Intelligent question setting exercise method and device based on AI and storage medium

Abstract

The invention relates to the technical field of data processing, in particular to an AI-based intelligent question setting exercise method and device and a storage medium, and the method comprises the steps: collecting target identity data, behavior data and question data; determining a target behavior index according to the target behavior data, and determining a question complexity index according to the question data; determining a target type according to the target behavior index and the topic complexity index; performing intelligent question setting and pushing to the target according to the target historical exercise question and the target type in the monitoring period; and judging a target state based on the question difficulty, the target question answering accuracy and the target question answering code line number in the management period, and updating a process of performing intelligent question setting and pushing to the target in the next management period according to the target state. The exercise efficiency of the user is effectively improved, and the learning efficiency of the user is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, and particularly relates to an AI-based intelligent question generation and practice method and device and storage medium. BACKGROUND

[0002] Traditional programming training systems mainly rely on correct answer rates and question difficulty labels to achieve content recommendation, which has significant limitations: single evaluation dimension cannot capture key behavior characteristics such as debugging strategy optimization and code refactoring during coding; static matching mechanism uses fixed difficulty and user level corresponding rules, which is difficult to dynamically adapt to the stage fluctuations of individual ability growth, and is easy to cause the disconnection between training intensity and real level; the feedback mechanism mainly relies on manual intervention, which is difficult to identify potential knowledge gaps of users in a timely manner due to high response delay; the question complexity evaluation ignores the code structure characteristics, which is essentially different from the actual requirements of maintainability and readability in software engineering practice.

[0003] The above defects make it difficult for traditional methods to achieve precise ability improvement guidance, and it is necessary to introduce fine-grained behavior tracking and dynamic adaptive models to build an intelligent training system. SUMMARY

[0004] The present application aims to provide an AI-based intelligent question generation and practice method and device and storage medium to solve at least one of the problems in the prior art.

[0005] To achieve the above-mentioned purpose, the present application adopts the following technical solutions:

[0006] An AI-based intelligent question generation and practice method, comprising:

[0007] Collecting target identity data, target behavior data and question data;

[0008] Determining a target behavior index according to the target behavior data, and determining a question complexity index according to the question data;

[0009] Determining a target type according to the target behavior index and the question complexity index;

[0010] Intelligently generating and pushing questions to the target according to the target historical practice questions and the target type in the monitoring period;

[0011] Judging the target state based on the question difficulty, the target correct answer rate and the target code line number in the management period, and updating the intelligent question generation and push process to the target in the next management period according to the target state.

[0012] Optionally, the duration of the target single key pause in the analysis period that is greater than or equal to the pause threshold is extracted and taken as the long pause duration, the total duration of the target long pause duration in the analysis period is counted and taken as the pause duration, and the difference between the analysis period duration and the pause duration is taken as the effective duration, denoted as Ts; the number of target repair operations in the analysis period is counted and denoted as FB; the target debugging time in the analysis period is counted and denoted as Tr;

[0013] The target debugging time Tr is compared with the preset debugging time Ty to determine the correction factor;

[0014] The effective duration Ts, the target repair operation number FB, and the correction factor in the analysis period are data fused to determine the target behavior index XW.

[0015] Optionally, the question knowledge point density coefficient D1 is determined according to the number of knowledge points and the knowledge point difficulty coefficient of the question;

[0016] The data of the nested level E1, the condition number E2, and the loop number E3 in the standard answer of the question are fused to determine the answer difficulty coefficient D2;

[0017] The question complexity index DN is constructed based on the question knowledge point density coefficient D1 and the answer difficulty coefficient D2.

[0018] Optionally, when the question complexity index DN is less than or equal to the complexity threshold N0, if the target behavior index XW is less than or equal to the behavior index threshold W0, the target type is determined to be a first type of user, otherwise, the user is determined to be a second type of user;

[0019] When the question complexity index DN is greater than the complexity threshold N0, if the target behavior index XW is less than or equal to {W0×{1-η×ln[4×(DN-N0) / (DN+N0)+1] / ln5}}, the target type is determined to be a first type of user, otherwise, the user is determined to be a second type of user, and η is an adjustment factor.

[0020] Optionally, the knowledge point tags of the wrong questions in the target historical practice questions in the monitoring period are extracted, and the occurrence frequency of each knowledge point tag in the user's historical wrong questions in the monitoring period is counted, the occurrence frequency of the kth knowledge point tag is denoted as Zk, and the knowledge point tags with the top three occurrence frequencies Zk are taken as the practice knowledge points.

[0021] The question complexity index of each question in the question bank is calculated, and the question complexity index of the jth question in the question bank is denoted as DNj; the average question complexity of the target historical practice questions is counted as DNp, and the standard deviation of the question complexity of the target historical practice questions is calculated as σ(DN).

[0022] Optionally, when the target type is a type of users, if DNj∈[DNp-0.2×σ(DN), DNp], the question is determined as a selected question, otherwise, it is determined as a rejected question, if the knowledge point label of the selected question contains more than two exercise knowledge points, the selected question is determined as a recommended question, and the recommended question is taken as a recommended question bank of the next monitoring period, and the recommended question is randomly selected from the question bank and pushed to the user.

[0023] When the target type is a type of users, if DNj∈[DNp+0.5×σ(DN), +∞], the question is determined as a selected question, otherwise, it is determined as a rejected question, if the knowledge point label of the selected question contains more than two exercise knowledge points, the selected question is determined as a recommended question, and the recommended question is taken as a recommended question bank of the next monitoring period, and the recommended question is randomly selected from the question bank and pushed to the user.

[0024] Optionally, when the ratio of the target answer code line number to the question standard answer code line number is less than or equal to 1, the target code conciseness of the question is set to 1, when the ratio of the target answer code line number to the question standard answer code line number is greater than 1 and less than 2, the target code conciseness of the question is set to (2-target answer code line number to the question standard answer code line number), and when the ratio of the target answer code line number to the question standard answer code line number is greater than or equal to 2, the target code conciseness of the question is set to 0.

[0025] The average value of the question complexity of the target answer question in the management period is ND, the average value of the target code conciseness in the management period is JD, and the target state index is constructed according to ND, JD and the target question correct rate ZQ, and the expression of the target state index is:

[0026] ZT=h1×(ND×ZQ)+h2×JD; ZT is the target state index, h1 is the efficiency weight, h2 is the conciseness weight, and h1+h2=1.

[0027] Optionally, when the target state index ZT is less than the state threshold zt, it is determined that the target state is an abnormal state, and the behavior index threshold of the next management period is updated as W0', W0'=W0×{1+γ×exp[3×(zt-ZT)-3]}, γ is a correction factor, to update the process of intelligent question pushing to the target in the next management period, otherwise, the process of intelligent question pushing to the target in the next management period.

[0028] According to another aspect of the present application, an AI-based intelligent question generation exercise device is also provided, comprising:

[0029] A data acquisition unit is configured to acquire target identity data, behavior data and question data.

[0030] a characteristic construction unit configured to determine a target behavior index according to target behavior data and determine a question complexity index according to question data;

[0031] a group division unit configured to determine a target type according to the target behavior index and the question complexity index;

[0032] an intelligent pushing unit configured to intelligently push questions to the target according to the target historical practice questions in the monitoring period and the target type;

[0033] an updating unit configured to judge a target state based on the question difficulty, the target correct answer rate and the target code line number in the management period, and update the process of intelligently pushing questions to the target in the next management period according to the target state.

[0034] According to still another aspect of the present application, a computer readable storage medium is provided, which stores a computer program, wherein the computer program is used to control an electronic device where the computer readable storage medium is located to execute the AI-based intelligent question pushing practice method when running.

[0035] The present application has the following advantages: the present application realizes precise quantitative management and personalized recommendation of the programming learning process through multi-dimensional data fusion and intelligent analysis technology, constructs a dynamic user ability portrait and a question difficulty model based on deep analysis of behavior data and question characteristics, accurately classifies users through a nonlinear classification rule, and automatically generates a hierarchical training question bank in combination with weak knowledge point tracking, breaks through the traditional evaluation mode which only relies on the answer results, comprehensively evaluates technical skills through process behavior analysis, dynamically matches user ability growth through an adaptive threshold adjustment mechanism, avoids recommendation lag, forms a sustainable learning closed loop through long-term state monitoring and feedback optimization, and improves training efficiency and code quality. BRIEF DESCRIPTION OF DRAWINGS

[0036] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0037] Figure 1 The flowchart of the AI-based intelligent question pushing practice method of the present embodiment is shown.

[0038] Figure 2 The flowchart of the target type determination method of the present embodiment is shown.

[0039] Figure 3A flowchart of a push update method for the embodiment.

[0040] Figure 4 A structural schematic diagram of an AI-based intelligent question setting and practicing device of the embodiment. DETAILED DESCRIPTION

[0041] In order to more clearly illustrate the present application, the present application will be further described below in conjunction with preferred embodiments and the accompanying drawings. Like components are denoted by the same reference numerals in the drawings. It should be understood by those skilled in the art that the following specific description is illustrative rather than limiting, and should not limit the scope of protection of the present application.

[0042] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects and do not necessarily describe a specific order or sequence. It should be understood that the data thus used can be interchanged under appropriate circumstances, so that the embodiments of the present application described herein can be implemented. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device including a series of steps or units does not necessarily limit to those steps or units clearly listed, but can include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0043] Specifically, the present embodiment is applied to a programming education platform. Before the course starts, the pre-test is used to determine the starting line of the students, and targeted practice questions are dynamically pushed according to the weaknesses of the students, so as to improve the learning efficiency of the students.

[0044] Referring to Figure 1 The flowchart of the AI-based intelligent question setting and practicing method of the embodiment is shown, which includes:

[0045] In step S101, target identity data, target behavior data and question data are collected. The identity data includes a target ID, the target behavior data includes key pause for writing code, number of repair operations and debugging time consumption, and the question data includes number of knowledge points, difficulty coefficient of knowledge points, nesting level in standard answer of question, number of conditions and number of loops. The nesting level in the standard answer of the question is the maximum depth of AST number in the standard answer of the question. The number of conditions in the standard answer of the question is the total number of branches of if / elif / else in the code of the standard answer of the question. The total number of for, while loops and comprehensions in the standard answer of the question; the target is a user, and the number of repair operations is the number of unnecessary modifications, including variable renaming and space adjustment. The type of unnecessary modification number is not specifically limited in the present embodiment, and can be freely set by those skilled in the art according to the needs.

[0046] Exemplary, in the present embodiment, the target identity data can be collected through interaction, the behavior data is the behavior data of the pre-test questions made by the user, the pre-test questions are test questions used to determine the target ability before the learning stage is started, and the pre-test questions are also included in the questions described in the present embodiment. In the present embodiment, the setting of the pre-test questions is not specifically limited, and those skilled in the art can freely set according to the needs.

[0047] Exemplary, in the present embodiment, the key pause of writing code can be captured in real time for the IDE plug-in, the number of repair operations can be collected through the code difference comparison tool and the syntax tree change detection, the debugging time consumption can be captured through the debugging protocol message, the number of knowledge points and the difficulty coefficient of knowledge points can be collected through expert annotation, the nesting level in the standard answer of the question can be collected through the Python AST module analysis, the condition number can be collected through the regular expression, and the number of loops can be collected through the syntax tree traversal and node type statistics. In the present embodiment, the data collection method is not specifically limited, and those skilled in the art can freely set according to the needs.

[0048] Exemplary, the knowledge point in the question is variable declaration and assignment, the difficulty coefficient of the knowledge point can be set to 0.25, the knowledge point is a basic operator, the difficulty coefficient of the knowledge point can be set to 0.35, the knowledge point is the integrity of the conditional branch, the difficulty coefficient of the knowledge point can be set to 0.60, the knowledge point is the loop boundary control, the difficulty coefficient of the knowledge point can be set to 0.65, the knowledge point is the function parameter passing mechanism, the difficulty coefficient of the knowledge point can be set to 0.8, the knowledge point is the dummy head node of the linked list, the difficulty coefficient of the knowledge point can be set to 0.78, and the knowledge point is the post-order traversal of the binary tree, the difficulty coefficient of the knowledge point can be set to 0.88. In the present embodiment, the setting of the difficulty coefficient of the knowledge point is not specifically limited, and those skilled in the art can freely set according to the needs.

[0049] Please continue to refer to Figure 1 As shown, the AI-based intelligent question generation and practice method further includes:

[0050] In step S102, the target behavior index is determined according to the target behavior data, and the question complexity index is determined according to the question data.

[0051] Specifically, the original data is converted into quantitative indicators, and a bidirectional evaluation system of users and questions is constructed. The behavior index comprehensively describes the smoothness of the user's thinking and the debugging optimization ability by fusing the effective time, the repair behavior and the debugging efficiency; the complexity index quantifies the essential characteristics of the question from the dual perspectives of the knowledge point combination difficulty and the code structure complexity. This data dimensioning method converts the fuzzy behavior mode into a calculable index, lays an interpretable mathematical foundation for subsequent intelligent decision-making, and avoids the deviation of subjective experience judgment.

[0052] Please refer to Figure 2As shown, the target type determination method comprises:

[0053] Step S201, determining a target behavior index according to target behavior data.

[0054] Specifically, the time length of the target single key pause in the analysis period that is greater than or equal to the pause threshold is extracted and taken as the long pause time length, the total time length of the target long pause time length in the analysis period is counted and taken as the pause time length, the difference between the analysis period time length and the pause time length is taken as the effective time length, denoted as Ts; the number of target repair operations in the analysis period is counted, denoted as FB; the target debugging time in the analysis period is counted and denoted as Tr;

[0055] The target debugging time Tr is compared with the preset debugging time Ty to determine a correction factor, in the case of Tr less than or equal to Ty, the correction factor is set to 1, otherwise, the correction factor is set to 1-(Tr-Ty) / Ty, and if the correction factor is less than 0, the value of the correction factor is 0;

[0056] The effective time length Ts, the target repair operation number FB and the correction factor in the analysis period are data fused to determine the target behavior index, and the expression of the target behavior index is:

[0057] XW=w1×(Ts / Tz)+w2×(1-FB / Fmax)+w3×correction factor, wherein w1 is a first weight factor, w2 is a second weight factor, w3 is a third weight factor, w1+w2+w3=1, Tz is the time length of the analysis period, and Fmax is a repair operation number threshold.

[0058] Specifically, by weightedly fusing the key features of the user behavior, a comprehensive index for quantitatively evaluating the real-time ability of the user is constructed. The differentiated design of the weight factors reflects the contribution of different behaviors to the ability evaluation, effectively distinguishes the potential differences of the users, and avoids the one-sidedness of single-dimensional evaluation.

[0059] For example, in the present embodiment, the analysis period is the time length of the target pre-test question, in the present embodiment, the pause threshold can be set to 3 seconds, the preset debugging time can be set to the historical average debugging time of the question, the repair operation number threshold can be set to 8 times, the first weight factor can be set to 0.5, the second weight factor can be set to 0.3, and the third weight factor can be set to 0.2; the setting of the above data is not specifically limited in the present embodiment, and can be freely set by those skilled in the art according to the needs.

[0060] Please continue to refer to Figure 2 As shown, the target type determination method further comprises:

[0061] Step S202, determining a question complexity index according to question data.

[0062] Specifically, the knowledge point density coefficient D1 of the question is determined according to the number of knowledge points in the question and the knowledge point difficulty coefficient of the knowledge points in the question, In the formula, A is the number of knowledge points in the question, and as is the knowledge point difficulty coefficient of the ath knowledge point in the question.

[0063] The answer difficulty coefficient is determined by data fusion according to the nesting level E1, the condition number E2 and the loop number E3 in the standard answer of the question. The expression of the answer difficulty coefficient is as follows:

[0064] D2 = x1 * 1-1 / (E1+1) 0.5 +x2 * (E2 / J1) + x3 * (E3 / J2), wherein x1 is

[0065] the nesting weight, x2 is the condition weight, x3 is the loop weight, x1+x2+x3=1, D2 is the answer difficulty coefficient, J1 is the condition number threshold, and J2 is the loop number threshold.

[0066] The question complexity index DN is constructed based on the question knowledge point density coefficient D1 and the answer difficulty coefficient D2, DN=(D1+D2) / 2.

[0067] Specifically, the knowledge point correlation and the code structure feature are combined, and the comprehensive difficulty of the question is quantified by a mathematical model. The knowledge point density coefficient reflects the complexity of the knowledge system behind the question, and the answer difficulty coefficient captures the challenge of the solution path from the code logic level. The combination of the two makes the question difficulty evaluation cover both explicit knowledge requirements and implicit thinking load, and is closer to the real problem solving experience.

[0068] For example, in the present embodiment, the condition number threshold can be set to 10, the loop number threshold can be set to 8, the nesting weight can be set to 0.55, the condition weight can be set to 0.25, and the loop weight can be set to 0.2. The setting of the above data is not specifically limited in the present embodiment, and can be freely set by those skilled in the art according to the needs.

[0069] Please continue to refer to Figure 1 As shown in FIG. 1, the AI-based intelligent question setting and practicing method further comprises:

[0070] Step S103, determining the target type according to the target behavior index and the question complexity index.

[0071] Specifically, when the question complexity index DN is less than or equal to the complexity threshold N0, if the target behavior index XW is less than or equal to the behavior index threshold W0, it is determined that the target type is a first type of user, otherwise, it is determined that the user is a second type of user.

[0072] When the topic complexity index DN is greater than the complexity threshold N0, if the target behavior index XW is less than or equal to {W0 x {1-η x ln[4 x (DN-N0) / (DN+N0)+1] / ln5}}, the target type is determined to be a first type of user, otherwise, the user is determined to be a second type of user, and η is an adjustment factor.

[0073] For example, in the embodiment, the complexity threshold can be set to 1.5, the behavior index threshold can be set to 0.55, and the adjustment factor can be set to 0.18. In the embodiment, the above settings are not specifically limited, and a person skilled in the art can freely set them according to needs.

[0074] Specifically, the classification mechanism based on the dynamic threshold and the nonlinear discrimination rule realizes fine identification of the user type. Different determination logics are set in different intervals of the topic complexity to fully consider the interaction between the user behavior and the topic difficulty, so as to avoid misjudgment of a single index. The introduction of the adjustment factor gives the system adaptability, which can automatically adjust the classification boundary according to the change of the group ability to adapt to the target requirements in different teaching stages. In essence, the step establishes a dynamic matching relationship model between the user ability and the topic difficulty.

[0075] Please continue to refer to Figure 1 As shown, the AI-based intelligent question generation exercise method further includes:

[0076] In step S104, the target is intelligently pushed with a question according to the target historical exercise questions in the monitoring period and the target type.

[0077] Specifically, the knowledge point tags of the wrong questions in the target historical exercise questions in the monitoring period are extracted, and the occurrence frequencies of the knowledge point tags in the user historical wrong questions in the monitoring period are counted. The occurrence frequency of the kth knowledge point tag is denoted as Zk, and the knowledge point tags with the top three occurrence frequencies Zk are taken as the exercise knowledge points.

[0078] The topic complexity indexes of the questions in the question bank are calculated, and the topic complexity index of the jth question in the question bank is denoted as DNj. The average topic complexity of the target historical exercise questions is DNp, and the standard deviation of the topic complexity of the target historical exercise questions is σ(DN).

[0079] When the target type is a first type of user, if DNj∈[DNp-0.2×σ(DN), DNp], the question is determined to be a selected question, otherwise, it is determined to be a rejected question. If the knowledge point tags of the selected question contain more than two exercise knowledge points, the selected question is taken as a recommended question, and the recommended question is taken as a recommended question bank in the next monitoring period, and the recommended question is randomly selected from the question bank and pushed to the user.

[0080] When the target type is a second type of user, if DNj is in [DNp+0.5xσ(DN), +∞], the question is determined to be a selected question, otherwise, it is rejected as a question. If the knowledge point label of the selected question contains more than two exercise knowledge points, the selected question is recommended as a recommended question, and the recommended question is used as the recommended question bank of the next monitoring period, and the recommended question is randomly selected from the question bank and pushed to the user.

[0081] Specifically, the frequency of each knowledge point label in the user history wrong question is the ratio of the number of wrong questions with the knowledge point label to the total number of wrong questions.

[0082] Specifically, the frequency of the wrong question label is associated with the knowledge point to accurately identify the user's knowledge gap; according to the user type, the question bank screening rule is dynamically adjusted (the first type of user focuses on consolidation, and the second type of user strengthens challenge), ensuring that the exercise question can not only fill the gap, but also not exceed the current ability level. This mechanism solves the contradiction between low efficiency and frustration in the traditional mass exercise mode.

[0083] For example, in this embodiment, the monitoring period can be set to 7 days, and the setting of the monitoring period is not limited in this embodiment, and those skilled in the art can freely set it according to the needs.

[0084] For example, in this embodiment, each question corresponds to a knowledge point label, which is generated by an expert standard, such as the question "Calculate the sum of the first n terms of the Fibonacci sequence, and use the recursive function to implement it, and handle the case n≤0", the knowledge point label is "recursive function design, boundary condition processing, generator expression and nested function definition"; the above setting is not limited in this embodiment, and those skilled in the art can freely set it according to the needs.

[0085] Please continue to refer to Figure 1 As shown in the figure, the AI-based intelligent question generation and exercise method further includes:

[0086] Step S105, judging the target state based on the difficulty of the question in the management period, the target answer rate and the target answer code line number, and updating the next management period to the target according to the target state Intelligent question generation and pushing process.

[0087] Specifically, the joint analysis of code simplicity and accuracy takes into account the efficiency and quality of user answers; the construction of the state index reflects the trend of comprehensive ability growth; the function of abnormal state trigger threshold adjustment makes the system have self-renewal activity, avoiding the "ability stagnation trap" caused by the solidification of the recommendation strategy. This step is the core guarantee to maintain the long-term learning motivation of the user.

[0088] Please refer to Figure 3 As shown in the figure, the push update method includes:

[0089] Step S301, judging the target state based on the question difficulty, the target answer correctness and the target answer code line number in the management period.

[0090] Specifically, when the ratio of the target answer code line number to the question standard answer code line number is less than or equal to 1, the target code conciseness of the question is set to 1; when the ratio of the target answer code line number to the question standard answer code line number is greater than 1 and less than 2, the target code conciseness of the question is set to (2-the ratio of the target answer code line number to the question standard answer code line number); and when the ratio of the target answer code line number to the question standard answer code line number is greater than or equal to 2, the target code conciseness of the question is set to 0.

[0091] The average value of the question complexity of the target answer questions in the management period is ND, the average value of the target code conciseness in the management period is JD, and the target state index is constructed according to ND, JD and the target answer correctness ZQ, and the expression of the target state index is:

[0092] ZT = h1 x (ND x ZQ) + h2 x JD; ZT is the target state index, h1 is the efficiency weight, h2 is the conciseness weight, and h1 + h2 = 1.

[0093] Specifically, the objective evaluation of the long-term performance of the user is realized through the hierarchical quantification of the code conciseness and the dynamic synthesis of the state index. The conciseness index not only reflects the level of coding skills, but also deeply maps the understanding degree of the user to the nature of the problem; the internal law of the ability growth is revealed through the cross-period data aggregation analysis, thereby providing a scientific basis for adjusting the recommendation strategy.

[0094] For example, in the present embodiment, the management period can be set to 30 days, and the setting of the management period is not specifically limited in the present embodiment, and can be freely set by the person skilled in the art according to the needs.

[0095] For example, in the present embodiment, the line number of the user's submitted code and the question standard answer code line number can be counted through an automatic code analysis tool (such as the ast module of Python), the efficiency weight can be set to 0.7, and the conciseness weight can be set to 0.3, and the above settings are not specifically limited in the present embodiment, and can be freely set by the person skilled in the art according to the needs.

[0096] Please continue to refer to Figure 3 As shown in the figure, the push update method comprises:

[0097] Step S302, updating the target according to the target state to intelligently push the questions in the next management period.

[0098] Specifically, when the target state index ZT is less than the state threshold zt, it is determined that the target state is an abnormal state, and the behavior index threshold of the next management period is updated as W0', W0' = W0 x {1 + γ x exp[3 x (zt-ZT)-3]}, γ is a correction factor, to update the process of intelligent question pushing to the target in the next management period, otherwise, the process of intelligent question pushing to the target in the next management period is not performed.

[0099] Specifically, when W0' is greater than or equal to 0.75, the value of W0' is 0.75.

[0100] Specifically, the threshold dynamic adjustment mechanism gives the system active adaptive ability. When the user state is abnormal, the strictness of the recommendation strategy is corrected through the nonlinear adjustment of the exponential function, which avoids overreaction to accidental fluctuations and can correct the negative trend of continuous decline in time. This "intelligent buffer" design balances the timeliness and stability of system intervention, and ensures the smooth transition of the learning path.

[0101] Exemplarily, in the present embodiment, the state threshold can be set to 0.6, and the correction factor can be set to 0.12. In the present embodiment, the above settings are not specifically limited, and a person skilled in the art can freely set them according to the needs.

[0102] Please refer to Figure 4 As shown in the figure, the AI-based intelligent question setting practice device comprises:

[0103] A data acquisition unit is configured to acquire target identity data, behavior data and question data.

[0104] A feature construction unit is configured to determine a target behavior index according to the target behavior data, and determine a question complexity index according to the question data.

[0105] A group division unit is configured to determine a target type according to the target behavior index and the question complexity index.

[0106] An intelligent pushing unit is configured to intelligently push questions to the target according to the target historical practice questions and the target type in the monitoring period.

[0107] An updating unit is configured to determine a target state based on the question difficulty, the target correct answer rate and the target code line number in the management period, and update the process of intelligent question pushing to the target in the next management period according to the target state.

[0108] The AI-based intelligent question setting practice device provided in the embodiments of the present application can execute the AI-based intelligent question setting practice method provided in any embodiment of the present application, and has the corresponding function modules and beneficial effects of the execution method.

[0109] The application also provides a computer readable storage medium, which is a tangible physical storage medium, and can store the computer program and various types of data used in the program; the physical storage medium includes, but is not limited to, a random access memory, a read-only memory, an optical disc, a hard disk, and other existing physical storage media or combinations of the media.

[0110] Those of ordinary skill in the art will understand that all or some of the steps in the above disclosed method and system can be implemented as software, firmware, hardware, and appropriate combinations thereof. Some or all of the physical components can be implemented as software executed by a processor, such as a central processing unit, a digital signal processor, or a microprocessor, or as hardware, or as an integrated circuit, such as an application specific integrated circuit. Such software can be distributed on computer readable media, which can include computer storage media (or non-transitory media) and communication media (or transitory media). As is well known to those of ordinary skill in the art, the term computer storage media includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storage of information such as computer readable programs, data structures, program modules or other data. Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disks (DVD) or other optical disk storage, magnetic cassettes, magnetic tapes, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to store the desired information and that can be accessed by a computer. Furthermore, it is well known to those of ordinary skill in the art that communication media typically embodies computer readable programs, data structures, program modules or other data in a modulated data signal such as a carrier wave or other transport mechanism and includes any information delivery media.

[0111] Obviously, the above embodiments of the present application are only examples for clearly illustrating the present application, and are not intended to limit the embodiments of the present application. Based on the above description, other different forms of changes or variations can be made by those of ordinary skill in the art, and it is impossible to enumerate all the embodiments here. Any obvious changes or variations derived from the technical solutions of the present application are still within the protection scope of the present application.

Claims

1. An AI-based intelligent question-setting practice method, characterized in that: include: Collect target identity data, target behavior data and question data; Determine the target behavior index based on the target behavior data, and determine the question complexity index based on the question data; Determine the target type based on the target behavior index and the question complexity index; Intelligently push questions to the target based on the target's historical practice questions and target type within the monitoring period; The target status is determined based on the difficulty of the questions within the management cycle, the target answer accuracy rate and the target answer code lines, and the process of intelligently pushing questions to the target in the next management cycle is updated according to the target status.

2. The AI-based intelligent question-setting practice method according to claim 1, characterized in that: Extract the duration of the target single key pause greater than or equal to the pause threshold within the analysis period and use it as the long pause duration. Count the total duration of the target long pauses within the analysis period as the pause duration, and take the difference between the analysis period duration and the pause duration as the effective duration, recorded as Ts. The number of target patching operations within the statistical analysis period is recorded as FB; the target debugging time within the statistical analysis period is recorded as Tr; Compare the target debugging time Tr with the preset debugging time Ty to determine the correction factor; The effective time Ts within the analysis period, the target number of repair operations FB and the correction factor are fused to determine the target behavior index XW.

3. The AI-based intelligent question-setting practice method according to claim 2, characterized in that: Determine the knowledge point density coefficient D1 of the question based on the number of knowledge points and the difficulty coefficient of the knowledge points; Data fusion is performed based on the nesting level E1, condition number E2, and loop number E3 in the standard answer to the question to determine the difficulty coefficient D2 of the answer; The question complexity index DN is constructed based on the question knowledge point density coefficient D1 and the solution difficulty coefficient D2.

4. The AI-based intelligent question-setting practice method according to claim 3, characterized in that: When the question complexity index DN is less than or equal to the complexity threshold N0, if the target behavior index XW is less than or equal to the behavior index threshold W0, the target type is determined to be a Class I user; otherwise, the user is determined to be a Class II user; When the question complexity index DN is greater than the complexity threshold N0, if the target behavior index XW is less than or equal to {W0×{1-η×ln[4×(DN-N0) / (DN+N0)+1] / ln5}}, the target type is determined to be a Class I user; otherwise, the user is determined to be a Class II user, and η is the adjustment factor.

5. The AI-based intelligent question-setting practice method according to claim 4, characterized in that: Extract the knowledge point labels of the wrong questions in the target historical practice questions during the monitoring period, and count the occurrence frequency of each knowledge point label in the user's historical wrong questions during the monitoring period. Record the occurrence frequency of the kth knowledge point label as Zk, and select the knowledge point labels with the top three occurrence frequencies Zk as the practice knowledge points; Calculate the question complexity index of each question in the question bank, and record the question complexity index of the jth question in the question bank as DNj; calculate the average question complexity of the target history practice questions as DNp, and calculate the standard deviation of the question complexity of the target history practice questions as σ(DN).

6. The AI-based intelligent question-setting practice method according to claim 5, characterized in that: When the target type is a class of users, if DNj∈[DNp-0.2×σ(DN), DNp], the question is determined to be a candidate question. Otherwise, it is considered a rejected question. If the knowledge point tag of the candidate question contains more than two practice knowledge points, the candidate question is considered a recommended question and the recommended question is used as the recommended question bank for the next monitoring cycle. A recommended question is randomly selected from the question bank and pushed to the user. When the target type is a second-class user, if DNj∈[DNp+0.5×σ(DN),+∞], the question is judged as a candidate question, otherwise, it is treated as a eliminated question. If the knowledge point label of the candidate question contains more than two practice knowledge points, the candidate question is treated as a recommended question, and the recommended question is used as the recommended question bank for the next monitoring cycle. The recommended question is randomly selected from the question bank and pushed to the user.

7. The AI-based intelligent question-setting practice method according to claim 6, characterized in that: When the ratio of the target answer code line number to the target standard answer code line number is less than or equal to 1, the target code simplicity of the question is set to 1. When the ratio of the target answer code line number to the target standard answer code line number is greater than 1 and less than 2, the target code simplicity of the question is set to (2-the ratio of the target answer code line number to the target standard answer code line number). When the ratio of the target answer code line number to the target standard answer code line number is greater than or equal to 2, the target code simplicity of the question is set to 0. The average value of the complexity of the target questions in the statistical management cycle is ND, and the average value of the target code simplicity in the statistical management cycle is JD. The target state index is constructed based on ND, JD and the target answer accuracy ZQ. The expression of the target state index is: ZT=h1×(ND×ZQ)+h2×JD; ZT is the target state index, h1 is the efficiency weight, h2 is the simplicity weight, and h1+h2=1.

8. The AI-based intelligent question-setting practice method according to claim 7, characterized in that: When the target state index ZT is less than the state threshold zt, the target state is determined to be an abnormal state, and the behavior index threshold of the next management cycle is updated to W0', W0'=W0×{1+γ×exp[3×(zt-ZT)-3]}, γ is the correction factor, so as to update the process of pushing intelligent questions to the target in the next management cycle. Otherwise, the process of pushing intelligent questions to the target will not be carried out in the next management cycle.

9. An AI-based intelligent question-setting practice device, applied to the AI-based intelligent question-setting practice method according to any one of claims 1 to 8, characterized in that: include: A data collection unit, used to collect target identity data, behavior data and question data; A feature construction unit, configured to determine a target behavior index based on the target behavior data and a question complexity index based on the question data; The group division unit is used to determine the target type based on the target behavior index and the question complexity index; An intelligent push unit is used to push intelligent questions to the target based on the target's historical practice questions and target type within the monitoring period; The update unit is used to determine the target status based on the difficulty of the questions within the management cycle, the target answer accuracy rate and the target number of answer code lines, and to update the next management cycle according to the target status to push intelligent questions to the target.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, wherein the computer program is used to control the electronic device where the computer-readable storage medium is located to execute the AI-based intelligent question-setting practice method according to any one of claims 1 to 8 during operation.

Images