Online examination system

Abstract

The invention discloses an online examination system, and relates to the technical field of online examination systems. The intelligent question bank generation module is used for automatically generating a question bank by utilizing an intelligent algorithm according to knowledge points, difficulty coefficients and question type information; meanwhile, according to historical answering conditions of the examinees, difficulty and types of questions in the question bank are dynamically adjusted so as to adapt to requirements of different examinees; the intelligent invigilation module is used for monitoring behaviors of examinees in real time through a computer vision technology and automatically identifying and recording suspected cheating behaviors; meanwhile, the natural language processing technology is utilized to analyze the voice of the examinee in real time, and whether the cheating suspicion exists or not is judged. Through the machine learning algorithm, cheating behaviors can be effectively identified and prevented, the fairness of the examination is ensured, and in addition, the scoring accuracy and the reliability of the system are further improved through the application of the auto-encoder and the integrated learning technology.

Description

Technical Field

[0001] This invention relates to the field of online examination system technology, and more particularly to an online examination system. Background Technology

[0002] With the development of educational technology, the methods of education and examination are also constantly evolving, and more and more online examination systems are being developed. However, traditional online examination systems have many problems, such as low security, susceptibility to cheating, lack of effective proctoring mechanisms, unfair scoring, and insufficient system intelligence. These problems seriously affect the quality and credibility of online examinations.

[0003] A search revealed Chinese patent application CN202210884920.9, which discloses an online examination assistance system and an online examination monitoring method, belonging to the field of online examination technology. The system includes an input terminal providing an input channel for candidates to input control commands and answer information; the examination terminal includes a candidate client and a first camera device; the candidate client displays question information and receives the candidate's input answer information under control commands; the first camera device, positioned in a first position, captures the candidate's facial information so that the candidate client can perform face verification and liveness verification and monitor for the presence of cheating devices within the candidate's line of sight; an earphone blocks wireless signals to prevent cheating and monitors whether the candidate is wearing the earphone correctly; the monitoring terminal has a second camera device positioned in a second position, capturing the view of the candidate's facing direction so that the candidate client can monitor for the presence of cheating devices in the candidate's facing direction. The examination assistance system in the aforementioned patent has the following shortcomings: it uses camera footage to determine the presence of abnormal behavior but does not provide specific determination mechanisms or related solutions, and the system also lacks the function of providing targeted exam training, requiring further improvement. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies by proposing an online examination system.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: An online examination system includes: Intelligent question bank generation module: Based on knowledge points, difficulty level, and question type information, it automatically generates a question bank using intelligent algorithms; at the same time, it dynamically adjusts the difficulty and type of questions in the question bank according to the candidate's historical answering behavior to meet the needs of different candidates; Intelligent proctoring module: Through computer vision technology, it monitors the behavior of candidates in real time, automatically identifies and records suspected cheating behavior; at the same time, it uses natural language processing technology to analyze the candidates' voices in real time to determine whether there is any suspicion of cheating. Intelligent scoring module: Employing deep learning technology, this module intelligently scores test takers' answers. First, it uses natural language processing technology to perform semantic analysis on the answers. Then, it matches the answers with the standard answers and calculates the similarity. Finally, it provides a score based on the similarity score. Intelligent feedback module: Based on the examinee's answers, it uses intelligent algorithms to provide the examinee with personalized error analysis and learning suggestions; at the same time, it provides teachers with detailed exam reports; Intelligent anti-cheating module: It uses facial recognition, fingerprint recognition, and IP address detection technologies to prevent candidates from cheating; at the same time, it uses big data analysis technology to mine exam data and discover potential cheating behaviors. Intelligent scheduling module: Automatically allocates examination resources according to the needs of the examination task; at the same time, dynamically adjusts the resource allocation strategy according to the progress of the examination. Intelligent security module: It uses encryption technology to protect the security of exam data; at the same time, it uses blockchain technology to achieve decentralized storage of data, preventing data from being tampered with or leaked. Intelligent interaction module: Provides an operation interface and supports access from multiple terminal devices; Intelligent optimization module: Automatically optimizes system performance through statistical analysis of exam data.

[0006] Preferably, the intelligent question bank generation module includes: Knowledge Point Analysis Unit: Analyze the distribution of knowledge points in the course, and determine the assessment frequency and weight of each knowledge point based on the teaching syllabus and past exam trends; Difficulty coefficient calculation unit: Using psychometric methods such as item response theory, the unit calculates the difficulty and discrimination of the questions to ensure that the difficulty distribution of the questions in the question bank is reasonable. Diverse question types: The unit is designed with a variety of question types and supports the handling of complex question types such as formula editor and graph questions to adapt to different subjects and exam requirements; Dynamically Adjustable Unit: Based on the test taker's answer data, the question bank is dynamically adjusted to achieve personalized question recommendations.

[0007] Preferably, the intelligent proctoring module includes: Real-time video monitoring unit: Monitors examinee behavior in real time through cameras and uses image processing technology to detect suspicious actions; Speech recognition and analysis unit: Collects the candidate's voice during the answering process, verifies the candidate's identity through voiceprint recognition technology, and analyzes the sound waveform to identify abnormal sound events; Biometric identification unit: Integrates biometric identification technologies such as fingerprint recognition and facial recognition to ensure the accuracy of the candidate's identity; Behavioral Pattern Analysis Unit: Constructs a behavioral model of test takers and analyzes abnormal behavioral patterns using machine learning techniques.

[0008] Preferably, the intelligent scoring module includes: Natural Language Processing Unit: Performs word segmentation and semantic understanding on subjective question answers, and calculates the semantic similarity between the answers and the reference answers; Deep learning scoring network: Construct a deep neural network model and train the scoring model to automatically learn the scoring criteria; Scoring calibration unit: Regularly calibrates the scoring model to ensure the consistency of scoring standards and the stability of scoring results; Performance Statistical Analysis Unit: This unit performs statistical analysis on exam results, including performance distribution and identification of difficult and easy questions, providing decision support for teachers.

[0009] Preferably, the intelligent anti-cheating module includes: Identity verification system: Combining one or more methods such as password, mobile verification code, and electronic signature to strictly verify the identity of candidates logging into the system; Behavioral monitoring unit: Monitors candidates' behavior during the exam, uses machine learning technology to distinguish between normal and abnormal behavior, and provides early warnings of potential cheating risks; Pattern Recognition and Analysis Unit: Utilizes data analysis tools to establish a pattern recognition model of normal test taker behavior, and achieves anomaly detection through clustering and classification algorithms.

[0010] Preferably, the intelligent scheduling module includes: Resource allocation unit: Efficiently allocates the hardware resources required for the exam to ensure the smooth operation of the exam system; Load balancing unit: When multiple users take the exam at the same time, it realizes the load balancing of the server and avoids service bottlenecks and response delays caused by uneven resource allocation.

[0011] Preferably, the intelligent security module includes: Encryption and decryption unit: Employs high-standard encryption algorithms to protect the security of exam questions and candidate information, preventing data from being stolen or leaked during transmission and storage; Access Control Unit: Set up multi-layered firewalls and access control to ensure that only authorized users can access specific exam resources and data; Security Audit Unit: Records and audits the system's operation logs to promptly detect and respond to security incidents.

[0012] Preferably, the intelligent optimization module includes: Performance monitoring and analysis unit: Real-time monitoring of system performance indicators, including response time and system load, and analysis and location of performance bottlenecks; Automatic optimization execution unit: Based on monitoring data, it automatically adjusts system configuration and service deployment to achieve optimal resource allocation and utilization; Prediction and Expansion Unit: Use predictive algorithms to assess future resource needs, expand and allocate resources in advance, and avoid the impact of insufficient resources on the examination.

[0013] Preferably, when the intelligent proctoring module and the behavior pattern analysis unit perform analysis using machine learning technology, the specific steps include the following: S1: Data collection and preprocessing, collecting data from the examination system; cleaning noise and normalizing the collected data to ensure data quality and consistency; S2: Feature extraction, extracting features based on data; S3: Model training. The model is trained using labeled normal behavior data and employs the Support Vector Machine (SVM) algorithm. The specific formula is as follows:

[0014] Where W is the normal vector of the hyperplane, b is the intercept of the hyperplane, and C is the regularization parameter. It is a slack variable; S4: Anomaly scoring and threshold setting. The model generates an anomaly score for each behavior instance. If the score exceeds a certain threshold, it is judged as an abnormal behavior. The threshold setting is based on the trade-off between precision and recall to ensure that the false positive and false negative rates are within an acceptable range. S5: The system analyzes the candidates' behavior in real time and immediately alerts the monitoring personnel when an anomaly is detected.

[0015] Preferably, it also includes: a graphic question processing and scoring module, wherein the graphic question processing and scoring module includes: Graphical Input Unit: Provides a graphical input interface, supports various drawing tools and functions; records the candidate's drawing process, including the drawing order, modification history, and data representation of the final graphic; Image recognition and analysis unit: Standardizes images, including resizing and enhancing clarity; extracts key features from images using image processing techniques, compares the extracted features with standard answers or templates, and calculates the similarity. Deep learning scoring network: Uses convolutional neural networks to train models to identify and score the quality of graphics; scores graphics comprehensively based on multiple dimensions such as accuracy, completeness, and aesthetics. Scoring Calibration and Feedback Unit: Regularly calibrates the scoring model to ensure consistency and accuracy of scoring; provides candidates with specific error analysis and improvement suggestions.

[0016] Preferably, the pattern recognition and analysis unit utilizes data analysis tools to establish a pattern recognition model of normal examinee behavior, and achieves anomaly detection through clustering and classification algorithms, specifically: S1: Based on clustering algorithms such as K-means, candidates are grouped according to behavioral characteristics. The squared error function is then minimized using K-means, specifically with the following formula:

[0017] Where E is the sum of the squared errors between the data points and the cluster centers they are associated with. It's a data point. It is the cluster center; S2: The classification algorithm further analyzes the data points marked as "potential anomalies" by the clustering algorithm. The decision tree divides the data by recursively selecting the best splitting attribute, maximizing the purity of the child nodes.

[0018] The beneficial effects of this invention are as follows: 1. This invention uses machine learning algorithms to effectively identify and prevent cheating, ensuring the fairness of examinations. Furthermore, the application of autoencoders and ensemble learning technologies further improves the accuracy of scoring and the reliability of the system.

[0019] 2. The system of the present invention has the ability to monitor the behavior of candidates in real time and can provide immediate feedback based on the candidates' answers. It also facilitates teachers to track and evaluate students' learning progress in real time.

[0020] 3. The intelligent question bank generation module of this invention can dynamically adjust the question bank based on knowledge point analysis, difficulty coefficient calculation, and diversified question type processing, providing each candidate with a personalized exam experience, which is conducive to targeted training and can help improve the candidate's personal level.

[0021] 4. By setting up a graphic question processing and scoring module, the online examination system can more comprehensively assess candidates' abilities, especially in subjects requiring spatial perception and creativity. Furthermore, the addition of this module will make the examination system more complete and multifunctional, meeting a wider range of educational needs. Attached Figure Description

[0022] Figure 1 This is a flowchart illustrating behavioral pattern analysis in an online examination system proposed in this invention. Detailed Implementation

[0023] The technical solution of the present invention will be further described in detail below with reference to specific embodiments.

[0024] Example 1: An online examination system, comprising: Intelligent question bank generation module: Based on information such as knowledge points, difficulty level, and question type, it automatically generates a question bank using intelligent algorithms; at the same time, it dynamically adjusts the difficulty and type of questions in the question bank according to the candidate's historical answering behavior to meet the needs of different candidates; Intelligent proctoring module: Through computer vision technology, it monitors the behavior of candidates in real time, automatically identifies and records suspected cheating behavior; at the same time, it uses natural language processing technology to analyze the candidates' voices in real time to determine whether there is any suspicion of cheating. Intelligent scoring module: Employing deep learning technology, this module intelligently scores test takers' answers. First, it uses natural language processing technology to perform semantic analysis on the answers. Then, it matches the answers with the standard answers and calculates the similarity. Finally, it provides a score based on the similarity score. Intelligent feedback module: Based on the examinee's answer, it uses intelligent algorithms to provide the examinee with personalized error analysis and learning suggestions; at the same time, it provides teachers with detailed exam reports, including the examinee's score distribution, knowledge mastery, etc. Intelligent anti-cheating module: It adopts a variety of technical means, such as facial recognition, fingerprint recognition, and IP address detection, to prevent candidates from cheating; at the same time, it uses big data analysis technology to mine exam data and discover potential cheating behaviors. Intelligent scheduling module: Automatically allocates examination resources, such as server resources and network bandwidth, according to the needs of the examination task; at the same time, it dynamically adjusts the resource allocation strategy according to the progress of the examination to ensure the smooth conduct of the examination process. Intelligent security module: It uses encryption technology to protect the security of exam data; at the same time, it uses blockchain technology to achieve decentralized storage of data, preventing data from being tampered with or leaked. Intelligent interaction module: Provides an operation interface and supports access from multiple terminal devices; Intelligent optimization module: Through statistical analysis of exam data, it automatically optimizes system performance and improves system stability and reliability.

[0025] The intelligent question bank generation module includes: Knowledge Point Analysis Unit: Analyze the distribution of knowledge points in the course, and determine the assessment frequency and weight of each knowledge point based on the teaching syllabus and past exam trends; Difficulty coefficient calculation unit: Using psychometric methods such as item response theory, the unit calculates the difficulty and discrimination of the questions to ensure that the difficulty distribution of the questions in the question bank is reasonable. Diverse question types: The unit is designed with a variety of question types and supports the handling of complex question types such as formula editor and graph questions to adapt to different subjects and exam requirements; Dynamic Adjustment Unit: Based on the test taker's answer data, the question bank is dynamically adjusted to achieve personalized question recommendations and improve the relevance and effectiveness of the question bank.

[0026] The intelligent proctoring module includes: Real-time video monitoring unit: Monitors examinee behavior in real time through cameras and uses image processing technology to detect suspicious actions; Speech recognition and analysis unit: Collects the candidate's voice during the answering process, verifies the candidate's identity through voiceprint recognition technology, and analyzes the sound waveform to identify abnormal sound events; Biometric identification unit: Integrates biometric identification technologies such as fingerprint recognition and facial recognition to ensure the accuracy of the candidate's identity; Behavioral Pattern Analysis Unit: Constructs a behavioral model of test takers and analyzes abnormal behavioral patterns, such as frequent eye shifts and abnormal keyboard and mouse operations, using machine learning techniques.

[0027] The intelligent scoring module includes: Natural Language Processing Unit: Performs word segmentation and semantic understanding on subjective question answers, and calculates the semantic similarity between the answers and the reference answers; Deep learning scoring network: Construct a deep neural network model, train the scoring model to automatically learn the scoring criteria, and improve the objectivity and accuracy of the scoring; Scoring calibration unit: Regularly calibrates the scoring model to ensure the consistency of scoring standards and the stability of scoring results; The performance statistics and analysis unit performs statistical analysis on exam results, including score distribution and identification of difficult and easy questions, to provide decision support for teachers.

[0028] The intelligent anti-cheating module includes: Identity verification system: Combining passwords, mobile verification codes, electronic signatures, and other methods, strictly verifies the identity of candidates logging into the system; Behavioral monitoring unit: Monitors candidates' behavior during the exam, uses machine learning technology to distinguish between normal and abnormal behavior, and provides early warnings of potential cheating risks; Pattern Recognition and Analysis Unit: Utilizes data analysis tools to establish pattern recognition models of normal test taker behavior, and achieves anomaly detection through algorithms such as clustering and classification.

[0029] The intelligent scheduling module includes: Resource allocation unit: Efficiently allocates the hardware resources required for the exam, such as CPU time, memory, and network bandwidth, to ensure the smooth operation of the exam system; Load balancing unit: When multiple users take the exam at the same time, it realizes the load balancing of the server and avoids service bottlenecks and response delays caused by uneven resource allocation.

[0030] The intelligent security module includes: Encryption and decryption unit: Employs high-standard encryption algorithms to protect the security of exam questions and candidate information, preventing data from being stolen or leaked during transmission and storage; Access Control Unit: Set up multi-layered firewalls and access control to ensure that only authorized users can access specific exam resources and data; Security Audit Unit: Records and audits the system's operation logs, promptly detects and responds to security incidents, and ensures the safe and reliable operation of the system.

[0031] The intelligent optimization module includes: Performance monitoring and analysis unit: Real-time monitoring of system performance indicators, such as response time and system load, to analyze and locate performance bottlenecks; Automatic optimization execution unit: Based on monitoring data, it automatically adjusts system configuration and service deployment to achieve optimal resource allocation and utilization; Prediction and Expansion Unit: Use predictive algorithms to assess future resource needs, expand and allocate resources in advance, and avoid the impact of insufficient resources on the examination.

[0032] The intelligent proctoring module and behavior pattern analysis unit, when performing analysis using machine learning technology, specifically include the following steps: S1: Data collection and preprocessing. Collect data from the examination system, such as behavioral data like candidates' mouse movements, keyboard usage, and eye gaze points; clean and normalize the collected data to ensure its quality and consistency. S2: Feature extraction, extracting features based on data, such as click frequency, page switching speed, and pause duration; S3: Model training. The model is trained using labeled normal behavior data and employs the Support Vector Machine (SVM) algorithm. The specific formula is as follows:

[0033] Where W is the normal vector of the hyperplane, b is the intercept of the hyperplane, and C is the regularization parameter. It is a slack variable; S4: Anomaly scoring and threshold setting. The model generates an anomaly score for each behavior instance. If the score exceeds a certain threshold, it is judged as an abnormal behavior. The threshold setting is based on the trade-off between precision and recall to ensure that the false positive and false negative rates are within an acceptable range. S5: The system analyzes the candidates' behavior in real time and immediately alerts the monitoring personnel when an anomaly is detected.

[0034] Example 2: An online examination system, which is based on Example 1: The system also includes: a graphic question processing and scoring module, which includes: Graphical Input Unit: Provides a graphical input interface, supports various drawing tools and functions, such as line, curve, shape, and color selection; records the candidate's drawing process, including the drawing order, modification history, and data representation of the final graphic; Image recognition and analysis unit: Standardizes images, including size adjustment and sharpness enhancement; extracts key features from images using image processing techniques, such as line thickness, angle size, and image proportion; compares the extracted features with standard answers or templates and calculates the similarity. Deep learning scoring network: Uses convolutional neural networks to train models to identify and score the quality of graphics; scores graphics comprehensively based on multiple dimensions such as accuracy, completeness, and aesthetics. Scoring Calibration and Feedback Unit: Regularly calibrates the scoring model to ensure consistency and accuracy of scoring; provides test takers with specific error analysis and improvement suggestions to help them understand their mistakes and improve their skills.

[0035] Example 3: An online examination system, which is based on Example 1: The pattern recognition and analysis unit utilizes data analysis tools to establish a pattern recognition model of normal examinee behavior, and achieves anomaly detection through clustering and classification algorithms, specifically: S1: Based on clustering algorithms such as K-means, candidates are grouped according to behavioral characteristics. The squared error function is then minimized using K-means, specifically with the following formula:

[0036] Where E is the sum of the squared errors between the data points and the cluster centers they are associated with. It's a data point. It is the cluster center; S2: The classification algorithm further analyzes the data points marked as "potential anomalies" by the clustering algorithm. The decision tree divides the data by recursively selecting the best splitting attribute, maximizing the purity of the child nodes.

[0037] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. An online examination system, characterized in that, include: Intelligent question bank generation module: Automatically generates a question bank based on knowledge points, difficulty level, and question type information using intelligent algorithms; At the same time, the difficulty and type of questions in the question bank are dynamically adjusted based on the candidates' past answer performance to meet the needs of different candidates; Intelligent proctoring module: Through computer vision technology, it monitors the behavior of candidates in real time, automatically identifies and records suspected cheating behavior; at the same time, it uses natural language processing technology to analyze the candidates' voices in real time to determine whether there is any suspicion of cheating. Intelligent scoring module: Employing deep learning technology, it intelligently scores test takers' answers. First, it uses natural language processing technology to perform semantic analysis on the answers; then, it matches the answers with the standard answers and calculates the similarity; finally, it gives a score based on the similarity. Intelligent feedback module: Based on the examinee's answers, it uses intelligent algorithms to provide the examinee with personalized error analysis and learning suggestions; at the same time, it provides teachers with detailed exam reports; Intelligent anti-cheating module: It uses facial recognition, fingerprint recognition, and IP address detection technologies to prevent candidates from cheating; at the same time, it uses big data analysis technology to mine exam data and discover potential cheating behaviors. Intelligent scheduling module: Automatically allocates examination resources according to the needs of the examination task; at the same time, dynamically adjusts the resource allocation strategy according to the progress of the examination. Intelligent security module: It uses encryption technology to protect the security of exam data; at the same time, it uses blockchain technology to achieve decentralized storage of data, preventing data from being tampered with or leaked. Intelligent interaction module: Provides an operation interface and supports access from multiple terminal devices; Intelligent optimization module: Automatically optimizes system performance through statistical analysis of exam data.

2. The online examination system according to claim 1, characterized in that, The intelligent question bank generation module includes: Knowledge Point Analysis Unit: Analyze the distribution of knowledge points in the course, and determine the assessment frequency and weight of each knowledge point based on the teaching syllabus and past exam trends; Difficulty coefficient calculation unit: Using psychometric methods such as item response theory, the unit calculates the difficulty and discrimination of the questions to ensure that the difficulty distribution of the questions in the question bank is reasonable. Diverse question types: The unit is designed with a variety of question types and supports the handling of complex question types such as formula editor and graph questions to adapt to different subjects and exam requirements; Dynamically Adjustable Unit: Based on the test taker's answer data, the question bank is dynamically adjusted to achieve personalized question recommendations.

3. The online examination system according to claim 1, characterized in that, The intelligent proctoring module includes: Real-time video monitoring unit: Monitors examinee behavior in real time through cameras and uses image processing technology to detect suspicious actions; Speech recognition and analysis unit: Collects the candidate's voice during the answering process, verifies the candidate's identity through voiceprint recognition technology, and analyzes the sound waveform to identify abnormal sound events; Biometric identification unit: Integrates biometric identification technologies such as fingerprint recognition and facial recognition to ensure the accuracy of the candidate's identity; Behavioral Pattern Analysis Unit: Constructs a behavioral model of test takers and analyzes abnormal behavioral patterns using machine learning techniques.

4. The online examination system according to claim 1, characterized in that, The intelligent scoring module includes: Natural Language Processing Unit: Performs word segmentation and semantic understanding on subjective question answers, and calculates the semantic similarity between the answers and the reference answers; Deep learning scoring network: Construct a deep neural network model and train the scoring model to automatically learn the scoring criteria; Scoring calibration unit: Regularly calibrates the scoring model to ensure the consistency of scoring standards and the stability of scoring results; Performance Statistical Analysis Unit: This unit performs statistical analysis on exam results, including performance distribution and identification of difficult and easy questions, providing decision support for teachers.

5. An online examination system according to claim 3, characterized in that, The intelligent anti-cheating module includes: Identity verification system: Combining one or more methods such as password, mobile verification code, and electronic signature to strictly verify the identity of candidates logging into the system; Behavioral monitoring unit: Monitors candidates' behavior during the exam, uses machine learning technology to distinguish between normal and abnormal behavior, and provides early warnings of potential cheating risks; Pattern Recognition and Analysis Unit: Utilizes data analysis tools to establish a pattern recognition model of normal test taker behavior, and achieves anomaly detection through clustering and classification algorithms.

6. The online examination system according to claim 1, characterized in that, The intelligent scheduling module includes: Resource allocation unit: Efficiently allocates the hardware resources required for the exam to ensure the smooth operation of the exam system; Load balancing unit: When multiple users take the exam at the same time, it realizes the load balancing of the server and avoids service bottlenecks and response delays caused by uneven resource allocation.

7. The online examination system according to claim 1, characterized in that, The intelligent security module includes: Encryption and decryption unit: Employs high-standard encryption algorithms to protect the security of exam questions and candidate information, preventing data from being stolen or leaked during transmission and storage; Access Control Unit: Set up multi-layered firewalls and access control to ensure that only authorized users can access specific exam resources and data; Security Audit Unit: Records and audits system operation logs to promptly detect and respond to security incidents; The intelligent optimization module includes: Performance monitoring and analysis unit: Real-time monitoring of system performance indicators, including response time and system load, and analysis and location of performance bottlenecks; Automatic optimization execution unit: Based on monitoring data, it automatically adjusts system configuration and service deployment to achieve optimal resource allocation and utilization; Prediction and Expansion Unit: Use predictive algorithms to assess future resource needs, expand and allocate resources in advance, and avoid the impact of insufficient resources on the examination.

8. An online examination system according to claim 5, characterized in that, When the intelligent proctoring module and behavior pattern analysis unit perform analysis using machine learning technology, the specific steps include the following: S1: Data collection and preprocessing, collecting data from the examination system; cleaning noise and normalizing the collected data to ensure data quality and consistency; S2: Feature extraction, extracting features based on data; S3: Model training. The model is trained using labeled normal behavior data and employs the Support Vector Machine (SVM) algorithm. The specific formula is as follows: ; Where W is the normal vector of the hyperplane, b is the intercept of the hyperplane, and C is the regularization parameter. It is a slack variable; S4: Anomaly scoring and threshold setting. The model generates an anomaly score for each behavior instance. If the score exceeds a certain threshold, it is judged as an abnormal behavior. The threshold setting is based on the trade-off between precision and recall to ensure that the false positive and false negative rates are within an acceptable range. S5: The system analyzes the candidates' behavior in real time and immediately alerts the monitoring personnel when an anomaly is detected.

9. An online examination system according to claim 1, characterized in that, Also includes: A graph-based question processing and scoring module, comprising: Graphical Input Unit: Provides a graphical input interface, supports various drawing tools and functions; records the candidate's drawing process, including the drawing order, modification history, and data representation of the final graphic; Image recognition and analysis unit: Standardizes images, including resizing and enhancing clarity; extracts key features from images using image processing techniques, compares the extracted features with standard answers or templates, and calculates the similarity. Deep learning scoring network: Uses convolutional neural networks to train models to identify and score the quality of graphics; scores graphics comprehensively based on multiple dimensions such as accuracy, completeness, and aesthetics. Scoring Calibration and Feedback Unit: Regularly calibrates the scoring model to ensure consistency and accuracy of scoring; provides candidates with specific error analysis and improvement suggestions.

10. An online examination system according to claim 5, characterized in that, The pattern recognition and analysis unit utilizes data analysis tools to establish a pattern recognition model of normal examinee behavior, and achieves anomaly detection through clustering and classification algorithms, specifically: S1: Based on clustering algorithms such as K-means, candidates are grouped according to behavioral characteristics. The squared error function is then minimized using K-means, specifically with the following formula: ; Where E is the sum of the squared errors between the data points and the cluster centers they are associated with. It's a data point. It is the cluster center; S2: The classification algorithm further analyzes the data points marked as "potential anomalies" by the clustering algorithm. The decision tree divides the data by recursively selecting the best splitting attribute, maximizing the purity of the child nodes.

Images