A student ability evaluation method and system based on an ELO ranking model

Abstract

This invention provides a student ability evaluation method and system based on the ELO rating model, belonging to the field of data analysis technology. The method specifically includes: acquiring and preprocessing student answer behavior data to obtain the student's actual score for each answer; predefining at least two distinct ability dimensions, establishing a dynamic scoring model for each ability dimension to dynamically update the student's score on each ability dimension based on the student's answer behavior data; initializing model parameters and the student's scores on each ability dimension; the dynamic scoring model includes: an ELO rating model and a non-ELO dimension evaluation model; updating the student's score on each ability dimension based on the answer behavior data and actual scores, using the dynamic scoring model corresponding to each ability dimension; and integrating the student's scores across all ability dimensions to generate a visualized ability profile and an ability evaluation report. This invention aims to achieve a more refined, accurate, visible, and comprehensive continuous evaluation of student abilities.

Description

Technical Field

[0001] This invention relates to the field of data analysis technology, and in particular to a student ability evaluation method and system based on the ELO rating model. Background Technology

[0002] With the deepening development of educational informatization and personalized learning concepts, how to scientifically, dynamically, and comprehensively assess students' learning abilities has become one of the core issues in the field of educational technology. Traditional evaluation methods mainly rely on static, summative assessments, such as mid-term / final exams and standardized tests. These methods typically have the following common limitations:

[0003] Evaluation lag: The feedback cycle is long, making it impossible to reflect the fluctuations and progress of students' abilities in the learning process in real time.

[0004] The evaluation criteria are too simplistic: they often use the total score or the score of a single subject as the evaluation standard, which makes it difficult to depict the specific strengths and weaknesses of students in different knowledge areas or ability dimensions.

[0005] Lack of dynamic adaptability: The evaluation model is fixed and cannot be updated and adjusted based on students' continuous learning behavior data, making it difficult to accurately capture the dynamic changes in students' abilities.

[0006] Insufficient visualization and interpretability: Evaluation results are mostly presented in numbers or grades, lacking intuitive and vivid visualizations (such as ability maps and growth curves), which is not conducive to students, teachers and parents intuitively understanding the composition and development trend of abilities.

[0007] In recent years, the rise of learning analytics and educational data mining technologies has offered a possibility to overcome the aforementioned limitations. These technologies aim to achieve more refined and timely evaluation and feedback by collecting, analyzing, and modeling data on students' learning processes. Currently, research and practice in this field largely focus on using machine learning models to predict academic performance, identify learning difficulties, or construct subject knowledge systems through knowledge graphs. However, how to systematically apply a dynamic scoring mechanism that can continuously, stably, and quantitatively reflect the results of the "game" between students and questions to multidimensional ability assessment, and achieve intuitive visualization, remains a direction that urgently needs in-depth exploration in the field of educational evaluation technology.

[0008] Existing programming education assessment methods mostly use project scoring, unit testing, and code correctness checks. While these methods can assess the final output, they are difficult to continuously and dynamically reveal the real-time status and evolution path of students' abilities at different levels, such as "syntax", "algorithmic thinking", "debugging", and "engineering practice". Furthermore, they cannot present these ability values ​​in a unified and comparable quantitative form to drive personalized learning resource recommendations.

[0009] To address the need for dynamic ability assessment, some research has begun to draw inspiration from competitive rating systems. Among these, the Elo rating system (ELO rating model) has attracted attention due to its simplicity, real-time updates, and strong interpretability. Designed by physicist Arpad Elo for ranking chess players, its core principle is to dynamically adjust participants' ratings based on game results (win / draw / loss), quantifying the relative skill gap between players. Currently, a few attempts have been made to apply the ELO rating model to educational assessment scenarios, such as matching the difficulty of a single knowledge point with student ability in adaptive learning systems. These applications typically treat a problem as an "opponent," and the student's correct or incorrect answer as the "game result," updating the student's ability rating on that knowledge point accordingly.

[0010] However, the current application of the ELO rating model in the field of student ability assessment still has the following main limitations:

[0011] Most existing methods use the ELO rating model to assess students' overall abilities in single, general knowledge points or subjects, providing only a comprehensive ability score or a score limited to a single knowledge point. They fail to deconstruct and independently score students' abilities from multiple key and orthogonal dimensions for comprehensive quantitative evaluation. This results in evaluations that "see the forest but not the trees," failing to accurately identify students' specific weaknesses and strengths. Consequently, subsequent personalized learning recommendations lack precise targeting and struggle to provide truly suitable learning resources or pathways.

[0012] Existing methods for applying ELO rating models typically set globally fixed adjustment coefficients for all types of questions or ability dimensions, ignoring the inherent differences in the patterns of ability variation across different dimensions. This uniform approach reduces the model's sensitivity and accuracy across different dimensions.

[0013] Existing technologies primarily output numerical or simple grades, lacking mature solutions for transforming multidimensional dynamic scoring data into integrated, visualized ability maps and long-term progress curves. This makes it difficult for teachers to quickly and intuitively grasp the overall ability distribution of a class or the ability structure of individuals, and students also find it difficult to vividly perceive their own ability composition and growth trajectory, significantly diminishing the teaching feedback value and learning motivation effect of the evaluation results.

[0014] The ELO rating system is essentially designed for "competitive games." However, the evaluation of some dimensions of student ability is not simply a competition between students and problems, but rather an assessment based on process data such as collaborative behavior and code review. Existing simple porting solutions based on the ELO rating system lack a framework for effectively integrating and complementing ELO dynamic scoring with traditional static evaluation and behavioral statistical analysis, resulting in an incomplete evaluation system. The student ability model constructed by these solutions is one-sided and cannot fully reflect the comprehensive practical and collaborative abilities emphasized in modern education (especially STEM education), limiting the breadth and depth of application of the evaluation system in real and complex learning scenarios.

[0015] In summary, existing methods for evaluating student abilities using the ELO rating model suffer from problems such as limited evaluation dimensions, poor model adaptability, unintuitive results, and incomplete system. Summary of the Invention

[0016] To address the shortcomings of the existing technologies, this invention proposes a student ability evaluation method and system based on the ELO rank model to dynamically assess students' abilities, aiming to achieve a more refined, accurate, visible, and comprehensive continuous evaluation of students' abilities.

[0017] On the one hand, this invention proposes a student ability evaluation method based on the ELO rating model, which includes the following process:

[0018] Acquire and preprocess student answer behavior data on the learning platform to obtain the actual score of each student's answer, and store the answer behavior data and the actual score;

[0019] Predefine at least two distinct ability dimensions, build an independent dynamic scoring model for each ability dimension, and initialize the model parameters and student scores on that ability dimension;

[0020] The dynamic scoring model is used to dynamically update the student's score on the corresponding ability dimension based on the student's answer behavior data; the dynamic scoring model includes: ELO level model and non-ELO dimension evaluation model.

[0021] Based on the answer behavior data and actual scores, the student's score on that ability dimension is updated using the dynamic scoring model corresponding to each ability dimension;

[0022] Integrate student scores across all ability dimensions to generate a visual ability profile and ability evaluation report.

[0023] Furthermore, the answer behavior data includes: student answer records and question metadata;

[0024] The student's answer record includes: student ID, question ID, submitted answer, standard answer, submission time, completion time, and process data;

[0025] The question metadata includes: question ID, question text / description, related knowledge point tag, preset difficulty level, and question type.

[0026] Furthermore, the preprocessing method is as follows:

[0027] Based on the preset invalid data conditions, the acquired answer behavior data is cleaned to obtain cleaned answer behavior data;

[0028] For each student's answer record, based on the question metadata, determine whether the question in that student's answer record is a binary scoring question or a continuously scoring question;

[0029] Based on the pre-set scoring rules, determine the student's actual score for this question. ;

[0030] The preset scoring rules are as follows:

[0031] For binary scoring questions, the submitted answer is compared with the standard answer. If they match, the student's actual score is awarded. Otherwise, the student's actual score ;

[0032] For continuously scored questions, the student's actual score is determined by running a preset automatic scoring standard or a scoring rule set by the teacher. ,and .

[0033] Furthermore, the method for establishing an independent dynamic scoring model for each capability dimension is as follows:

[0034] The predefined ability dimensions are divided into a first type of ability dimension and a second type of ability dimension; wherein the first type of ability dimension is a dimension for quantitative scoring based on a student's single answer behavior; and the second type of ability dimension is a dimension for quantitative evaluation based on a student's answer behavior data within a preset time window;

[0035] For the first category of capability dimensions, an ELO level model is established for each capability dimension;

[0036] For the second type of capability dimension, a non-ELO dimension evaluation model is established for each capability dimension.

[0037] Furthermore, the method for constructing the ELO rating model is as follows:

[0038] The model parameters of the ELO rating model include: student's ability dimensions The scoring and difficulty level of the questions and capability dimensions learning rate ;

[0039] The process of updating the score using the ELO rating model includes:

[0040] Based on the answering behavior data, and according to the current question difficulty value in the ELO rating model... Calculate the expected probability that a student will correctly answer the question on the corresponding ability dimension of the ELO rating model.

[0041] The actual score Multiplying the contribution weight of the question by the contribution weight of the question gives the question's contribution weight in the ability dimension. Contribution value ;

[0042] According to ability dimensions learning rate And the expected probability and contribution value obtained. Update students' ability dimensions The current rating on [platform name].

[0043] Furthermore, the non-ELO dimension evaluation model is as follows:

[0044] For any second-category capability dimension, determine several sub-indicators for evaluating that capability dimension;

[0045] Acquire students' answer behavior data within a preset time window, and calculate the score for each sub-indicator according to the preset indicator calculation rules;

[0046] The scores of each sub-indicator are weighted and summed, and then mapped to a preset score range that is consistent with the scoring scale of the ELO rating model to obtain the student's score on the second type of ability dimension.

[0047] The model parameters of the non-ELO dimension evaluation model include: the set of sub-indicators for the corresponding capability dimension of the model, and the weight of each sub-indicator.

[0048] Furthermore, the specific methods for initializing the model parameters and the student's score on this ability dimension are as follows:

[0049] For the ELO level model, based on the knowledge point tag of each question, the corresponding ability dimension of the question is marked, and the contribution weight of the question on the corresponding ability dimension is assigned.

[0050] For any newly registered student, set the student's initial scores in each ability dimension to the default values;

[0051] For any new question, a cold start phase is initiated. Based on the question's preset difficulty level and according to preset assignment rules, the initial question difficulty value is determined, and an initial uncertainty range is set for the initial question difficulty value.

[0052] Calculate the effective difficulty value of the question based on the initial question difficulty value and the initial uncertainty range;

[0053] The effective question difficulty value is used as the current question difficulty value of the question and applied to the scoring update process of the ELO level model. When the accumulated student answer records exceed the preset number threshold, a preset parameter estimation algorithm is used to simultaneously invert the student's initial score on the ability dimension and the actual question difficulty value based on the accumulated student answer records.

[0054] The actual question difficulty value obtained from the inversion is used as the current question difficulty value, and the cold start phase is exited, transitioning to the normal ELO rating model scoring update process;

[0055] Set the learning rate for each capability dimension;

[0056] For non-ELO dimension evaluation models, set the set of sub-indicators for the corresponding capability dimension of the model, as well as the weight of each sub-indicator.

[0057] Furthermore, the specific method for integrating student scores across all ability dimensions to generate a visual ability profile and ability evaluation report is as follows:

[0058] Normalize the student's current scores across all ability dimensions to the same scale and use them as coordinate points for each dimension in the radar chart to create a personal ability radar chart for the student.

[0059] Regularly record each student's scores on each ability dimension, and draw a line graph of the student's ability score-time to obtain the student's progress curve. Then, by aggregating the data of all students in the class, generate a trend chart of the class average ability or a heat map of ability distribution.

[0060] Based on the individual ability radar chart and ability score-time line graph, generate individual learning reports for students according to a preset template; based on the class average ability trend chart or ability distribution heat map, generate class learning analysis reports according to a preset template.

[0061] Based on students' personal learning reports, and using their current scores in each ability dimension as key features, a resource recommendation engine is used to filter out questions whose difficulty level matches the student's current score, and then provide these questions to the student for personalized practice.

[0062] Furthermore, the student ability assessment method also includes a calibration mechanism; the calibration mechanism includes:

[0063] Periodic calibration process: When the periodic calibration triggering condition is met, a number of students are randomly selected, and their actual answer performance on the pre-constructed anchor question bank is obtained; the actual answer performance is: for any anchor question in the anchor question bank, the average of the actual scores of all selected students on that anchor question is taken as the student's actual answer performance on that anchor question;

[0064] Calculate the student's expected performance based on the student's current score on the ability dimension and the difficulty value of the anchor question;

[0065] When the actual answer performance is higher than the expected answer performance, and the difference between the actual answer performance and the expected answer performance exceeds a preset deviation threshold, it is determined that there is a systematic deviation, and the student's score and question difficulty value are calibrated using a preset global adjustment amount.

[0066] Manual intervention process: In response to the correction request issued by the teacher, the current score of the specified student on the specified ability dimension is corrected and recorded for the next score update;

[0067] Boundary control process: After each scoring update process is completed, check whether the updated score exceeds the preset scoring threshold range. If it does, the score is corrected to the scoring threshold range according to the scoring correction formula for the next scoring update; otherwise, no scoring correction is required.

[0068] Check if the current question difficulty value exceeds the preset difficulty threshold range. If it does, adjust the current question difficulty value to the difficulty threshold range according to the difficulty correction formula for the next score update; otherwise, no difficulty correction is needed.

[0069] On the other hand, this invention proposes a student ability evaluation system based on the ELO rating model, which includes:

[0070] The data preprocessing module is used to acquire and preprocess students' answer behavior data on the learning platform to obtain the actual score of each student's answer, and store the answer behavior data and the actual score.

[0071] The scoring model building module is used to predefine at least two distinct ability dimensions, build an independent dynamic scoring model for each ability dimension, and initialize the model parameters and the student's score on that ability dimension.

[0072] The dynamic evaluation module is used to update the student's score on each ability dimension based on the answer behavior data and actual scores, using the dynamic scoring model corresponding to each ability dimension.

[0073] The scoring visualization module is used to integrate student scores across all ability dimensions and generate visual ability profiles and ability evaluation reports.

[0074] The calibration module is used to execute the periodic calibration process when the periodic calibration trigger conditions are met; to execute the manual intervention process when a correction request is received from a teacher; and to execute the boundary control process after each scoring update process is completed.

[0075] The beneficial effects of adopting the above technical solution are as follows:

[0076] This invention introduces and improves the ELO rating model, and combines it with multidimensional ability decomposition and visualization techniques to achieve a more scientific, precise, real-time, and intuitive evaluation of students' abilities. Specific results are as follows:

[0077] This invention introduces a multi-dimensional, independent ELO rating model, decomposing the traditional single comprehensive ability assessment into multiple independent dimensions such as "grammar mastery," "algorithm implementation," "engineering practice," and "debugging." An independent ELO scoring system is established for each dimension, while non-adversarial dimensions such as "collaborative development ability" are evaluated using a fusion approach, with initial ability scores and question difficulty scores assigned separately. As a result, the system can track the dynamic changes of students across different ability dimensions and finely deconstruct their abilities, generating fine-grained ability profiles. This allows teachers to clearly identify each student's strengths and weaknesses, achieving a shift from "fuzzy overall evaluation" to "precise weakness identification," providing accurate data for targeted practice recommendations and instructional guidance.

[0078] This invention differentiates the learning rate parameter K in the ELO rating model based on the characteristics of different ability dimensions. For example, a smaller K value is set for "grammar mastery," which changes more slowly, while a larger K value is set for "algorithm implementation ability," which requires rapid response. This differentiated K value makes the model more closely reflect the growth patterns of different abilities, avoiding evaluation distortion under uniform parameters. Furthermore, for composite dimensions such as "engineering practice ability," this invention innovatively proposes a fusion evaluation mechanism that integrates dynamic ELO scores with static analysis / teacher scores according to weights. This breaks through the limitation of traditional ELO rating models being only applicable to "adversarial scenarios," extending its advantages to a broader field of comprehensive ability evaluation. By designing dimensionally differentiated adjustment coefficients (i.e., the learning rate parameter K) and a fusion evaluation mechanism, the overall sensitivity and validity of the system's scoring are significantly improved, making the scores more accurately reflect students' real-time levels and progress across various dimensions.

[0079] This invention uses dynamic ELO scores across various dimensions as data sources. Through a graphical algorithm, it automatically generates individual ability radar charts (presented as radar charts) and ability progress curves (presented as line graphs). The radar charts visually compare the strengths and weaknesses of abilities in each dimension at the current moment, while the growth curves clearly show the trends of each dimension's abilities over time. By automatically synthesizing multidimensional ELO score data into a visualized ability map and growth curve, students can intuitively perceive their ability structure and growth trajectory, enhancing their sense of purpose and accomplishment in learning. Simultaneously, teachers can quickly browse the overall ability radar chart distribution of the class, facilitating the identification of common weaknesses. This provides an intuitive and efficient decision support tool for adjusting teaching focus and implementing differentiated instruction, enabling evaluation data to truly drive teaching improvement and greatly enhancing the interpretability, intuitiveness, and teaching application value of the evaluation results.

[0080] This invention establishes an extended framework "centered on ELO dynamic scoring, compatible with other quantitative methods," allowing for easy integration of new ability dimensions based on test-taking assessment. Simultaneously, it introduces a three-tiered anti-distortion mechanism: a "data cold start strategy," "regular anchor question calibration," and "manual teacher correction," to combat question difficulty drift and initial / abnormal fluctuations in scoring. This framework ensures the system's scalability and maintainability, adapting to future needs with increased evaluation dimensions. By implementing anti-distortion mechanisms, the long-term reliability of the scoring system is effectively guaranteed, preventing model failure due to data accumulation or individual anomalies, ensuring the continuous reliability of evaluation results, and guaranteeing the system's long-term stability and reliability. This lays a solid foundation for building long-term learning profiles and conducting academic prediction.

[0081] In summary, the method of this invention uses the ELO rating model to evaluate student abilities, dynamically reflecting students' true levels and updating their abilities with each practice session. Furthermore, by dividing the evaluation into multiple dimensions, it generates a radar chart progress curve for each student. Moreover, this method can be flexibly expanded to include any ability assessment based on "problem-solving," and its accuracy increases with the duration of student use. Combining the ability points obtained from the graph with recommendation functions and other modules can provide students with more precise learning services. Attached Figure Description

[0082] Figure 1 This is a flowchart of a student ability evaluation method based on the ELO rating model in this embodiment;

[0083] Figure 2 This is a schematic diagram illustrating the principle of a student ability evaluation method based on the ELO rating model in this embodiment;

[0084] Figure 3 This is a schematic diagram of the personal capability radar chart in this embodiment;

[0085] Figure 4 This is a structural diagram of a student ability evaluation system based on the ELO rating model in this embodiment. Detailed Implementation

[0086] To facilitate understanding of this application, specific embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and embodiments. The following embodiments are illustrative of the invention but are not intended to limit its scope. Rather, these embodiments are provided to provide a more thorough and complete understanding of the disclosure of this application.

[0087] Example 1:

[0088] This embodiment presents a student ability assessment method based on the ELO rating model, such as... Figure 1 and Figure 2 As shown, the method includes the following steps:

[0089] The system acquires and preprocesses students' answering behavior data on the learning platform to obtain the actual score for each answer, and stores the answering behavior data and the actual score.

[0090] In this embodiment, taking a programming learning platform as an example, raw and structured answer behavior data is continuously collected and stored from the learning platform's backend database to provide input for subsequent evaluation.

[0091] The answer behavior data includes: student answer records and question metadata.

[0092] The student's answer record includes: student ID, question ID, submitted answer, standard answer, submission time, completion time, and process data.

[0093] In this embodiment, each student answer record includes the following fields: Student ID, Question ID, Submitted Answer, Standard Answer, Submission Time, Completion Time, and Process Data. It should be noted that the Submitted Answer represents the final result submitted by the student. For objective questions, the Submitted Answer represents the selected option; for programming questions, the Submitted Answer represents the output of the program. The Process Data takes different forms depending on the question type; for example, for programming questions, the Process Data is the code content.

[0094] The question metadata includes: question ID, question text / description, related knowledge point tag, preset difficulty level, and question type.

[0095] In this embodiment, for any question, its corresponding question metadata includes the following fields: question ID, question text / description, related knowledge point tag (such as loop syntax, quick sort), preset difficulty level (such as beginner, intermediate, advanced), and question type (such as multiple choice, fill-in-the-blank, programming, debugging).

[0096] The preprocessing method is as follows:

[0097] Based on preset invalid data conditions, the acquired answer behavior data is cleaned to obtain cleaned answer behavior data.

[0098] In this embodiment, the invalid data conditions include, but are not limited to: duplicate student answer records, student answer records that are submitted with very short submission times and no valid content, and student answer records that are missing key fields, such as student ID, question ID, answer content, etc., which are empty or have incorrect formatting.

[0099] Methods for cleaning invalid data from the acquired answer behavior data include, but are not limited to: if there are duplicate student answer records, only the last student answer record will be retained; if a submitted student answer record is extremely short and contains no valid content, then that student answer record will be considered a test submission and removed.

[0100] For each student's answer record, based on the question metadata, determine whether the question in that student's answer record is a binary scoring question or a continuously scoring question.

[0101] Based on the pre-set scoring rules, determine the student's actual score for this question. .

[0102] The preset scoring rules are as follows:

[0103] For binary scoring questions, the submitted answer is compared with the standard answer. If they match, the student's actual score is awarded. Otherwise, the student's actual score .

[0104] For continuously scored questions, the student's actual score is determined by running a pre-set automatic scoring standard or a scoring rubric developed by the teacher, based on the process data. ,and .

[0105] In this embodiment, each student's answer record is automatically graded, comparing the submitted answer with the standard answer (for programming questions, this is done by running test cases), and generating the actual score. . The rules for determining the value are as follows:

[0106] For objective questions with a binary scoring system, such as multiple choice and fill-in-the-blank questions: A correct answer results in... If you answer incorrectly... .

[0107] For subjective questions or partially graded questions (programming questions, project tasks), which are continuously graded: the grading will be based on the percentage of test cases passed or the grading criteria set by the teacher. Desirable Continuous values ​​within an interval. For example, for any programming problem, running predefined test cases on the process data (i.e., code content) in a student's answer record, if 80% of the test cases pass... .

[0108] Predefine at least two distinct ability dimensions, establish an independent dynamic scoring model for each ability dimension, and initialize the model parameters and student scores on that ability dimension.

[0109] In this embodiment, by defining a multi-dimensional competency system and establishing an evaluation framework, the abstract concept of "competency" is transformed into computable model parameters. Specifically, based on the teaching objectives, a set of interrelated yet relatively independent competency dimensions are predefined. For example, for programming education, the following five core dimensions are defined: syntax mastery, algorithm implementation, engineering practice, debugging, and collaborative development.

[0110] The dynamic scoring model is used to dynamically update the student's score on the corresponding ability dimension based on the student's answer behavior data; the dynamic scoring model includes: ELO level model and non-ELO dimension evaluation model.

[0111] The method for establishing an independent dynamic scoring model for each capability dimension is as follows:

[0112] The predefined ability dimensions are divided into a first type of ability dimension and a second type of ability dimension; wherein the first type of ability dimension is a dimension for quantitative scoring based on a student's single answer behavior; and the second type of ability dimension is a dimension for quantitative evaluation based on a student's answer behavior data within a preset time window.

[0113] For the first type of capability dimension, an ELO level model is established for each capability dimension.

[0114] The method for constructing the ELO rating model is as follows:

[0115] The model parameters of the ELO rating model include: student's ability dimensions The scoring and difficulty level of the questions and capability dimensions learning rate .

[0116] The process of updating the score using the ELO rating model includes:

[0117] Based on the answering behavior data, and according to the current question difficulty value in the ELO rating model... Calculate the expected probability that a student will correctly answer the question on the corresponding ability dimension of the ELO rating model.

[0118] ;

[0119] in, For students in terms of ability dimension The expected probability of correctly answering the question; For students in terms of ability dimension The current score; it should be noted that 400 is a scaling constant used to control the conversion relationship between the score difference and the expected win rate. Its value is an empirical value set by Dr. Elo in the original chess ELO system.

[0120] The actual score Multiplying the contribution weight of the question by the contribution weight of the question gives the question's contribution weight in the ability dimension. Contribution value .

[0121] According to ability dimensions learning rate And the expected probability and contribution value obtained. Update students' ability dimensions The current rating on [platform name].

[0122] ;

[0123] in, This indicates the updated student's ability dimension. The current rating on [platform name].

[0124] In this embodiment, the above process is repeated for each valid answer, so as to achieve continuous and dynamic calibration of students' abilities and question difficulty.

[0125] For the second type of capability dimension, a non-ELO dimension evaluation model is established for each capability dimension.

[0126] The non-ELO dimension evaluation model is as follows:

[0127] For any second-category capability dimension, determine several sub-indicators for evaluating that capability dimension.

[0128] Acquire students' answer behavior data within a preset time window, and calculate the score for each sub-indicator according to the preset indicator calculation rules.

[0129] In this embodiment, the length of the time window is set based on a combination of the standard answer time for the questions, the historical distribution of students' answer times, and the maximum time limit allowed for the exam.

[0130] The scores of each sub-indicator are weighted and summed, and then mapped to a preset score range that is consistent with the scoring scale of the ELO rating model to obtain the student's score on the second type of ability dimension.

[0131] In this embodiment, for the second capability dimension, such as collaborative development capability, data is extracted from version control tool (such as Git) logs, including student ID, number of commits, number of merge requests, number and content of code review comments, etc. Scores for each sub-indicator are calculated according to preset formulas. For example: Git Collaboration Activity = (Number of valid commits per individual / Average number of commits per project) * Standardization coefficient; Code Review Contribution = (Number of valid review comments / Total review opportunities) * Standardization coefficient; The scores of each sub-indicator are weighted and summed to map a score range comparable to the ELO rating model scale (e.g., 0-2500).

[0132] The model parameters of the non-ELO dimension evaluation model include: the set of sub-indicators for the corresponding capability dimension of the model, and the weight of each sub-indicator.

[0133] The specific methods for initializing the model parameters and the student's score on this ability dimension are as follows:

[0134] For the ELO level model, based on the knowledge point tag of each question, the corresponding ability dimension of the question is marked, and the contribution weight of the question on the corresponding ability dimension is assigned.

[0135] In this embodiment, for each question in the question bank, its main ability dimensions and contribution weights are manually or through algorithms (such as tag mapping algorithms based on knowledge point tags). Specifically, this embodiment adopts a hybrid strategy of "rules + machine learning": when a new question is added to the question bank, initial ability dimension weights are first given through rules (such as tag matching) or classification models. For example, a programming question of "using recursion to implement binary tree traversal" can be labeled as: algorithm implementation ability (weight 70%) and syntax mastery ability (weight 30%). This means that the student's performance on this question will simultaneously affect the ELO scores of these two dimensions in this proportion. After accumulating enough answer data, the weights are fine-tuned through backpropagation of the ELO rank model to make the contribution of the question on different ability dimensions more consistent with the actual student performance.

[0136] For any newly registered student, set the initial scores for each ability dimension to the default values.

[0137] For any new question, a cold start phase is initiated. Based on the question's preset difficulty level and according to preset assignment rules, the initial question difficulty value is determined, and an initial uncertainty range is set for the initial question difficulty value.

[0138] Based on the initial question difficulty value and the initial uncertainty range, calculate the effective question difficulty value for the question.

[0139] ;

[0140] in, The effective difficulty value of the question; This is the initial difficulty level of the question; This is an uncertain offset; the initial value can be set to 200. direction (i.e.) or The difficulty level can be determined randomly or based on the initial correct or incorrect answer. This design is equivalent to treating the difficulty level as a range rather than a fixed point, in order to reduce the excessive impact of the initial single answer result on the score.

[0141] In this embodiment, during the initial run when historical answer data is lacking, a phased strategy is used to initialize the question difficulty values ​​and student scores across ability dimensions. Specifically, the initial scores of all newly registered students in each ability dimension are set to default values. For newly added questions, their initial difficulty level is assigned by subject matter experts or teachers based on the knowledge points, complexity, and question type they test, in conjunction with their preset difficulty level. For example, the initial difficulty level for beginner-level questions is [insert initial difficulty value here]. The initial difficulty level corresponding to the intermediate difficulty is [value missing]. The initial difficulty level corresponding to the advanced difficulty is [value missing]. In actual scoring calculations, a relatively large initial uncertainty is set for questions in this "cold start" phase. Specifically, this is achieved by using a dynamic [variable] in the expected win rate calculation. Replace fixed .

[0142] The effective question difficulty value is used as the current question difficulty value of the question and applied to the scoring update process of the ELO level model. When the accumulated student answer records exceed the preset number threshold, a preset parameter estimation algorithm is used to simultaneously invert the student's initial score on the ability dimension and the actual question difficulty value based on the accumulated student answer records.

[0143] The actual difficulty value of the question obtained from the inversion is used as the current difficulty value of the question, and the cold start phase is exited, transitioning to the normal ELO rating model scoring update process.

[0144] In this embodiment, the effective question difficulty value is incorporated into the model and dynamically updated as student answer data accumulates. Once a certain amount of initial answer records have been accumulated (i.e., the cumulative number of student answer records exceeds a preset threshold, such as the first 100 students each completing approximately 20-50 questions), a batch calibration calculation is initiated. Using this batch of data, a preset parameter estimation algorithm, such as maximum likelihood estimation or Bayesian inference, is employed to simultaneously invert the initial scores of all participating students on the ability dimension and the actual question difficulty value. After this calibration is completed, the calibrated actual question difficulty value is used as the official difficulty score for that question, and the uncertainty offset is removed, transitioning to the standard ELO dynamic update process. The above steps are used to correct the initial difficulty setting deviation of the questions and prevent "difficulty drift."

[0145] Set the learning rate for each capability dimension.

[0146] In this embodiment, a baseline is set for each capability dimension. Values, i.e., the learning rate of each capability dimension. The values ​​are usually set smaller to ensure a stable level of difficulty. For example, the grammar mastery ability dimension. Algorithm implementation capability dimension This value determines the update range of the ability score on this dimension.

[0147] For non-ELO dimension evaluation models, set the set of sub-indicators for the corresponding capability dimension of the model, as well as the weight of each sub-indicator.

[0148] Based on the answer behavior data and actual scores, the student's scores on each ability dimension are updated using the dynamic scoring model corresponding to each ability dimension.

[0149] Integrate student scores across all ability dimensions to generate a visual ability profile and ability evaluation report.

[0150] The specific method for integrating student scores across all ability dimensions to generate a visual ability profile and ability evaluation report is as follows:

[0151] Normalize the student's current scores across all ability dimensions to the same scale and use them as coordinate points for each dimension in the radar chart to create a personal ability radar chart for the student.

[0152] In this embodiment, as Figure 3As shown, taking the latest ratings of a student in the 2021 cohort across all dimensions (including ELO and non-ELO dimensions) as an example, this example is based on the five defined core dimensions, further divided into 17 competency dimensions such as "information acquisition and synthesis", "software requirements analysis and design", "teamwork" and "effective communication" for evaluation. All the latest ratings are normalized to the same scale (e.g., 0-100) and used as coordinate points for each dimension, which are then plotted as radar charts to visually display the overall structure and relative strengths of the student's abilities.

[0153] Regularly record each student's scores in each ability dimension and draw a line graph of the student's ability score-time to obtain the student's progress curve. Then, by aggregating the data of all students in the class, generate a trend chart of the class average ability or a heat map of ability distribution.

[0154] In this embodiment, each student's score snapshots across various dimensions are recorded periodically (e.g., weekly), and their ability score-time line graph is plotted to form a progress curve. Furthermore, by aggregating the data of all students in the class, a class average ability trend chart or an ability distribution heatmap can be generated.

[0155] Based on the individual ability radar chart and ability score-time line chart, a student's individual learning report is generated according to a preset template; based on the class average ability trend chart or ability distribution heat map, a class learning analysis report is generated according to a preset template.

[0156] Based on students' personal learning reports, and using their current scores in each ability dimension as key features, a resource recommendation engine is used to filter out questions whose difficulty level matches the student's current score, and then provide these questions to the student for personalized practice.

[0157] In this embodiment, individual learning reports and class learning analysis reports containing the aforementioned visualization charts are generated. Students are categorized by ability dimension. The current rating is used as a key feature and input into the resource recommendation engine. The recommendation engine then selects questions from the question bank that match the difficulty level with the current rating, based on strategies such as "prioritizing improvement of weaknesses" or "continuously deepening strengths." Match questions with a target win rate (i.e., an expected win rate within a certain ideal range, such as 40%-60%) to provide personalized practice recommendations.

[0158] like Figure 2 As shown, based on the generated visualized ability profile and ability evaluation report, it is possible to realize application scenarios such as dynamic adjustment of students' learning paths, early warning of learning difficulties, optimization of teaching resources and strategies, and visualization of learning paths.

[0159] The student ability assessment method also includes a calibration mechanism; the calibration mechanism includes:

[0160] Periodic calibration process: When the periodic calibration trigger condition is met, a number of students are randomly selected, and their actual answer performance on the pre-constructed anchor question bank is obtained; the actual answer performance is: for any anchor question in the anchor question bank, the average of the actual scores of all selected students on that anchor question is taken as the student's actual answer performance on that anchor question.

[0161] In this embodiment, after long-term use, the difficulty level of the questions may change due to changes in the overall level of students, outdated question content, or changes in exposure strategies. A systemic "drift" occurs. To address this, this embodiment establishes a periodic calibration mechanism, specifically as follows: a batch of questions with stable knowledge points, clear expression, and verified quality are selected to form an "anchor question bank." Anchor questions... The value is not subject to regular dynamic updates; its difficulty is determined by experts and remains relatively fixed throughout the system's lifecycle, serving as a calibration benchmark. The calibration trigger condition is set to be executed automatically every quarter or semi-annually. During each calibration execution, a random sample of active students completes the anchor question test. The actual performance (average score rate) of these students on the anchor questions is analyzed.

[0162] Calculate the student's expected performance based on their current score on the competency dimension and the difficulty value of the anchor question.

[0163] When the actual answer performance is higher than the expected answer performance, and the difference between the actual answer performance and the expected answer performance exceeds a preset deviation threshold, a systematic deviation is determined to exist, and a preset global adjustment amount is used to calibrate the student's score in the ability dimension and the difficulty value of the questions.

[0164] In this embodiment, the actual answer performance is compared with the expected answer calculated based on the student's current score on the ability dimension and the fixed difficulty value of the anchor questions. If a systematic bias is found (such as the actual win rate of all anchor questions being significantly higher than expected), it indicates that the current scoring scale may have shifted overall. At this point, a global linear calibration is triggered: a global adjustment is introduced. The scores for all students on the ability dimension and the difficulty values ​​of all questions are proportionally shifted to align the expected performance on the anchor questions back with the actual performance, as shown below:

[0165] ;

[0166] ;

[0167] in, The score after calibration; Scoring before calibration; The difficulty level of the questions after calibration; The difficulty level of the questions before calibration.

[0168] The regular calibration mechanism ensures that the entire scoring system remains consistent with an absolute benchmark over the long term, effectively preventing scoring inflation or deflation.

[0169] Manual intervention process: In response to the correction request issued by the teacher, the current score of the specified student on the specified ability dimension is corrected and recorded for the next score update.

[0170] In this embodiment, teachers are provided with the ability to view each student's scoring history and answer records across all ability dimensions. If a teacher believes, based on sufficient evidence (such as a student's prolonged absence due to illness or system misjudgment), that the current score is severely distorted, they can issue a correction request and initiate a manual correction process. This process corrects the current score to the target ability value or resets it to the most recent reasonable value. This manual intervention is recorded, and the student's current score is updated to the specified value. Subsequently, the model will continue to dynamically update based on the new value.

[0171] Boundary control process: After each scoring update process is completed, check whether the updated score exceeds the preset scoring threshold range. If it does, the score is corrected to the scoring threshold range according to the scoring correction formula for the next scoring update; otherwise, no scoring correction is required.

[0172] Check if the current question difficulty value exceeds the preset difficulty threshold range. If it does, adjust the current question difficulty value to the difficulty threshold range according to the difficulty correction formula for the next score update; otherwise, no difficulty correction is needed.

[0173] In this embodiment, to avoid computational overflow or unrealistic extreme values, reasonable upper and lower limits are set for the current score and question difficulty value. Suggested range: Lower limit of score. Lower limit of the difficulty value of the question Rating cap Maximum difficulty level of the question Immediately after each score update calculation, boundary checks and clipping are performed, i.e.:

[0174] ;

[0175] ;

[0176] in, The revised score; The score before correction; This is the revised difficulty level of the question; The original difficulty level of the questions is shown below; this measure ensures the stability and interpretability of the system's numerical values.

[0177] In this embodiment, to handle extreme cases that the model cannot cover and to give teachers the necessary control, the following manual interfaces and security measures are established: (1) Manual intervention interface: Provide teachers with a management backend, where teachers can view the scoring history and answer records of any student. If the model scoring is considered to be obviously distorted (e.g., a student's score drops sharply due to an accidental mistake), the teacher can manually reset the student's score in a specific dimension to a reasonable value, and the model will continue to run on this basis. (2) Ability score boundary setting: To prevent numerical overflow or extreme cases, reasonable upper and lower limits are set for the score (e.g., [500, 2500]).

[0178] Through the above complete embodiments, those skilled in the art can build and run a complete student ability evaluation system that can automatically acquire data, dynamically and accurately evaluate, intuitively present results, and support personalized learning. This embodiment not only provides the core ELO level model formula, but also gives a full-link implementation plan from the data source to the final application, ensuring the reproducibility and practicality of the technology.

[0179] Example 2:

[0180] This embodiment presents a student ability evaluation system based on the ELO rating model, such as... Figure 4 As shown, the system includes:

[0181] The data preprocessing module is used to acquire and preprocess students' answer behavior data on the learning platform to obtain the actual score of each student's answer, and to store the answer behavior data and the actual score.

[0182] The scoring model building module is used to predefine at least two distinct ability dimensions, build an independent dynamic scoring model for each ability dimension, and initialize the model parameters and the student's score on that ability dimension.

[0183] The dynamic evaluation module is used to update the student's score on each ability dimension based on the answer behavior data and actual scores, using the dynamic scoring model corresponding to each ability dimension.

[0184] The scoring visualization module is used to integrate student scores across all ability dimensions to generate visual ability profiles and ability evaluation reports.

[0185] The calibration module is used to execute the periodic calibration process when the periodic calibration trigger conditions are met; to execute the manual intervention process when a correction request is received from a teacher; and to execute the boundary control process after each scoring update process is completed.

[0186] Example 3:

[0187] This embodiment proposes an electronic device, including: one or more processors, and a memory, wherein the memory is used to store instructions, and when the instructions are executed by the one or more processors, the one or more processors execute the student ability evaluation method based on the ELO level model.

[0188] The electronic device may be a mobile phone, computer, or tablet computer, etc., and includes a memory and a processor. The memory stores a computer program, which, when executed by the processor, implements a student ability evaluation method based on the ELO grading model as described in the embodiment. It is understood that the electronic device may also include input / output (I / O) interfaces and communication components.

[0189] The processor is used to execute all or part of the steps in the student ability evaluation method based on the ELO grading model as described in the above embodiments. The memory is used to store various types of data, which may include, for example, instructions for any application or method in the electronic device, as well as application-related data.

[0190] The processor can be implemented as an Application Specific Integrated Circuit (ASIC), Digital Signal Processor (DSP), Programmable Logic Device (PLD), Field Programmable Gate Array (FPGA), controller, microcontroller, microprocessor, or other electronic components, and is used to execute the student ability evaluation method based on the ELO level model described in the above embodiments.

[0191] Example 4:

[0192] This embodiment proposes a computer-readable storage medium that stores executable instructions. When these instructions are executed, if they are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium.

[0193] The computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the student ability evaluation method based on the ELO rating model described in the various embodiments of this application.

[0194] The aforementioned storage media include: flash memory, hard disks, multimedia cards, card-type memory (e.g., SD (Secure Digital Memory Card) or DX (Memory Data Register, MDR) memory), random access memory (RAM), static random-access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic storage, disks, optical discs, servers, APP (Application) app stores, and other media capable of storing program verification codes. These media store computer programs, which, when executed by a processor, can implement the various steps of the student ability evaluation method based on the ELO level model described above.

[0195] Example 5:

[0196] This embodiment proposes a computer program product, including a computer program or instructions, which, when executed by a processor, implements the aforementioned student ability evaluation method based on the ELO grading model.

[0197] Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or part of the technical solution, can be embodied in the form of a computer program product.

[0198] The various embodiments in this application are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0199] The scope of protection of this application is not limited to the embodiments described above. Obviously, those skilled in the art can make various modifications and variations to this disclosure without departing from the scope and spirit of this disclosure. If such modifications and variations fall within the scope of this disclosure and its equivalents, then the intent of this disclosure also includes these modifications and variations.

Claims

1. A student ability assessment method based on the ELO rating model, characterized in that, This method includes the following steps: Acquire and preprocess student answer behavior data on the learning platform to obtain the actual score of each student's answer, and store the answer behavior data and the actual score; Predefine at least two distinct ability dimensions, build an independent dynamic scoring model for each ability dimension, and initialize the model parameters and student scores on that ability dimension; The dynamic scoring model is used to dynamically update the student's score on the corresponding ability dimension based on the student's answer behavior data; The dynamic scoring model includes: an ELO rank model and a non-ELO dimension evaluation model; Based on the answer behavior data and actual scores, the student's score on that ability dimension is updated using the dynamic scoring model corresponding to each ability dimension; Integrate student scores across all ability dimensions to generate a visual ability profile and ability evaluation report.

2. The student ability evaluation method based on the ELO rating model according to claim 1, characterized in that, The answer behavior data includes: student answer records and question metadata; The student's answer record includes: student ID, question ID, submitted answer, standard answer, submission time, completion time, and process data; The question metadata includes: question ID, question text / description, related knowledge point tag, preset difficulty level, and question type.

3. The student ability evaluation method based on the ELO rating model according to claim 2, characterized in that, The preprocessing method is as follows: Based on the preset invalid data conditions, the acquired answer behavior data is cleaned to obtain cleaned answer behavior data; For each student's answer record, based on the question metadata, determine whether the question in that student's answer record is a binary scoring question or a continuously scoring question; Based on the pre-set scoring rules, determine the student's actual score for this question. ; The preset scoring rules are as follows: For binary scoring questions, the submitted answer is compared with the standard answer. If they match, the student's actual score is awarded. ; Otherwise, the student's actual score ; For continuously scored questions, the student's actual score is determined by running a preset automatic scoring standard or a scoring rule set by the teacher. ,and .

4. The student ability evaluation method based on the ELO rating model according to claim 3, characterized in that, The method for establishing an independent dynamic scoring model for each capability dimension is as follows: The predefined ability dimensions are divided into a first type of ability dimension and a second type of ability dimension; wherein the first type of ability dimension is a dimension for quantitative scoring based on a student's single answer behavior; and the second type of ability dimension is a dimension for quantitative evaluation based on a student's answer behavior data within a preset time window; For the first category of capability dimensions, an ELO level model is established for each capability dimension; For the second type of capability dimension, a non-ELO dimension evaluation model is established for each capability dimension.

5. The student ability evaluation method based on the ELO rating model according to claim 4, characterized in that, The method for constructing the ELO rating model is as follows: The model parameters of the ELO rating model include: student's ability dimensions The scoring and difficulty level of the questions and capability dimensions learning rate ; The process of updating the score using the ELO rating model includes: Based on the answering behavior data, and according to the current question difficulty value in the ELO rating model... Calculate the expected probability that a student will correctly answer the question on the corresponding ability dimension of the ELO rating model. The actual score Multiplying the contribution weight of the question by the contribution weight of the question gives the question's contribution weight in the ability dimension. Contribution value ; According to ability dimensions learning rate And the expected probability and contribution value obtained. Update students' ability dimensions The current rating on [platform name].

6. The student ability evaluation method based on the ELO rating model according to claim 5, characterized in that, The non-ELO dimension evaluation model is as follows: For any second-category capability dimension, determine several sub-indicators for evaluating that capability dimension; Acquire students' answer behavior data within a preset time window, and calculate the score for each sub-indicator according to the preset indicator calculation rules; The scores of each sub-indicator are weighted and summed, and then mapped to a preset score range that is consistent with the scoring scale of the ELO rating model to obtain the student's score on the second type of ability dimension. The model parameters of the non-ELO dimension evaluation model include: the set of sub-indicators for the corresponding capability dimension of the model, and the weight of each sub-indicator.

7. The student ability evaluation method based on the ELO rating model according to claim 6, characterized in that, The specific methods for initializing the model parameters and the student's score on this ability dimension are as follows: For the ELO level model, based on the knowledge point tag of each question, the corresponding ability dimension of the question is marked, and the contribution weight of the question on the corresponding ability dimension is assigned. For any newly registered student, set the student's initial scores in each ability dimension to the default values; For any new question, a cold start phase is initiated. Based on the question's preset difficulty level and according to preset assignment rules, the initial question difficulty value is determined, and an initial uncertainty range is set for the initial question difficulty value. Calculate the effective difficulty value of the question based on the initial question difficulty value and the initial uncertainty range; The effective question difficulty value is used as the current question difficulty value of the question and applied to the scoring update process of the ELO level model. When the accumulated student answer records exceed the preset number threshold, a preset parameter estimation algorithm is used to simultaneously invert the student's initial score on the ability dimension and the actual question difficulty value based on the accumulated student answer records. The actual question difficulty value obtained from the inversion is used as the current question difficulty value, and the cold start phase is exited, transitioning to the normal ELO rating model scoring update process; Set the learning rate for each capability dimension; For non-ELO dimension evaluation models, set the set of sub-indicators for the corresponding capability dimension of the model, as well as the weight of each sub-indicator.

8. The student ability evaluation method based on the ELO rating model according to claim 7, characterized in that, The specific method for integrating student scores across all ability dimensions to generate a visual ability profile and ability evaluation report is as follows: Normalize the student's current scores across all ability dimensions to the same scale and use them as coordinate points for each dimension in the radar chart to create a personal ability radar chart for the student. Regularly record each student's scores on each ability dimension, and draw a line graph of the student's ability score-time to obtain the student's progress curve. Then, by aggregating the data of all students in the class, generate a trend chart of the class average ability or a heat map of ability distribution. Based on the individual ability radar chart and ability score-time line graph, generate the student's individual learning report according to the preset template; Based on the class average ability trend chart or ability distribution heat map, generate a class learning situation analysis report according to the preset template; Based on students' personal learning reports, and using their current scores in each ability dimension as key features, a resource recommendation engine is used to filter out questions whose difficulty level matches the student's current score, and then provide these questions to the student for personalized practice.

9. The student ability evaluation method based on the ELO rating model according to claim 8, characterized in that, The student ability assessment method also includes a calibration mechanism; the calibration mechanism includes: Periodic calibration process: When the periodic calibration triggering condition is met, a number of students are randomly selected, and their actual answer performance on the pre-constructed anchor question bank is obtained; the actual answer performance is: for any anchor question in the anchor question bank, the average of the actual scores of all selected students on that anchor question is taken as the student's actual answer performance on that anchor question; Calculate the student's expected performance based on the student's current score on the ability dimension and the difficulty value of the anchor question; When the actual answer performance is higher than the expected answer performance, and the difference between the actual answer performance and the expected answer performance exceeds a preset deviation threshold, it is determined that there is a systematic deviation, and the student's score and question difficulty value are calibrated using a preset global adjustment amount. Manual intervention process: In response to the correction request issued by the teacher, the current score of the specified student on the specified ability dimension is corrected and recorded for the next score update; Boundary control process: After each scoring update process is completed, check whether the updated score exceeds the preset scoring threshold range. If it does, the score is corrected to the scoring threshold range according to the scoring correction formula for the next scoring update; otherwise, no scoring correction is required. Check if the current question difficulty value exceeds the preset difficulty threshold range. If it does, adjust the current question difficulty value to the difficulty threshold range according to the difficulty correction formula for the next score update; otherwise, no difficulty correction is needed.

10. A student ability evaluation system based on the ELO ranking model, used to implement the student ability evaluation method based on the ELO ranking model as described in any one of claims 1-9, characterized in that, The system includes: The data preprocessing module is used to acquire and preprocess students' answer behavior data on the learning platform to obtain the actual score of each student's answer, and store the answer behavior data and the actual score. The scoring model building module is used to predefine at least two distinct ability dimensions, build an independent dynamic scoring model for each ability dimension, and initialize the model parameters and the student's score on that ability dimension. The dynamic evaluation module is used to update the student's score on each ability dimension based on the answer behavior data and actual scores, using the dynamic scoring model corresponding to each ability dimension. The scoring visualization module is used to integrate student scores across all ability dimensions and generate visual ability profiles and ability evaluation reports. The calibration module is used to execute the periodic calibration process when the periodic calibration trigger conditions are met; to execute the manual intervention process when a correction request is received from a teacher; and to execute the boundary control process after each scoring update process is completed.

Images