Learning early warning and intervention method based on course knowledge graph

Abstract

The invention discloses a learning early warning and intervention method based on a course knowledge graph. The learning early warning and intervention method comprises the following steps: establishing a mode layer of the course knowledge graph, determining a numerical value of a score attribute of each entity in a knowledge level, determining a numerical value of a difficulty attribute of each entity in the knowledge level, and generating a learning reaction matrix; calculating a numerical value of each entity capability value attribute in each student knowledge hierarchy according to the learning reaction matrix; calculating the value of the predicted score attribute of each entity in each student knowledge hierarchy; calculating the value of the predicted total score attribute of the learning object; and according to the updated curriculum knowledge graph, judging a learning early warning level. A learning early warning and dry prediction algorithm and a learning early warning and dry prediction process are defined and given, early warning of two dimensions of knowledge level entities and comprehensive learning conditions is realized, and knowledge level entity resources for effectively improving the academic level and intervention suggestion information of a learning sequence of the knowledge level entity resources are provided.

Description

Technical Field

[0001] The present invention relates to smart education data processing technology, and specifically to a learning early warning and intervention method based on a course knowledge graph. Background Art

[0002] Learning early warning is one of the hot areas of research and application of educational big data. By using scientific methodology to build early warning models, conducting in-depth analysis of data in the teaching process, and timely issuing early warning signals to students and teachers, and providing effective and targeted intervention suggestions, it is an important means to implement teaching students in accordance with their aptitude and improve teaching quality. As a structured form of knowledge representation, knowledge graphs have the technical advantage of interpretability due to their structural characteristics. They can reveal the intricate logical connections between knowledge points, as well as the inclusion and inclusion relationship between learning resources and specific knowledge points. They provide a reference for the systematic modeling of domain knowledge, the effective integration of learning resources, and the design of learning early warning models.

[0003] At present, relevant research results have been achieved at home and abroad. For example, Cheng et al. published the paper "The Impact of Online Review Exercises on Student Course Participation and Learning Performance - A Case Study of the Introductory Financial Accounting Course of an International Joint Venture University" in the Journal of Accounting Education in 2019. The paper analyzed the correlation between online learning and mid-term and final exam scores through students' online learning behavior data, and concluded that students' scores in mid-term and final exams were positively correlated with their efforts in completing online review exercises. Wang Cixiao et al. published the paper "Research on Early Warning of Collaborative Learning Based on Identification of Key Role Interaction Features" in the Journal of Audio-visual Education Research in 2024, which integrated time series and energy level information to construct a two-level (individual and group) role interaction feature calculation method, used a multi-layer linear model to identify the key role interaction features that affect learning outcomes, designed a collaborative learning early warning mechanism, and proposed three intervention strategies with different degrees of coercion: role perception, role suggestion, and role configuration. In the paper "Construction and Application of Human-Computer Collaborative Learning Intervention Model for Marginal Discussion Participation Behavior" published in the journal "Modern Educational Technology" in 2024, Liu Qingtang et al. constructed a human-computer collaborative learning intervention model for marginal discussion participation behavior based on teaching system design theory and learning intervention model. Through teaching practice application, it was shown that the model can effectively correct marginal discussion participation behavior.

[0004] The current learning early warning methods have the following two shortcomings:

[0005] (1) The granularity of warning data for learning warning is generally coarse, and most of them are limited to the two levels of qualified and unqualified.

[0006] (2) Learning warnings lack the necessary explainability, which makes it difficult for students to obtain specific and targeted learning improvement suggestions from the warnings. Summary of the invention

[0007] In order to address the shortcomings of the above-mentioned learning warning methods, the present invention is based on the technical advantages of knowledge graphs in revealing the logical associations and explainability of knowledge. By introducing attributes such as status, score, ability value, completion degree, etc. in the construction of course knowledge graphs and giving corresponding algorithms, a learning warning and intervention method based on course knowledge graphs is provided.

[0008] The present invention adopts the following technical solutions to achieve the above purpose. The learning warning and intervention method based on the course knowledge graph has the following steps:

[0009] Step 1: Establish the model layer of the course knowledge graph and determine the entities, attributes and relationships of the course knowledge graph. The specific steps are as follows:

[0010] S01: Divide the entities of the course knowledge graph into three categories: knowledge level, learning resources and learning objects, wherein the entities, attributes and relationships are determined for the knowledge level and learning resources respectively, and the entities and attributes are determined for the learning objects;

[0011] S02: In the knowledge hierarchy structure, three level entities, namely, assessment target, knowledge cluster and knowledge point, are set;

[0012] S03: Define ID, name, status, score, predicted score, ability value, difficulty and completion value in the attributes of the knowledge level, wherein ID and name record the basic information of the knowledge level entity; the status attribute reflects the activation state of the knowledge level entity, when the status value of the knowledge level entity is equal to 0, it is in an unactivated state, indicating that the knowledge level entity has not been taught; when the status value of the knowledge level entity is equal to 1, it is in an enabled state, indicating that the knowledge level entity has been taught; and the score, predicted score, ability value, difficulty and completion value are all obtained by calculation, wherein the predicted score, ability value and completion value need to be stored separately according to the student ID;

[0013] S04: defining three relationships of isPriorTo, isPartOf and isNecessaryFor in the relationship of the knowledge level, wherein the isPriorTo relationship indicates that the entities of the knowledge level are in a sequential order, the isPartOf relationship indicates that the entities of the knowledge level are in a containment and being contained relationship, and the isNecessaryFor relationship indicates that a certain knowledge level entity is a necessary condition for learning another knowledge level entity;

[0014] S05: setting three types of entities, namely, knowledge, cases and questions, in the learning resource;

[0015] S06: defining an ID, a name and content in the attributes of the learning resource, wherein the basic information of the learning resource is recorded using the ID and the name, and the above content is the content contained in the learning resource;

[0016] S07: defining three relationships, hasKnowledge, hasExample and hasTest, in the relationship of the learning resource, to define the content of the learning resource as knowledge, examples or questions of knowledge level entities;

[0017] S08: defining an ID, a name, a learning conversion rate and a predicted total score in the attributes of the learning object, recording basic information of the learning object using the ID and the name, and obtaining the learning conversion rate and the predicted total score by calculation;

[0018] Step 2: According to the principle of "from high to low, gradually issued", determine the value of the score attribute of each entity in the knowledge level. The specific steps are as follows:

[0019] S01: Determine the assessment score of each assessment objective based on the course syllabus as AOS i , where: i is the assessment target sequence number; the assessment target entity is defined as the central entity CE j , Assessment Target Score AOS i Defined as the central entity score CES j , then there is CES j =AOS i , where: j is the central entity serial number;

[0020] S02: For all the central entities, search for all the central entities CE in the course knowledge graph. j n related entities, and assign scores to all n related entities according to the relationship type, such as the lth related entity RE j,l The distribution score is REDS j,l , and the central entity score is adjusted to CEJS j The total score after issuance needs to be the same as the original central entity score CES j Equal, that is, Where: n is the central entity CE j The total number of related entities, RE j,l CE j The lth related entity, REDS j,l CE j The score of the lth related entity, CEJS j is the adjusted score of the j-th central entity;

[0021] S03: Calculate and update the scores of all relevant entities in S02 based on their scores before and after delivery; that is, the scores of relevant entities RE j,l RES j,l According to the relevant entity issued before the score REBS j,l With the distribution score REDS j,l , according to RES j,l =REBS j,l +REDS j,l Calculate update; where: RES j,l CE j The score of the lth related entity, REBS j,l CE j The score of the lth related entity before it is issued;

[0022] S04: All related entities in S02 are used as the central nodes of the new round; that is, the related entities RE j,l As the central node of the new round of CE j ;

[0023] S05: Repeat steps S02-S04 until all knowledge level entities are traversed and the scores are issued;

[0024] S06: After the distribution is completed, in order to make the course knowledge graph easier to understand, the entities at each knowledge level are sorted and their corresponding scores are represented in a unified symbol format; that is, the knowledge level entity is represented by the symbol E w Indicates that the corresponding knowledge level entity score is represented by the symbol ES w Indicates; where: w is the knowledge level entity serial number;

[0025] Step 3: Determine the value of the difficulty attribute of each entity in the knowledge level. The specific steps are as follows:

[0026] S01: Calculate the true average score probability Among them: TSP mg is the true average score probability, S fa The total score is the sum of the total scores of the previous students’ in-class tests and the total scores of the exams; S fg It is the total score obtained by adding the scores of previous students' in-class tests and examinations.

[0027] S02: Calculate the average model score x=-s;where: MSP mg is the average score rate of the model, x is the lower limit of the ability value, s is the upper limit of the ability value, e is the natural logarithm, MSP x is the lower limit of the model score rate, MSP s is the upper limit of the model score rate;

[0028] S03: Calculate the overall average capability value Where: Y am is the overall average capability value;

[0029] S04: Calculate the coefficient of the capability value to be determined Among them: q is the undetermined coefficient of capability value;

[0030] S05: Calculate the real average score rate of the knowledge level entity; that is, calculate the knowledge level entity E w The actual average scoring rate Among them: TSP mgw is the knowledge level entity E w The true average score rate, ES faw The knowledge level entity E in the past students' class tests and exams w The total score of the superposition, ES fgw The knowledge level entity E in the past students' class tests and exams w The total score of

[0031] S06: Calculate the average model score rate of the knowledge level entity; that is, calculate the knowledge level entity E w The average score rate of the model MSP mgw =TSP mgw (MSP s -MSP x )+MSP x ;

[0032] S07: Calculate the average capability value of the knowledge level entity; that is, calculate the average capability value of the knowledge level entity E w The average ability value

[0033] S08: Calculate the difficulty of the knowledge level entity; that is, calculate the difficulty of the knowledge level entity E w Difficulty

[0034] Step 4: Generate a learning response matrix for input into the early warning model for analysis. The specific steps are as follows:

[0035] S01: The type of learning project is represented by G. When the input data is class, online learning or homework, G = 0; when it is a class test or exam, G = 1;

[0036] S02: Calculate the completion value of each knowledge level entity of the learning object's class, online learning, homework, in-class test and examination items; that is: the knowledge level entity E of the kth student w EX k,w =Lh , online learning completion value Job Completion Value Completion value of class test Exam completion value Among them: aw , W aw , Q aw or F aw They are respectively about the knowledge level entity E w Total online learning time, homework score, class test score or exam score; k,gw , W k,gw , Q k,gw or F k,gw are the kth student's knowledge level entity E w Online learning time, homework scores, class test scores or exam scores; L h , O h , W h , Q h or F h The completion value weight coefficients of class, online learning, homework, test or examination are defined by the teacher according to the characteristics of the course, reflecting the contribution of each project to the completion value; L h , O h and W h Need to satisfy the relationship L h +O h +W h =1, that is, when the learning conversion rate of the kth student is α k =1, for knowledge level entity E w After completing the complete classroom and online learning and correctly completing the homework, the knowledge level entity E w Ability value EY k,w Will reach the average capability value EY mw level;

[0037] S03: When the item type G = 0, generate a learning response matrix:

[0038]

[0039] Where: k is the student serial number;

[0040] S04: When the project type G=1, generate a learning response matrix:

[0041]

[0042] Among them: QFES aw The knowledge level entity E in this class test or exam w Total score of QFES k,gwis the kth student knowledge level entity E in the class test or exam w The score value of

[0043] Step 5: According to the learning response matrix, the state values ​​of all knowledge level entities associated in the matrix are set to 1, that is, set to the enabled state of the taught course;

[0044] Step 6: Based on the learning response matrix, calculate the value of the capability attribute of each entity in the knowledge level of the learning object; that is, the kth student's ability attribute of each entity in the knowledge level E w The ability value is based on Calculation; where: EY k,w is the knowledge level entity E w The ability value of α is calculated for the first time. k =1;

[0045] Step 7: Calculate the predicted score attribute value of each student in the knowledge level; that is, calculate the knowledge level entity E of the kth student w The predicted score ES k,pw , where: ES k,pw is the knowledge level entity E of the kth student w The prediction score of ; the specific steps are as follows:

[0046] S01: The kth student knowledge level entity E w Model score rate MSP k,gw Using the Rasch model Calculate the ability value EY k,w The value range of MSP is determined by the teacher according to the course as [x,s); k,gw is the kth student knowledge level entity E w Model score rate;

[0047] S02: Due to the k-th student knowledge level entity E w The calculation model score rate interval [MSP x ,MSP s ) is not in the interval [0,1] that matches the truth, so we calculate the true score rate:

[0048] S03: Calculate the knowledge level entity E of the kth student w The predicted score ES k,pw =ES w ·TSP k,gw ; Among them: ES k,pw is the kth student knowledge level entity E w The prediction score of

[0049] Step 8: Calculate the value of the predicted total score attribute of the learning object; that is, the predicted total score of the kth student is calculated according to S k,ap =ΣES k,pw Calculate; where: S k,ap is the predicted total score of the kth student;

[0050] Step 9: When the project type G = 0, that is, the learning project is a class, online learning or homework, skip this step; when the project type G = 1, that is, the learning project is a class test or exam, execute this step and calculate the learning conversion rate of each student; that is, when the learning project is a class test or exam, calculate the learning conversion rate of the kth student as α k , the specific steps are as follows:

[0051] S01: The learning conversion rate of the kth student is α k , learning conversion rate α k The initial value of is 1, and the learning conversion rate of the yth in-class test or exam closest to the current one is calculated as α k,y ,calculate Where: S k,ary is the actual total score of the kth student in the yth class test or examination closest to the current one, E wy is the knowledge level entity E in the yth in-class test or exam closest to the current one w The assessment score of

[0052] S02: Calculate the update learning conversion rate Where: n means that the most recent n in-class tests or exams are included in the calculation of student conversion rate; QF h,y is the weight coefficient of the yth in-class test or examination closest to the current one, which is determined by the teacher based on the characteristics of the course; QF α,h QF α,1 and QF α,2 The two weight coefficients of learning conversion rate are class test and examination, which are determined by teachers according to the characteristics of the course. k The rationality of QF h,y and QF α,h Needs to satisfy relationship

[0053] Step 10: Based on the course knowledge graph and learning conversion rate updated in steps 4 to 9, calculate the real score rate of the knowledge level entities with a status value equal to 1 in the course knowledge graph of the learning object, and perform early warnings on the knowledge level entities respectively; that is, the bth knowledge level entity with a status value equal to 1 of the kth student is E G＝1,b ; Where: b is the serial number in the knowledge level entity sequence with a state value of 1; knowledge level entity E G＝1,b The true score rate is TSPk,gb When I ≤ TSP k,gb the warning level of the knowledge level entity E G＝1,b of the k-th student is "normal"; when Q gc ≤ TSP k,gb <Q gn the warning level of the knowledge level entity E G＝1,b of the k-th student is "prompt"; when V ≤ TSP k,gb <I, the warning level of the knowledge level entity E G＝1,b of the k-th student is "warning"; when TSP k,gb <V, the warning level of the knowledge level entity E G＝1,b of the k-th student is "urgent"; where: I, V, Z are the prompt threshold, warning threshold, and urgent threshold respectively, determined by the teacher according to the actual situation of the course;

[0054] Step Eleven: Calculate the learning warning level of the learning object based on the course knowledge graph and learning conversion rate updated in Steps Four to Nine; that is: the score of the knowledge level entity E G＝1,b with a status value of 1 in the course knowledge graph of the k-th student is ES G＝1,b , the predicted score is ES G＝1,k,pb , the total predicted score ES G＝1,k,p of the knowledge level entities with a status value of 1 in the course knowledge graph of the k-th student = ∑ES G＝1,k,pb and the cumulative total score ES G＝1 = ∑ES G＝1,b ; the total predicted score rate of the calculation model of the k-th student When ES G＝1 <H or I ≤ ESP G＝1,k,a,p then the learning warning level of the k-th student is "normal"; when H ≤ ES G＝1 and V ≤ ESP G＝1,k,a,p <I, the learning warning level of the k-th student is "prompt"; when H ≤ ES G＝1 and Z ≤ ESP G＝1,k,a,p <V then the learning warning level of the k-th student is "warning"; when H ≤ ES G＝1 and ESP G＝1,k,a,p <Z, the learning warning level of the k-th student is "urgent"; where: H is the warning start threshold, determined by the teacher according to the actual situation of the course, that is, when the cumulative total score ES G＝1 ≥H of the knowledge level entities with a status value of 1, it is the initial state at the end of the course learning; when the amount of teaching knowledge is sufficient to reflect the student warning status, the learning warning of the student is started;

[0055] Step 12: Read and determine the learning warning level, and assign online learning content and homework questions to students through the learning plan and warning system based on the learning warning level; the specific steps are as follows:

[0056] S01: Reading the learning warning level of the learning object;

[0057] S02: Determine the warning level of the learning object. If the warning level is normal or prompt, end this step and arrange online learning content and homework according to the teaching plan. If the warning level is warning or urgent, proceed to S03.

[0058] S03: Traverse all students whose warning levels are warning and emergency, and calculate the intervention improvement scores of each knowledge level entity whose status value is equal to 1; that is: the knowledge level entity E of the kth student G＝1,b The completion value is EX k,b , the intervention completion value is EX k,v,b =EX k,b +v, v is the intervention change value, which is determined by the teacher according to the actual course; according to the knowledge level entity E G＝1,b Completion value EX k,b and intervention completion value EX k,v,b Compute knowledge level entity E G＝1,b The predicted score ES k,pb and intervention prediction score ES k,v,pb , and then calculate the k-th student knowledge level entity E G＝1,b The intervention improved the ES score k,vx,pb =ES k,v,pb -EX k,pb ;

[0059] S04: Sort the knowledge level entities of the learning object based on the intervention effect; that is, the knowledge level entities of the student whose status value is equal to 1 are sorted according to the entity type order of "knowledge point", "knowledge cluster" and "assessment target", and then the knowledge level entities are further sorted in the entity types of "knowledge point", "knowledge cluster" and "assessment target" from high to low according to the intervention improvement score, so as to obtain the knowledge level entity sorting of the kth student based on the intervention effect, among which the knowledge level entity ranked q is used E G＝1,q express;

[0060] S05: Calculate the score required for the warning level of the learning object to be reduced to a prompt; that is, calculate the score ES required for the warning level of the k-th student to be reduced to a prompt k,vx,r =ES G＝1 ·(I-ESP G＝1,k,a,p );

[0061] S06: Determine the online learning content or homework of the learning object; sort the knowledge level entities based on the intervention effect of the student, and calculate the model intervention improvement score ES k,vx,pb The superposition reaches the warning level and is reduced to the required score ES k,vx,r The minimum number of knowledge-level entities required is z, and the first z knowledge-level entities are the knowledge-level entities that need to be mastered and the warning levels are reduced to prompts. According to the knowledge-level entities, corresponding learning resources of knowledge, cases and questions are given, and intervention content is given and distributed to students.

[0062] Furthermore, the learning objects are all individual students.

[0063] Furthermore, the intervention content is used as online learning content or homework questions.

[0064] The present invention can use the learning behavior data of students in class, online learning, in-class tests, exams, etc. to warn each student of each knowledge level entity and the overall learning status, and conduct learning intervention for students whose learning warning levels are "warning" and "urgent", provide warning levels to reduce to prompts to master knowledge level entities and their knowledge, cases, questions and other learning resources, and sort the knowledge level entities according to the difficulty of improving grades. Define and give learning warning and intervention algorithms and processes, realize warnings in two dimensions of each knowledge level entity and comprehensive learning status, and provide intervention suggestion information on knowledge level entity resources and their learning sequence that effectively improve academic level. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] Figure 1 It is an application process diagram of the knowledge graph-based learning early warning model of the present invention;

[0066] Figure 2 It is a learning early warning analysis flow chart of the present invention;

[0067] Figure 3 It is a flow chart of the learning intervention analysis of the present invention. DETAILED DESCRIPTION

[0068] The present invention is further described below with reference to the accompanying drawings and embodiments. Figures 1 to 3 , the learning warning and intervention method based on the course knowledge graph has the following steps:

[0069] Step 1: Establish the model layer of the "Industrial Internet Technology" course knowledge graph and determine the entities, attributes and relationships of the course knowledge graph. The specific steps are as follows:

[0070] S01: Divide the entities of the course knowledge graph into three categories: knowledge hierarchy, learning resources and learning objects, wherein the knowledge hierarchy and learning resources respectively determine entities, attributes and relationships, and the learning objects determine entities and attributes;

[0071] S02: In the knowledge hierarchy structure, the knowledge clusters are set to three levels, namely, knowledge cluster I, knowledge cluster II and knowledge cluster III. The knowledge hierarchy structure is formed from top to bottom: 7 assessment targets, such as platform foundation, project analysis and other entities; 20 knowledge cluster I, such as system architecture, basic technology and other entities; 35 knowledge cluster II, such as cloud computing, artificial intelligence and other entities; 30 knowledge cluster III, such as the application form of modeling and simulation, management and analysis of edge data and other entities; and 249 knowledge points, such as the concept of modeling and simulation, stream computing and other entities;

[0072] S03: Define ID, name, status, score, predicted score, ability value, difficulty and completion value in the attributes of the knowledge level, wherein ID and name record the basic information of each entity of the knowledge level; the status attribute reflects the activation status of the knowledge level entity, when the status value of the knowledge level entity is equal to 0, it is in the unactivated state, indicating that the knowledge level entity has not yet been taught; when the status value of the knowledge level entity is equal to 1, it is in the activated state, indicating that the knowledge level entity has been taught; the score, predicted score, ability value, difficulty and completion value are obtained by calculation, wherein the predicted score, ability value and completion value need to be stored separately according to the student's student number.

[0073] S04: In the relationship of the knowledge level, 133 isPriorTo relationships, 333 isPartOf relationships, and 53 isNecessaryFor relationships are formed. The isPriorTo relationship indicates that there is a sequence of precedence between the knowledge level entities; the isPartOf relationship indicates that there is a containment and being contained relationship between the knowledge level entities; the isNecessaryFor relationship indicates that an entity of a certain knowledge level is a necessary condition for learning an entity of another knowledge level.

[0074] S05: 297 knowledge entities, 67 case entities, and 77 topic entities are formed in the learning resources;

[0075] S06: defining an ID, a name and content in the attributes of the learning resource, wherein the basic information of the learning resource is recorded using the ID and the name, and the content is the learning resource content contained in the learning resource;

[0076] S07: defining three relationships, hasKnowledge, hasExample and hasTest, in the relationship of the learning resource, to define whether the content of the learning resource is knowledge, examples or questions of a knowledge level entity;

[0077] S08: Define ID, name, learning conversion rate and predicted total score in the attributes of the learning object (i.e., each student). The basic information of the learning object is recorded using the ID and name. The learning conversion rate and predicted total score of the learning object are obtained by calculation.

[0078] Step 2: According to the principle of "from high to low, gradually issued", determine the value of the score attribute of each entity in the knowledge level. The specific steps are as follows:

[0079] S01: Determine the assessment scores of the 7 assessment targets according to the course outline of Industrial Internet Technology, and define each assessment target entity as a central entity, and define the score as the central entity score; here, taking the i-th assessment target as an example, the assessment score of the assessment target is AOS i , the assessment target entity is defined as the central entity CE j , the corresponding score is CES j ; then there is CES j =AOS i ; Where: j is the central entity serial number;

[0080] S02: For all the central entities, search for all entities related to them in the course knowledge graph, assign scores to all related entities according to the relationship type, and adjust the central entity scores at the same time. The total score after assignment needs to be equal to the original central entity score; here, the i-th central entity CE j For example, find all the entities related to the central entity CE in the course knowledge graph. j n related entities, and assign scores to all n related entities according to the relationship type, such as the lth related entity RE j,l The distribution score is REDS j,l , and the central entity score is adjusted to CEJS j The total score after issuance needs to be the same as the original central entity score CES j Equal, that is, Where: n is the central entity CE j Total number of related entities;

[0081] S03: Calculate and update the scores of all related entities described in S02 based on the scores before and after the related entities are sent; here, the lth related entity RE j,l For example, the related entity REj,l RES j,l According to the relevant entity issued before the score REBS j,l With the distribution score REDS j,l , according to RES j,l =REBS j,l +REDS j,l Compute updates;

[0082] S04: All related entities described in S02 are used as the central nodes of the new round; here, the lth related entity RE j,l For example, the related entity RE j,l As the central node of the new round of CE j ;

[0083] S05: Repeat steps S02-S04 until all knowledge level entities are traversed and the scores are issued;

[0084] S06: After the distribution is completed, in order to make the course knowledge graph easier to understand, the entities at each knowledge level are sorted, and each knowledge level entity and its corresponding score are represented in a unified symbol format; taking the wth knowledge level entity as an example, the symbol E is used. w Indicates that the corresponding score is represented by the symbol ES w Indicates; the scores of some entities (see Table 1):

[0085] Table 1 Some entity scores

[0086]

[0087] Step 3: Determine the value of the difficulty attribute of each entity in the knowledge level. The specific steps are as follows:

[0088] S01: Calculate the true average scoring rate Among them: TSP mg is the true average score probability, S fa The total score is the sum of the total scores of the previous students’ in-class tests and the total scores of the exams; S fg It is the total score obtained by adding the scores of previous students' in-class tests and examinations.

[0089] S02: Calculate the model average score rate MSP mg =TSP mg (MSP s -MSP x )+MSP x , x=-s,where: MSP mg is the average score rate of the model, x is the lower limit of the ability value, s is the upper limit of the ability value, MSP xis the lower limit of the model score rate, MSP s is the upper limit of the model score rate, and e is the natural logarithm;

[0090] S03: Calculate the overall average capability value Where: Y am is the overall average capability value;

[0091] S04: Calculate the coefficient of the capability value to be determined Among them: q is the undetermined coefficient of capability value;

[0092] S05: Calculate the real average score rate of the knowledge level entity; here, the wth knowledge level entity E w Take the knowledge level entity E as an example. w The actual average scoring rate Among them: TSP mgw is the knowledge level entity E w The true average score rate, ES faw The knowledge level entity E in the past students' class tests and exams w The total score of the superposition, ES fgw The knowledge level entity E in the past students' class tests and exams w The total score of

[0093] S06: Calculate the model average score rate of the knowledge level entity; here, the wth knowledge level entity E w Take the knowledge level entity E as an example. w The average score rate of the model MSP mgw =TSP mgw (MSP s -MSP x )+MSP x ;

[0094] S07: Calculate the average capability value of the knowledge level entities; here, the wth knowledge level entity E w Take the knowledge level entity E as an example. w The average ability value

[0095] S08: Calculate the difficulty of the knowledge level entity; here, the wth knowledge level entity E w Take the knowledge level entity K as an example. i Difficulty Difficulty of some entities (see Table 2):

[0096] Table 2 Difficulty of some entities

[0097]

[0098] Step 4: Generate a learning response matrix for input into the early warning system for analysis (e.g. Figure 1 As shown in FIG. 1 , the learning behavior data of the classes 102, the in-class tests 105 and the examinations 106 conducted according to the teaching plan 101, as well as the learning behavior data of the online learning 103 and the homework 104 arranged by the teaching plan 101 and the early warning system 107, will be input into the early warning system in the form of a learning response matrix; the specific steps are as follows:

[0099] S01: The type of learning project is represented by G. When the input data is class 102, online learning 103 or homework 104, G=0; when it is a class test 105 or an exam 106, G=1;

[0100] S02: Calculate the completion value of each knowledge level entity of the learning object's class, online learning, homework, in-class test and examination items; here, the wth knowledge level entity E of the kth student is taken as w For example, the knowledge level entity E of the kth student w Completion value of class 102 EX k,w =L h , the knowledge level entity E of the kth student w Completion value of online learning 103 The knowledge level entity E of the kth student w The value of homework 104 The knowledge level entity E of the kth student w Completion value of class test 105 The knowledge level entity E of the kth student w Completion value of Exam 106 Among them: aw , W aw , Q aw or F aw They are respectively about the knowledge level entity E w Total time of online learning 103, total score of homework 104, total score of in-class test 105, total score of examination 106; k,gw , W k,gw , Q k,gw or F k,gw are the kth student's knowledge level entity E w Online learning 103 hours, homework 104 score, class test 105 score, exam 106 score; L h , O h , W h , Q h or F h are the completion value weight coefficients of class 102, online learning 103, homework 104, in-class test 105, and examination 106; L h, O h and W h Need to satisfy the relationship L h +O h +W h =1, which means that the learning conversion rate is α k =1 for students at knowledge level E w After completing the course 102 and online learning 103, and completing the homework 104 correctly, the knowledge level entity E w After the knowledge level entity E w Ability value EY k,w Will reach the average capability value EY mw level;

[0101] S03: When the project type G=0, a learning response matrix is ​​generated. The learning response matrix generated at a certain time is:

[0102]

[0103] Among them: the first column in the learning response matrix is ​​the student's student number, the second column is the knowledge level entity, and the third column is the completion value of the knowledge level entity corresponding to the student in this row;

[0104] S04: When the project type G=1, a learning response matrix is ​​generated. The learning response matrix generated at a certain time is:

[0105]

[0106] Among them: the first column of the learning response matrix is ​​the student's student number, the second column is the knowledge level entity, the third column is the completion value of the knowledge level entity corresponding to the student in this row; the fourth column is the total score of the knowledge level entity in this row in this assessment; the fifth column is the score value of the knowledge level entity corresponding to the student in this row in this assessment;

[0107] Step 5: Read the learning response matrix generated in S03 or S04, and set the state values ​​of all knowledge level entities associated in the matrix to 1, that is, set it to the enabled state of the course. This function is one of the functions of the early warning system 107; the specific process (such as Figure 2 As shown): Start 201 first reads the learning behavior data 202, and then updates the entity state attribute 203;

[0108] Step 6: According to the learning response matrix generated in S03 or S04, the value of the ability value attribute of each entity in the learning object (i.e., each student) is calculated. This function is Figure 1 One of the functions of the early warning system 107, the specific process is as follows Figure 2In the course knowledge graph update 204, the value of the ability value attribute of each student's knowledge level entity is updated; here, the knowledge level entity E of the kth student w Ability value EY k,w For example, according to Calculation; where: EY k,w is the knowledge level entity E w The ability value, α k is the learning conversion rate of the kth student. When it is calculated for the first time, α k = 1. The ability values ​​of some knowledge level entities of the student with the student number 210***20 at a certain stage of learning (see Table 3):

[0109] Table 3 Capability values ​​of some knowledge level entities

[0110]

[0111] Step 7: Calculate the value of the predicted score attribute of the learning object (i.e. each student) at the knowledge level entity. This function is Figure 1 One of the functions of the early warning system 107, the specific process is as follows Figure 2 In the course knowledge graph update 204, the predicted score attribute of each student's knowledge level entity is updated; here, the knowledge level entity E of the kth student is used. w The predicted score ES k,pw Take this as an example, the specific steps are as follows:

[0112] S01: The kth student knowledge level entity E w Model score rate MSP k,gw Using the Rasch model Calculation; Ability value EY k,w The value range of is determined by the teacher according to the course as [x,s);

[0113] S02: Due to the calculation of the score rate interval [MSP x ,MSP s ) is not in the interval [0,1] that matches the truth, so we calculate the true score rate

[0114] S03: Calculate the knowledge level entity E of the kth student w The predicted score ES k,pw =ES w ·TSP k,gw ; The predicted scores of some knowledge level entities of the student with the student number 210***20 at a certain stage of learning (see Table 4):

[0115] Table 4 Prediction scores of some knowledge level entities

[0116]

[0117] Step 8: Calculate the value of the predicted total score attribute of the learning object (i.e. each student). This function is Figure 1 One of the functions of the early warning system 107, the specific process is as follows Figure 2 In the course knowledge graph update 204, the predicted total score attribute of each student is updated; here, the predicted total score S of the kth student is k,ap For example, the total score is predicted according to S k,ap =∑ES k,pw Calculation. The predicted total score of some students at a certain stage of study (see Table 5):

[0118] Table 5 Predicted total scores of some students

[0119]

[0120] Step 9: When the project type G = 0, that is, the learning project is a class, online learning or homework, skip this step; when the project type G = 1, that is, the learning project is a class test or exam, execute this step. This function is Figure 1 One of the functions of the early warning system 107, the specific process is as follows Figure 2 In the process, when G=0, it is determined 205 whether the next step is to calculate the learning conversion rate 206 or determine the warning level of each knowledge level entity 208; when G=0, the learning conversion rate 206 is calculated and the learning conversion rate 207 is updated; here, the learning conversion rate α of the kth student when the learning project is a class test or an exam k Take this as an example, the specific steps are as follows:

[0121] S01: Learning conversion rate α k The initial value of is 1, and the learning conversion rate α after each in-class test or exam is calculated k , taking the learning conversion rate α of the yth in-class test or exam closest to the current one k,y For example, calculate S k,apy =∑(ES wy ·TSP k,gw );Where: S k,ary is the actual total score of the kth student in the yth class test or examination closest to the current one, E wy is the knowledge level entity E in the yth in-class test or exam closest to the current one w The assessment score of

[0122] S02: Calculate the update learning conversion rate Where: n means that the most recent n assessments are included in the calculation of student conversion rate; QF h,yis the learning conversion rate weight coefficient of the yth assessment closest to the current one, which is determined by the teacher based on the course characteristics; QF α,h QF α,1 and QF α,2 The two weight coefficients of learning conversion rate are class test and examination, which are determined by teachers according to the characteristics of the course. k The rationality of QF h,y and QF α,h Needs to satisfy relationship The learning conversion rate of some students at a certain stage of learning (see Table 6):

[0123] Table 6 Learning conversion rate of some students

[0124]

[0125] Step 10: Based on the course knowledge graph and learning conversion rate updated in steps 4 to 9, calculate the real score rate of the knowledge level entities whose status value is equal to 1 in the course knowledge graph of the learning object (i.e., each student), and perform early warnings on the knowledge level entities respectively; here, the bth knowledge level entity E of the kth student whose status value is equal to 1 G＝1,b For example, the knowledge level entity E G＝1,b The true score rate is TSP k,gb , when I≤TSP k,gb The knowledge level entity E of the kth student G＝1,b The knowledge level entity warning level is "normal", Q gc ≤TSP k,gb <Q gn The knowledge level entity E of the kth student G＝1,b The knowledge level entity warning level is "prompt", V≤TSP k,gb <I when the knowledge level entity E of the kth student G＝1,b The knowledge level entity warning level is "Warning", TSP k,gb <V when the knowledge level entity E of the kth student G＝1,b The knowledge level entity warning level is "urgent"; where: I, V, and Z are the prompt threshold, warning threshold, and emergency threshold respectively. This function is Figure 1 One of the functions of the early warning system 107, the specific process is as follows Figure 2 Determine the warning level of each knowledge level entity 208; the warning level of some knowledge level entities of the student with the student number 210***30 at a certain stage of learning (as shown in Table 7):

[0126] Table 7 Prediction scores of some knowledge level entities

[0127]

[0128] Step Eleven: Calculate the learning warning level of the learning object (i.e., each student) based on the updated curriculum knowledge graph and learning conversion rate from Step Four to Step Nine. Here, taking the k-th student as an example, calculate the score of the knowledge level entity E with a status value of 1 in the curriculum knowledge graph of the student as ES G＝1,b ; the predicted score is ES G＝1,b ; the total predicted score ES of the knowledge level entities with a status value of 1 in the curriculum knowledge graph of the k-th student G＝1,k,pb = ∑ES G＝1,k,p and the cumulative total score ES G＝1,k,pb = ∑ES G＝1 ; further calculate the total predicted score rate of the k-th student G＝1,b ; when ES < H or I ≤ ESP G＝1 the learning warning level of the k-th student is "Normal"; when H ≤ ES G＝1,k,a,p and V ≤ ESP G＝1 < I, the learning warning level of the k-th student is "Prompt"; when H ≤ ES G＝1,k,a,p and Z ≤ ESP G＝1 < V, the learning warning level of the k-th student is "Warning"; when H ≤ ES G＝1,k,a,p and ESP G＝1 < Z, the learning warning level of the k-th student is "Urgent"; where: H is the warning threshold, determined by the teacher according to the actual curriculum, that is, when the cumulative total score ES G＝1,k,a,p ≥ H at the initial stage of the course learning, the teaching knowledge volume is sufficient to reflect the student warning status, and the learning warning of the student is enabled. This function is G＝1 one of the functions of the warning system 107 in Figure 1 , and the specific process is as Figure 2 Determine the learning warning level 209; the learning warning levels of some students at a certain learning stage are shown in Table 8:

[0129] Table 8 Learning Warning Levels of Some Students

[0130]

[0131] Step Twelve: Read and judge the learning warning level, and assign online learning content and homework questions to students through the learning plan and warning system based on the learning warning level; the specific steps are as follows (as Figure 1 and Figure 3 shown):

[0132] S01: Start 301 by first reading the learning warning level 302 of the learning object (i.e., each student);

[0133] S02: The warning level of the learning object is determined 303. If the warning level is normal or prompt, the learning object is terminated directly 309, and the online learning 103 content and homework 104 are arranged according to the teaching plan 101. If the warning level is warning or urgent, the learning object continues S03.

[0134] S03: Traverse all students whose warning levels are warning and emergency, and calculate the intervention improvement scores of each knowledge level entity whose status value is equal to 1; here, the knowledge level entity E of the kth student whose status value is equal to 1 G＝1,b Take the knowledge level entity E as an example. G＝1,b Completion value EX k,b , generating the entity intervention completion value 304 of each knowledge level as EX k,v,b =EX k,b +v, where: v is the intervention change value; according to the knowledge level entity E G＝1,b Completion value EX k,b and intervention completion value EX k,v,b Compute knowledge level entity E G＝1,b The predicted score ES k,pb and intervention prediction score ES k,v,pb , and then calculate the k-th student knowledge level entity E G＝1,b The intervention improved the ES score k,vx,pb =ES k,v,pb -EX k,pb ;

[0135] S04: Sort the knowledge level entities of the learning object based on the intervention effect; here, taking the kth student as an example, the knowledge level entities of the student with a status value equal to 1 are sorted according to the entity type order of "knowledge point", "knowledge cluster", and "assessment target", and then the knowledge level entities are further sorted in the entity types of "knowledge point", "knowledge cluster", and "assessment target" according to the intervention improvement score from high to low, so that the knowledge level entity ranking 306 of the kth student based on the intervention effect is obtained, wherein the knowledge level entity ranked q is represented by E G＝1,q express;

[0136] S05: Calculate the score required for the warning level of the object to be reduced to a prompt; here, taking the kth student as an example, the warning level of the student is reduced to a prompt score of 307, which is ES k,vx,r =ES G＝1 ·(I-ESP G＝1,k,a,p );

[0137] S06: Determine the online learning content or homework of the learning object; here, taking the kth student as an example, the knowledge level entity ranking 306 based on the intervention effect of the student is calculated, and the intervention improvement score ES is calculated. k,vx,pbThe stacking reaches the improved score ES k,vx,r The minimum number of knowledge level entities required is z, and the first z knowledge level entities are the knowledge level entities that need to be mastered to reduce the warning level to the prompt. According to the knowledge level entities, corresponding knowledge, cases, questions and other learning resources are given, intervention content 308 is given, and the end 309 is ended; at the same time, the intervention content 308 is given as the online learning 103 content or the homework 104 topic (such as Figure 1 As shown), it is sent to students through the early warning system 107.

Claims

1. The learning warning and intervention method based on the course knowledge graph is characterized by: The steps are as follows: Step 1: Establish the model layer of the course knowledge graph and determine the entities, attributes and relationships of the course knowledge graph. The specific steps are as follows: S01: Divide the entities of the course knowledge graph into three categories: knowledge level, learning resources and learning objects, wherein the entities, attributes and relationships are determined for the knowledge level and learning resources respectively, and the entities and attributes are determined for the learning objects; S02: In the knowledge hierarchy structure, three level entities, namely, assessment target, knowledge cluster and knowledge point, are set; S03: Define ID, name, status, score, predicted score, ability value, difficulty and completion value in the attributes of the knowledge level, wherein ID and name record the basic information of the knowledge level entity; the status attribute reflects the activation state of the knowledge level entity, when the status value of the knowledge level entity is equal to 0, it is in an unactivated state, indicating that the knowledge level entity has not been taught; when the status value of the knowledge level entity is equal to 1, it is in an enabled state, indicating that the knowledge level entity has been taught; and the score, predicted score, ability value, difficulty and completion value are all obtained by calculation, wherein the predicted score, ability value and completion value need to be stored separately according to the student ID; S04: defining three relationships of isPriorTo, isPartOf and isNecessaryFor in the relationship of the knowledge level, wherein the isPriorTo relationship indicates that the entities of the knowledge level are in a sequential order, the isPartOf relationship indicates that the entities of the knowledge level are in a containment and being contained relationship, and the isNecessaryFor relationship indicates that a certain knowledge level entity is a necessary condition for learning another knowledge level entity; S05: setting three types of entities, namely, knowledge, cases and questions, in the learning resource; S06: defining an ID, a name and content in the attributes of the learning resource, wherein the basic information of the learning resource is recorded using the ID and the name, and the above content is the content contained in the learning resource; S07: defining three relationships, hasKnowledge, hasExample and hasTest, in the relationship of the learning resource, to define the content of the learning resource as knowledge, examples or questions of knowledge level entities; S08: defining an ID, a name, a learning conversion rate and a predicted total score in the attributes of the learning object, recording basic information of the learning object using the ID and the name, and obtaining the learning conversion rate and the predicted total score by calculation; Step 2: According to the principle of "from high to low, gradually issued", determine the value of the score attribute of each entity in the knowledge level. The specific steps are as follows: S01: Determine the assessment score of each assessment objective based on the course syllabus as AOS i , where: i is the assessment target sequence number; the assessment target entity is defined as the central entity CE j , Assessment Target Score AOS i Defined as the central entity score CES j , then there is CES j =AOS i , where: j is the central entity serial number; S02: For all the central entities, search for all the central entities CE in the course knowledge graph. j n related entities, and assign scores to all n related entities according to the relationship type, such as the lth related entity RE j,l The distribution score is REDS j,l , and the central entity score is adjusted to CEJS j The total score after issuance needs to be the same as the original central entity score CES j Equal, that is, Where: n is the central entity CE j The total number of related entities, RE j,l CE j The lth related entity, REDS j,l CE j The score of the lth related entity, CEJS j is the adjusted score of the j-th central entity; S03: Calculate and update the scores of all relevant entities in S02 based on their scores before and after delivery; that is, the scores of relevant entities RE j,l RES j,l According to the relevant entity issued before the score REBS j,l With the distribution score REDS j,l , according to RES j,l =REBS j,l +REDS j,l Calculate update; where: RES j,l CE j The score of the lth related entity, REBS j,l CE j The score of the lth related entity before it is issued; S04: All related entities in S02 are used as the central nodes of the new round; that is, the related entities RE j,l As the central node of the new round of CE j ; S05: Repeat steps S02-S04 until all knowledge level entities are traversed and the scores are issued; S06: After the distribution is completed, in order to make the course knowledge graph easier to understand, the entities at each knowledge level are sorted and their corresponding scores are represented in a unified symbol format; that is, the knowledge level entity is represented by the symbol E w Indicates that the corresponding knowledge level entity score is represented by the symbol ES w Indicates; where: w is the knowledge level entity serial number; Step 3: Determine the value of the difficulty attribute of each entity in the knowledge level. The specific steps are as follows: S01: Calculate the true average score probability Among them: TSP mg is the true average score probability, S fa The total score is the sum of the total scores of the previous students’ in-class tests and the total scores of the exams; S fg It is the total score obtained by adding the scores of previous students' in-class tests and examinations. S02: Calculate the model average score rate MSP mg =TSP mg (MSP s -MSP x )+MSP x , x=-s;where: MSP mg is the average score rate of the model, x is the lower limit of the ability value, s is the upper limit of the ability value, e is the natural logarithm, MSP x is the lower limit of the model score rate, MSP s is the upper limit of the model score rate; S03: Calculate the overall average capability value Where: Y am is the overall average capability value; S04: Calculate the coefficient of the capability value to be determined Among them: q is the undetermined coefficient of capability value; S05: Calculate the real average score rate of the knowledge level entity; that is, calculate the knowledge level entity E w The actual average scoring rate Among them: TSP mgw is the knowledge level entity E w The real average score rate, ES faw The knowledge level entity E in the past students' class tests and exams w The total score of the superposition, ES fgw The knowledge level entity E in the past students' class tests and exams w The total score of S06: Calculate the average model score rate of the knowledge level entity; that is, calculate the knowledge level entity E w The average score rate of the model MSP mgw =TSP mgw (MSP s -MSP x )+MSP x ; S07: Calculate the average capability value of the knowledge level entity; that is, calculate the average capability value of the knowledge level entity E w The average ability value S08: Calculate the difficulty of the knowledge level entity; that is, calculate the difficulty of the knowledge level entity E w Difficulty Step 4: Generate a learning response matrix for input into the early warning model for analysis. The specific steps are as follows: S01: The type of learning project is represented by G. When the input data is class, online learning or homework, G = 0; when it is a class test or exam, G = 1; S02: Calculate the completion value of each knowledge level entity of the learning object's class, online learning, homework, in-class test and examination items; that is: the knowledge level entity E of the kth student w EX k,w =L h , online learning completion value Job Completion Value Completion value of class test Exam completion value Among them: aw , W aw , Q aw or F aw They are respectively about the knowledge level entity E w Total online learning time, homework score, class test score or exam score; k,gw , W k,gw , Q k,gw or F k,gw are the kth student's knowledge level entity E w Online learning time, homework scores, class test scores or exam scores; L h , O h , W h , Q h or F h The completion value weight coefficients of class, online learning, homework, test or examination are defined by the teacher according to the characteristics of the course, reflecting the contribution of each project to the completion value; L h , O h and W h Need to satisfy the relationship L h +O h +W h =1, that is, when the learning conversion rate of the kth student is α k =1, for knowledge level entity E w After completing the complete classroom and online learning and correctly completing the homework, the knowledge level entity E w Ability value EY k,w Will reach the average capability value EY mw level; S03: When the item type G = 0, generate a learning response matrix: Where: k is the student serial number; S04: When the project type G=1, generate a learning response matrix: Among them: QFES aw The knowledge level entity E in this class test or exam w Total score of QFES k,gw is the kth student knowledge level entity E in the class test or exam w The score value of Step 5: According to the learning response matrix, the state values ​​of all knowledge level entities associated in the matrix are set to 1, that is, set to the enabled state of the taught course; Step 6: Based on the learning response matrix, calculate the value of the capability attribute of each entity in the knowledge level of the learning object; that is, the kth student's ability attribute of each entity in the knowledge level E w The ability value is based on Calculation; where: EY k,w is the knowledge level entity E w The ability value of α is calculated for the first time. k =1; Step 7: Calculate the predicted score attribute value of each student in the knowledge level; that is, calculate the knowledge level entity E of the kth student w The predicted score ES k,pw , where: ES k,pw is the knowledge level entity E of the kth student w The prediction score of ; the specific steps are as follows: S01: The kth student knowledge level entity E w Model score rate MSP k,gw Using the Rasch model Calculation, ability value EY k,w The value range of MSP is determined by the teacher according to the course as [x,s); k,gw is the kth student knowledge level entity E w Model score rate; S02: Due to the k-th student knowledge level entity E w The calculation model score rate interval [MSP x ,MSP s ) is not in the interval [0,1] that matches the truth, so we calculate the true score rate: S03: Calculate the knowledge level entity E of the kth student w The predicted score ES k,pw =ES w ·TSP k,gw ; Among them: ES k,pw is the kth student knowledge level entity E w The prediction score of Step 8: Calculate the value of the predicted total score attribute of the learning object; that is, the predicted total score of the kth student is calculated according to S k,ap =ΣES k,pw Calculate; where: S k,ap is the predicted total score of the kth student; Step 9: When the project type G = 0, that is, the learning project is a class, online learning or homework, skip this step; when the project type G = 1, that is, the learning project is a class test or exam, execute this step and calculate the learning conversion rate of each student; that is, when the learning project is a class test or exam, calculate the learning conversion rate of the kth student as α k , the specific steps are as follows: S01: The learning conversion rate of the kth student is α k , learning conversion rate α k The initial value of is 1, and the learning conversion rate of the yth in-class test or exam closest to the current one is calculated as α k,y ,calculate S k,apy =Σ(ES wy ·TSP k,gw );where: S k,ary is the actual total score of the kth student in the yth class test or examination closest to the current one, E wy is the knowledge level entity E in the yth in-class test or exam closest to the current one w The assessment score of S02: Calculate the update learning conversion rate Where: n means that the most recent n in-class tests or exams are included in the calculation of student conversion rate; QF h,y is the weight coefficient of the yth in-class test or examination closest to the current one, which is determined by the teacher based on the characteristics of the course; QF α,h QF α,1 and QF α,2 The two weight coefficients of learning conversion rate are class test and examination, which are determined by teachers according to the characteristics of the course. k The rationality of QF h,y and QF α,h Needs to satisfy relationship Step 10: Based on the course knowledge graph and learning conversion rate updated in steps 4 to 9, calculate the real score rate of the knowledge level entities with a status value equal to 1 in the course knowledge graph of the learning object, and perform early warnings on the knowledge level entities respectively; that is, the bth knowledge level entity with a status value equal to 1 of the kth student is E G＝1,b ; Where: b is the serial number in the knowledge level entity sequence with a state value of 1; knowledge level entity E G＝1,b The true score rate is TSP k,gb , when I≤TSP k,gb When the k-th student's knowledge level entity E G＝1,b The knowledge level entity warning level is "normal"; when Q gc ≤TSP k,gb <Q gn When the k-th student's knowledge level entity E G＝1,b The knowledge level entity warning level is "prompt"; when V≤TSP k,gb <I, the knowledge level entity E of the kth student G＝1,b The knowledge level entity warning level is "warning"; when TSP k,gb <V, the knowledge level entity E of the kth student G＝1,b The knowledge level entity warning level is "urgent"; where: I, V, and Z are respectively the prompt threshold, warning threshold, and emergency threshold, which are determined by the teacher according to the actual situation of the course; Step Eleven: Calculate the learning warning level of the learning object based on the updated curriculum knowledge graph and learning conversion rate in Steps Four to Nine; that is: the score of the knowledge level entity E with a status value of 1 in the curriculum knowledge graph of the k-th student is ES G＝1,b ; the predicted score is ES G＝1,b ; the total predicted score ES G＝1,k,pb of the knowledge level entities with a status value of 1 in the curriculum knowledge graph of the k-th student G＝1,k,p = ∑ES G＝1,k,pb and the cumulative total score ES G＝1 = ΣES G＝1,b ; the total predicted score rate of the calculation model of the k-th student When ES G＝1 < H or I ≤ ESP G＝1,k,a,p Then the learning warning level of the k-th student is "Normal"; when H ≤ ES G＝1 and V ≤ ESP G＝1,k,a,p < I, then the learning warning level of the k-th student is "Prompt"; when H ≤ ES G＝1 and Z ≤ ESP G＝1,k,a,p < V, then the learning warning level of the k-th student is "Warning"; when H ≤ ES G＝1 and ESP G＝1,k,a,p < Z, then the learning warning level of the k-th student is "Urgent"; where: H is the warning start threshold, determined by the teacher according to the actual curriculum, that is, when the cumulative total score ES G＝1 of the knowledge level entities with a status value of 1 is ≥ H, it is the initial state at the end of the curriculum learning; when the teaching knowledge amount is sufficient to reflect the student warning status, the student learning warning is turned on; Step 12: Read and determine the learning warning level, and assign online learning content and homework questions to students through the learning plan and warning system based on the learning warning level; the specific steps are as follows: S01: Reading the learning warning level of the learning object; S02: Determine the warning level of the learning object. If the warning level is normal or prompt, end this step and arrange online learning content and homework according to the teaching plan. If the warning level is warning or urgent, proceed to S03. S03: Traverse all students whose warning levels are warning and emergency, and calculate the intervention improvement scores of each knowledge level entity whose status value is equal to 1; that is: the knowledge level entity E of the kth student G＝1,b The completion value is EX k,b , the intervention completion value is EX k,v,b =EX k,b +v, v is the intervention change value, which is determined by the teacher according to the actual course; according to the knowledge level entity E G＝1,b Completion value EX k,b and intervention completion value EX k,v,b Compute knowledge level entity E G＝1,b The predicted score ES k,pb and intervention prediction score ES k,v,pb , and then calculate the k-th student knowledge level entity E G＝1,b The intervention improved the ES score k,vx,pb =ES k,v,pb -EX k,pb ; S04: Sort the knowledge level entities of the learning object based on the intervention effect; that is, the knowledge level entities of the student whose status value is equal to 1 are sorted according to the entity type order of "knowledge point", "knowledge cluster" and "assessment target", and then the knowledge level entities are further sorted in the entity types of "knowledge point", "knowledge cluster" and "assessment target" according to the intervention improvement score from high to low, so as to obtain the knowledge level entity sorting of the kth student based on the intervention effect, among which the knowledge level entity ranked q is used E G＝1,q express; S05: Calculate the score required for the warning level of the learning object to be reduced to a prompt; that is, calculate the score ES required for the warning level of the k-th student to be reduced to a prompt k,vx,r =ES G＝1 ·(I-ESP G＝1,k,a,p ); S06: Determine the online learning content or homework of the learning object; sort the knowledge level entities based on the intervention effect of the student, and calculate the model intervention improvement score ES k,vx,pb The superposition reaches the warning level and is reduced to the required score ES k,vx,r The minimum number of knowledge-level entities required is z, and the first z knowledge-level entities are the knowledge-level entities that need to be mastered and the warning levels are reduced to prompts. According to the knowledge-level entities, corresponding learning resources of knowledge, cases and questions are given, and intervention content is given and distributed to students.

2. The learning warning and intervention method based on course knowledge graph according to claim 1 is characterized in that: The learning objects are all individual students.

3. The learning warning and intervention method based on course knowledge graph according to claim 1 is characterized in that: The intervention content serves as online learning content or homework questions.

Images