analyzing the network behavior pattern of the examinee and calculating the examinee's abnormal behavior pattern index is specifically as follows: Obtain the actual network activity data of the examinee and the normal network activity data corresponding to the normal network activity; The actual network activity data is obtained by: collecting the actual network activity data through a network monitoring tool, wherein the network activity data includes data packet size, transmission rate, connection frequency and test duration; The standard network activity data is obtained by: determining the normal network activity data corresponding to the normal network activity based on historical data and statistical analysis; The abnormal behavior pattern recognition formula is as follows: Among them, BEH is the abnormal behavior pattern index, N j is the jth type of data in the actual network activity data, P j is the j-th type of data in the normal network activity data, j is the index of the data type in the network activity data, and j=1,2,3,4.

5. The method for detecting cheating in online examinations according to claim 1, characterized in that: The process of using sensors to collect the examinee's physiological data and calculate the examinee's psychological stress index is specifically as follows: Using a biosensor to monitor the examinee's physiological signals in real time, the physiological signals including the examinee's heartbeats per minute α and breathing rate β; comparing the recorded heartbeats α with a preset heart rate threshold, and if it exceeds the threshold range, calculating the difference Δα between the examinee's heart rate and the preset heart rate range; Compare the recorded respiratory rate β with the preset respiratory rate threshold. If it exceeds the threshold range, calculate the difference Δβ between the candidate's respiratory rate and the preset respiratory rate range. The pressure index calculation formula is as follows: Among them, HRV is the psychological stress index, Δα is the difference between the examinee's heart rate and the preset heart rate range, Δβ is the difference between the examinee's breathing rate and the preset breathing rate range, μ is the preset proportional coefficient of Δα, ρ is the preset proportional coefficient of Δβ, μ>ρ>0, and C1 is the constant correction index.

6. The method for detecting cheating in online examinations according to claim 1, characterized in that: The formula for comprehensive evaluation using weighted average is as follows: <h2 style=";text-align:left;direction:ltr">F<h2 style=";text-align:left;direction:ltr"> total <h2 style=";text-align:left;direction:ltr"> = w1*PSI+w2*BEH+w3*HRV Among them, F total is a comprehensive evaluation index, PSI is the probability of answer text plagiarism, BEH is the abnormal behavior pattern index, HRV is the psychological stress index, and w1, w2, and w3 are the weights of PSI, BEH, and HRV respectively.

7. The method for detecting cheating in online examinations according to claim 6, characterized in that: The specific logic of generating the weights of each parameter in the comprehensive evaluation model through the hierarchical analysis method is as follows: The three indicators of answer text plagiarism probability, abnormal behavior pattern index and psychological stress index are marked, and the relative importance values ​​between them are determined by the nine-scale method to construct a judgment matrix, in which the indicator of answer text plagiarism probability PSI is marked as 1, the indicator of abnormal behavior pattern index BEH is marked as 2, and the indicator of psychological stress index HRV is marked as 3. The constructed judgment matrix is: B=[b hk ] 3×3 Among them, h, k are the labels of the indicators, b hk represents the importance of index h to index k, and b hh =1, Each element in the judgment matrix is ​​normalized, and the specific formula of the normalization process is: Among them, c hk for b hk The normalized value.

8. The method for detecting cheating in online examinations according to claim 7, characterized in that: The calculation formulas for the weights of the answer text plagiarism probability, abnormal behavior pattern index, and psychological stress index are as follows: Among them, c 1k represents the importance of the probability of plagiarism of the answer text after normalization relative to the indicator k, c 2k represents the importance of the abnormal behavior pattern index after normalization relative to the index k, c 2k It indicates the importance of the normalized psychological stress index relative to the index k.

9. The method for detecting cheating in online examinations according to claim 1, characterized in that: The comprehensive evaluation index F total With the preset threshold F cheat Compare and if the threshold is exceeded, it is determined that the examinee has cheated: When F total ≤F cheat When the examinee is judged to have no cheating behavior; When F total >F cheat When the test is conducted, the examinee is judged to have cheated.

10. An online examination cheating behavior detection system, characterized by: The online examination cheating behavior detection system is used to execute the online examination cheating behavior detection method according to any one of claims 1 to 9, comprising: The content security analysis module is used to collect the answer texts of candidates in real time, analyze the answer texts of candidates using NLP technology, detect plagiarism by comparing the answer texts with the standard answer materials in the database, and calculate the probability of plagiarism of the answer texts of candidates; Social network analysis module, used to monitor the network activities of candidates, record key information, analyze the network behavior patterns of candidates, and calculate the abnormal behavior pattern index of candidates; The psychological stress analysis module is used to collect the examinee's physiological data using sensors and calculate the examinee's psychological stress index; The data fusion and decision-making module is used to fuse the probability of answer text plagiarism, abnormal behavior pattern index and psychological pressure index, use weighted average to calculate the comprehensive evaluation index, and compare it with the preset cheating threshold of the exam. If it exceeds the threshold, it is judged as cheating behavior; The result feedback and processing module is used to display the test results to the invigilator and take corresponding measures based on the test results, such as warning, recording or disqualification from the exam.

Images