A teacher matching method and system based on big data analysis

Abstract

The application discloses a teacher matching method and system based on big data analysis, and relates to the technical field of data processing, which comprises the following steps: intercepting a state transition segment; searching a teacher historical teaching record according to a falling direction represented by the state transition segment, and reserving a historical teaching segment corresponding to the falling direction in a knowledge state trajectory before teaching; generating a teacher teaching effectiveness parameter indexed by a knowledge point range in which a mastery probability is converted from continuous falling to continuous rising according to the knowledge state trajectory before and after the historical teaching segment; and outputting a teacher matching result according to the result of point-by-point offsetting of the mastery probability falling amount in the state transition segment based on the teacher teaching effectiveness parameter. The application is beneficial to improving the adaptability of the teacher recommendation result to the knowledge state change process of students.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, specifically to a teacher matching method and system based on big data analysis. Background Technology

[0002] With the continuous development of online education platforms, these platforms have accumulated a large amount of learning process data and teaching service data. Teacher recommendation methods are gradually shifting from manual screening to automated matching based on data models. Current teacher matching typically involves calculations based on course category, teacher experience, historical evaluations, and student needs tags, which can meet basic screening requirements. However, the related processing relies heavily on static profiles and comprehensive evaluation data, making it difficult to reflect the dynamic changes in students' knowledge status during the answering process, and also failing to depict teachers' actual ability to improve specific learning status changes.

[0003] In actual teaching services, students' mastery of knowledge points changes with continuous answering, knowledge point dependencies, and learning feedback. Errors in subsequent knowledge points also reflect a decline in mastery of prior knowledge points. If teacher matching remains at the level of course tags or overall teacher ratings, the recommended results are difficult to establish a stable data correlation with changes in students' current knowledge status. Although existing platforms can infer students' mastery through knowledge tracking models, the lack of a fine-grained calculation mechanism between teachers' historical teaching effectiveness and students' current knowledge status results in insufficient adaptation of teacher recommendations to students' actual learning needs. Summary of the Invention

[0004] The purpose of this invention is to provide a teacher matching method and system based on big data analysis to solve the problems mentioned in the background.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] In a first aspect, the present invention provides a teacher matching method based on big data analysis, comprising:

[0007] Input the target student's answer sequence into the knowledge tracking model to form the target student's knowledge state trajectory, which records the mastery probability of each knowledge point at the time of answering. Then, extract the state transition segments along the knowledge point dependency path, where the mastery probability continuously decreases from the subsequent knowledge point to the previous knowledge point.

[0008] Retrieve the teacher's historical teaching records based on the descent direction represented by the state transition segment, and retain the historical teaching segments corresponding to the knowledge state trajectory before the teaching and the descent direction;

[0009] Based on the knowledge state trajectory before and after the historical teaching segment, teacher teaching effectiveness parameters are generated, with the range of knowledge points whose mastery probability changes from continuous decline to continuous increase as the index.

[0010] Based on the teacher teaching effectiveness parameters, the decrease in the mastery probability of each knowledge point in the state transition segment is calculated on a knowledge point-by-knowledge point basis to obtain the candidate teacher matching value, and the teacher matching result is output according to the candidate teacher matching value.

[0011] Secondly, the present invention provides a teacher matching system based on big data analysis, implemented based on the method described above, comprising:

[0012] The tracking module is used to input the target student's answer sequence into the knowledge tracking model, form the target student's knowledge state trajectory by recording the mastery probability of each knowledge point at the time of answering, and extract the state transition segment where the mastery probability continuously decreases from the subsequent knowledge point to the previous knowledge point along the knowledge point dependency path.

[0013] The retrieval module is used to retrieve the teacher's historical teaching records in the descent direction represented by the state transition segment, and retain the historical teaching segments corresponding to the knowledge state trajectory before the teaching and the descent direction.

[0014] The evaluation module is used to generate teacher teaching effectiveness parameters based on the knowledge status trajectory before and after the teaching of the historical teaching segments. The teacher teaching effectiveness parameters are indexed by the range of knowledge points whose mastery probability changes from continuous decline to continuous increase.

[0015] The matching module is used to perform a knowledge-point-by-knowledge-point offsetting calculation on the decrease in the mastery probability of each knowledge point in the state transition segment based on the teacher teaching effectiveness parameters, obtain the candidate teacher matching value, and output the teacher matching result according to the candidate teacher matching value.

[0016] The technical effects and advantages provided by the present invention in the above technical solution are as follows:

[0017] This invention transforms the target student's answering process into a state transition segment that continuously descends along a knowledge point-dependent path. It then uses this descent direction to retrieve the teacher's historical teaching records, shifting the teacher recommendation basis from course tags and comprehensive evaluations to the path correspondence of knowledge state changes before instruction. Simultaneously, by comparing the knowledge state trajectories before and after historical teaching segments, it generates teacher teaching effectiveness parameters indexed by the range of knowledge points that transition from continuous decline to continuous rise. These parameters are then used to offset the decrease in the target student's mastery probability on a knowledge point-by-knowledge-point basis. This enables a calculable correlation between teacher matching results and the student's current knowledge state change process, reducing the impact of static profiles on recommendation results and improving the adaptability of teacher recommendations to students' actual learning needs. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0019] Figure 1 A flowchart illustrating a teacher matching method based on big data analysis provided in an embodiment of the present invention;

[0020] Figure 2 This is a schematic diagram of a teacher matching system based on big data analysis, provided as an embodiment of the present invention. Detailed Implementation

[0021] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, they are provided to make the description of this application more complete and comprehensive, and to fully convey the concept of the exemplary embodiments to those skilled in the art. The drawings are merely illustrative illustrations of this application and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted.

[0022] Furthermore, the described features, structures, or characteristics can be combined in any suitable manner in one or more exemplary embodiments. Numerous specific details are provided in the following description to give a full understanding of the exemplary embodiments disclosed in this application. However, those skilled in the art will recognize that the technical solutions disclosed in this application can be practiced with one or more specific details omitted, or other methods, components, steps, etc., can be employed. In other instances, well-known structures, methods, implementations, or operations are not shown or described in detail to avoid obscuring various aspects of the disclosure of this application.

[0023] Example 1

[0024] like Figure 1As shown, this embodiment discloses a teacher matching method based on big data analysis, applicable to teacher recommendation in online education platforms. The online education platform stores the target student's answer sequence, the correspondence between question knowledge points, the knowledge point dependency path, and the teacher's historical teaching records. The target student's answer sequence originates from the target student's continuous answer records on the online education platform. The correspondence between question knowledge points originates from the platform's question bank's labeled data on the knowledge points tested in the questions. The knowledge point dependency path originates from the sequential learning dependency order between different knowledge points in the course knowledge system. The teacher's historical teaching records originate from the historical teaching service data stored on the online education platform, used to record the teacher's answer data and teaching data before and after teaching history students.

[0025] The online education platform generates a knowledge state trajectory of the target student based on the target student's answer sequence, and extracts state transition segments where the target student's mastery probability continuously decreases from the knowledge point dependency path. Subsequently, it retrieves teachers' historical teaching records based on these state transition segments, filtering out historical teaching segments corresponding to the target student's decreasing direction. Then, it generates teacher teaching effectiveness parameters based on the changes in knowledge state before and after the teaching of these historical segments. Finally, it outputs teacher matching results based on the knowledge point-by-knowledge point offsetting effect of the decrease in mastery probability in the target student's state transition segments, using the teacher teaching effectiveness parameters. Specific methods include:

[0026] S101: Input the target student's answer sequence into the knowledge tracking model to form the target student's knowledge state trajectory, which records the mastery probability of each knowledge point at the time of answering, and extract the state transition segment along the knowledge point dependency path where the mastery probability continuously decreases from the subsequent knowledge point to the previous knowledge point.

[0027] The online education platform reads the target student's answer sequence and sorts it according to the time of answering. Each answer record in the target student's answer sequence includes the answer time, question identifier, and answer result. The question identifier is used to retrieve the knowledge point tested by the question from the knowledge point correspondence. The answer result indicates the target student's answer status for the question, and the answer result can be marked as correct, incorrect, or a score-based result saved by the platform.

[0028] The knowledge tracing model employs a pre-trained deep knowledge tracing model. The model comprises an input representation layer, a temporal state update layer, and a probability output layer. The input representation layer converts the question identifier, knowledge points, and answer result into vectors. For each answer time... Corresponding question identifier Knowledge Points and the answer results The input representation layer forms the input vector: ;

[0029] In the formula, Indicates the title identifier Embedded vector, Representing knowledge points Embedded vector, Indicates the answer result Embedded vector, This represents vector concatenation. The temporal state update layer updates the learning state of the target student according to the answer time, and can use a gated recurrent unit network or a long short-term memory network. When using a gated recurrent unit network, the state update form is as follows: ;

[0030] In the formula, Indicates the time of answering. The corresponding learning state vector, This represents the learning state vector corresponding to the previous response. The probability output layer outputs the mastery probability of each knowledge point based on the learning state vector:

[0031] ;

[0032] In the formula, express function, This represents the output weight matrix. This represents the output bias vector. Therefore, Each component is located at to between.

[0033] For the time to answer The knowledge tracing model outputs a probability vector of the target student's mastery of knowledge points:

[0034] ;

[0035] In the formula, Indicates the target student's answer time Knowledge points The probability of mastering This represents the number of knowledge points participating in knowledge tracking on the online education platform. The online education platform arranges the mastery probability vectors of each knowledge point according to the time of answering, forming a knowledge state trajectory for the target student.

[0036] ;

[0037] In the formula, This represents the number of answer records in the target student's answer sequence. The target student's knowledge state trajectory records the probability of mastering each knowledge point by the target student at the time of answering, and is used to characterize the changes in the target student's knowledge state during the continuous answering process.

[0038] The training data for the knowledge tracking model comes from historical student answer sequences stored on the online education platform. Each historical answer record includes the answer time, question identifier, knowledge point, and answer result. During training, the historical student's answer records before the current answer time are input into the knowledge tracking model, and the actual answer result of the question corresponding to the current answer time is used as the supervision label. The training loss function can adopt binary classification cross-entropy loss:

[0039] ;

[0040] In the formula, Indicates the number of training samples. Indicates the first The actual response results of each training sample This represents the probability of mastering the corresponding knowledge point output by the knowledge tracing model. After model training, the knowledge tracing model outputs the probability of mastering the knowledge point as it changes according to the target student's answer sequence. The above model structure, training data, and loss function enable the formation process of the target student's knowledge state trajectory to be feasible.

[0041] The target student's answer sequence is indexed by the answer time, the answer time corresponds to the question identifier, the question identifier points to the knowledge point through the question knowledge point correspondence, and the knowledge point has a path position in the knowledge point dependency path.

[0042] Each answer record in the target student's answer sequence is indexed by the answer time. The online education platform determines the order of different answer records based on the answer time. The question identifier is used to find the corresponding knowledge point in the question-knowledge point mapping. The knowledge point dependency path represents the learning dependency order between knowledge points. The path position of a knowledge point in the knowledge point dependency path determines its position relative to other knowledge points. Knowledge points at the beginning of the learning dependency order are called prerequisite knowledge points, and knowledge points at the end of the learning dependency order are called subsequent knowledge points.

[0043] As a data example, the knowledge point dependency path in a middle school mathematics linear function course can be sequentially included: equation transformation, linear function expression, linear function graph, and linear function word problems. Equation transformation is at the beginning, and linear function word problems are at the end. If, after a target student makes consecutive mistakes on linear function word problems, the probability of mastering the linear function graph and linear function expression also decreases with each attempt, then the target student's learning status shows a continuous decline along this knowledge point dependency path, moving from subsequent knowledge points to earlier ones. This continuous decline reflects the direction of the transmission of the target student's current learning problem along the knowledge point dependency path.

[0044] Specifically, the step of extracting state transition segments along the knowledge point dependency path where the mastery probability continuously decreases from subsequent knowledge points to preceding knowledge points includes:

[0045] Based on the knowledge point mastery probability read from the target student's knowledge state trajectory at the time of answering, a knowledge point mastery probability sequence is formed.

[0046] The online education platform reads the probability of mastery of each knowledge point from the target student's knowledge status trajectory at the time of answering the question. For each knowledge point... The online education platform generates a sequence of mastery probabilities for each knowledge point:

[0047] ;

[0048] In the formula, Indicates the target students' knowledge points The sequence of mastery probabilities is arranged according to the time of answering. For each knowledge point in the knowledge point dependency path, the mastery probability is read in the same way. The mastery probability sequences corresponding to multiple knowledge points together form the knowledge point mastery probability sequence. This sequence serves as the basic data for path configuration and the extraction of continuous descending trajectory segments.

[0049] Configure the knowledge point mastery probability sequence to the knowledge point dependency path corresponding to the target student's answer sequence to obtain the path mastery probability sequence;

[0050] The online education platform determines the knowledge points indicated by each question identifier in the target student's answer sequence based on the correspondence between the questions and the knowledge points, and then reads the path location of each knowledge point based on the knowledge point dependency path. For a knowledge point dependency path: ;

[0051] In the formula, This indicates the prerequisite knowledge points in the path. This indicates the subsequent knowledge points in the path. This indicates the number of knowledge points contained in the knowledge point dependency path. The online education platform allocates the knowledge point mastery probability sequence to the knowledge point dependency path according to the knowledge point's position within that path. (For the answering time...) The path mastery probability sequence is represented as:

[0052] ;

[0053] In the formula, Indicates the target student's answer time The sequence of mastery probabilities along the knowledge point's dependency path. The online education platform arranges these sequences according to the time of answering. This process obtains a path mastery probability sequence. By doing so, the mastery probabilities scattered across different knowledge points are configured into the same knowledge point dependency path, enabling subsequent processing to determine whether the mastery probability continuously decreases along the direction from subsequent knowledge points to preceding knowledge points.

[0054] In the path mastery probability sequence, a continuously descending trajectory segment is extracted along the direction from the subsequent knowledge point to the preceding knowledge point to form the state transition segment.

[0055] The change in the probability of mastery between adjacent answering times in the read path probability sequence of the online education platform. For knowledge points... During the answering period The decrease in the probability of mastery within is expressed as: ;

[0056] In the formula, Indicates the start time of the answering period. Indicates the end of the answering period. Indicates the target students' knowledge points The decrease in the probability of mastery. When When the value is positive, it indicates that the probability of the target student mastering the knowledge point has decreased.

[0057] The online education platform starts with subsequent knowledge points in the knowledge point dependency path and sequentially reads the decrease in mastery probability for each knowledge point along the direction pointing to the preceding knowledge point. When multiple knowledge points on adjacent path positions have decreases in mastery probability, the online education platform extracts the corresponding continuous trajectory segment as a state transition segment. The state transition segment includes the starting knowledge point, the ending knowledge point, the direction of descent, the answering time period, and the decrease in mastery probability for each knowledge point.

[0058] As a data example, if the decrease in the probability of target students mastering the three knowledge points of linear function application problems, linear function graphs, and linear function expressions is as follows: ;

[0059] Furthermore, if the three knowledge points are arranged consecutively along the knowledge point dependency path from the subsequent knowledge point to the preceding knowledge point, the online education platform will extract this continuous downward trajectory segment as a state transition segment. This state transition segment indicates that the target student's current learning state is continuously decreasing along the knowledge point dependency path. This processing combines the teacher's recommendation criteria with the direction of change in the target student's knowledge state, which can reduce the recommendation bias caused by matching solely based on course tags or incorrect knowledge points.

[0060] S102: Retrieve the teacher's historical teaching records according to the descent direction represented by the state transition segment, and retain the historical teaching segments corresponding to the knowledge state trajectory before teaching and the descent direction;

[0061] The online education platform uses the resulting state transition segments as the retrieval basis. The descent direction represented by these state transition segments includes the starting knowledge point, the ending knowledge point, and a continuous descent path from subsequent knowledge points to preceding knowledge points. The online education platform searches for historical student records in the teacher's historical teaching records that correspond to the pre-lecture knowledge state trajectory and this descent direction, and retains the corresponding teacher's historical teaching records as historical teaching segments. This process shifts the selection criterion for teacher historical teaching records from the course name to the path change direction of the pre-lecture knowledge state trajectory, providing a data foundation for subsequent evaluation of whether teachers have processed a continuous descent process corresponding to the target student's current learning state.

[0062] The teacher's historical teaching record is indexed by the start time of the teaching session. The start time of the teaching session corresponds to the teacher's identifier, the teacher's identifier corresponds to the teaching content sequence, the teaching content sequence corresponds to the historical student answer sequence, and the historical student answer sequence includes answer records before and after the start time of the teaching session.

[0063] The teacher's historical teaching records are indexed by the start time of the lesson. The start time is used to distinguish between pre-lesson and post-lesson response records. The start time corresponds to a teacher identifier, used to identify the teacher who conducted the lesson. The teacher identifier corresponds to a sequence of teaching content, used to identify the teaching content covered in that lesson. The sequence of teaching content corresponds to a sequence of historical student responses. This sequence includes responses before and after the start time of the lesson. The historical student response sequence before the start time is used to form a pre-lesson knowledge state trajectory, and the historical student response sequence after the start time is used to form a post-lesson knowledge state trajectory.

[0064] No. A teacher's historical teaching record is represented as follows:

[0065] ;

[0066] In the formula, Indicates the first The start time of each teacher's historical teaching record. This indicates the corresponding teacher identifier. This indicates the corresponding sequence of teaching content. This indicates the sequence of historical student responses prior to the start of the lesson. This indicates the sequence of history students' answers after the start of the lesson.

[0067] Specifically, the retention of the historical teaching segment corresponding to the pre-teaching knowledge state trajectory and the descent direction includes:

[0068] Using the start time of the lecture in the teacher's historical lecture record as the boundary, the sequence of historical student answers before the start time of the lecture is input into the knowledge tracking model to form a knowledge state trajectory before the lecture.

[0069] Online education platforms use the start time of lessons from the teacher's historical teaching records. As a boundary, read the sequence of historical student answers before the start of the lesson. Online education platforms will Input the knowledge tracing model to obtain the probability of a student's mastery of knowledge points before the start of the lesson, and form a knowledge state trajectory before the start of the lesson:

[0070] ;

[0071] In the formula, Indicates the first In the teacher's history lesson record, the time when history students answered questions before the start of the lesson. The corresponding knowledge point mastery probability vector. The pre-instruction knowledge state trajectory is used to determine whether there is a continuous downward direction corresponding to the target student's state transition segment before receiving instruction.

[0072] Configure the state transition segment and the pre-lecture knowledge state trajectory to the same knowledge point dependency path, and calculate the path difference between the pre-lecture knowledge state trajectory and the state transition segment;

[0073] The online education platform configures the state transition segments of the target students and the pre-lesson knowledge state trajectories of historical students onto the same knowledge point dependency path. After configuration, the order of knowledge points in the state transition segments remains consistent with the order of knowledge points in the pre-lesson knowledge state trajectories. The online education platform reads the decrease in mastery probability of each knowledge point on the same path for both students and calculates the path difference. The path difference is used to compare the differences in the magnitude and pattern of the decrease between the target students and historical students.

[0074] The calculation of the path difference between the pre-lecture knowledge state trajectory and the state transition segment includes:

[0075] Based on the knowledge point dependency path of the state transition segment, the decrease in the mastery probability of each knowledge point within the state transition segment is read to form a sequence of changes in the target student's mastery probability. and read the first The decrease in the probability of mastery of the same knowledge point within the knowledge state trajectory before the lesson is taught forms a historical sequence of changes in students' mastery probability. ;

[0076] The target students will master the use of probability change sequences. express:

[0077] ;

[0078] In the formula, This represents the first segment in the target student's state transition sequence. The decrease in the probability of mastering a knowledge point corresponding to each path location. (This is used in the context of a sequence showing changes in the probability of historical students mastering a knowledge point.) express:

[0079] ;

[0080] In the formula, Indicates the first The decrease in the probability of mastering the knowledge point corresponding to the same path position in the pre-lesson knowledge state trajectory. The sequence of changes in the mastery probability of the target students and the sequence of changes in the mastery probability of historical students use the same order of knowledge points, and the two can be compared point by point.

[0081] Calculate the first according to the following formula. Path difference between the pre-lesson knowledge state trajectory and the state transition segment :

[0082] ;

[0083] In the formula, This represents the first difference of the probability change sequence mastered by the target student. This represents the first difference of the sequence of historical student mastery probability changes. express Norm.

[0084] in, This is used to calculate the difference in the decrease in the probability of mastering each knowledge point on the same path between target students and historical students. This is used to calculate the difference in descent patterns between the two along the knowledge point-dependent path. The path difference reflects both the magnitude of the descent and the pattern of path change, ensuring that the retrieval criteria for teachers' historical teaching records correspond to the target student's state transition segments. This process can reduce the misselection of historical teaching records caused by comparing only the descent of a single knowledge point.

[0085] As a supplementary explanation, if the target students master the probability change sequence as follows: ;

[0086] No. The history students mastered the probability change sequence as follows: ;

[0087] The difference in the decrease in the probability of mastery is: ;

[0088] The target students will master the first difference of probability change sequences: ;

[0089] History students' understanding of the first difference of probability change sequences is as follows: ;

[0090] The differences in descent patterns are as follows: ;

[0091] Therefore, the path difference is: ;

[0092] The path difference quantity in the above example simultaneously represents the difference in the decrease in mastery probability and the difference in the path decrease pattern. By filtering teachers' historical teaching records using this path difference quantity, a definite correspondence can be established between the retained historical teaching segments and the target student's current state transition segments on the knowledge point dependency path.

[0093] Using teacher identifiers as units, the corresponding teacher's historical teaching records are sorted in ascending order according to the path difference, and the teacher's historical teaching record at the top of the list is read and saved as the historical teaching segment.

[0094] The online education platform groups teachers' historical teaching records according to teacher identifiers. When a teacher corresponds to multiple historical teaching records, the platform calculates the path difference for each record and sorts them in ascending order of path difference. The teacher's historical teaching record ranked first has the lowest path difference between the pre-lesson knowledge state trajectory and the target student's state transition segment. The platform reads this record and saves it as the corresponding teacher's historical teaching segment.

[0095] As a specific data example, teachers For the three teacher history teaching records, the path differences are as follows: , and The platform sorts the three teacher's historical teaching records in ascending order by path difference, and reads the path difference as follows: Teachers' historical teaching records are preserved for teachers. The corresponding historical lesson segment. (Teacher) For two corresponding teacher history teaching records, the path difference is as follows: and The platform read path difference is... Teachers' historical teaching records are preserved for teachers. The platform retrieves the corresponding historical teaching segments. If two historical teaching records for the same teacher have the same path difference, the platform reads the earlier historical teaching record in the order of the teaching start time. If the teaching start times are the same, the platform reads the earlier historical teaching record in the database in ascending order of the record number.

[0096] S103: Based on the knowledge state trajectory before and after the historical teaching segment, generate teacher teaching effectiveness parameters, with the range of knowledge points whose mastery probability changes from a continuous decrease to a continuous increase as the index;

[0097] The online education platform retrieves the start time of each lesson from historical lecture segments, along with the teacher's identifier, the sequence of student responses before the start time, and the sequence of student responses after the start time. The sequence of student responses before the start time is used to form the pre-lecture knowledge state trajectory. The sequence of student responses after the start time is used to form the post-lecture knowledge state trajectory. The teacher identifier is used to aggregate historical lecture segments belonging to the same teacher.

[0098] The input data for this step includes historical teaching segments, the pre-teaching knowledge state trajectory, the historical student response sequence after the start of the lesson, and the knowledge point dependency paths corresponding to the state transition segments. The output data is teacher teaching effectiveness parameters. These parameters are indexed by knowledge point range and represent the extent of improvement a teacher makes in continuously declining knowledge state within a given knowledge point range. These parameters originate from changes in the knowledge state trajectory before and after the lesson, rather than from teacher evaluation texts or course tags, providing teachers with data supporting their recommendations based on the changes in the target students' knowledge state.

[0099] Specifically, based on the knowledge state trajectory before and after the historical teaching segment, teacher teaching effectiveness parameters are generated, including:

[0100] Input the sequence of historical students' answers after the start of the lesson in the historical teaching segment into the knowledge tracking model to form a knowledge state trajectory after the lesson.

[0101] The online education platform retrieves the sequence of historical student responses from historical lesson segments, starting after the beginning of each lesson. This sequence includes the response time, question identifier, and response result. The platform sorts the sequence according to the response time and inputs it into a knowledge tracing model. This model is the same as the one used to form the target student's knowledge state trajectory and the pre-lesson knowledge state trajectory, ensuring consistent output accuracy for mastery probabilities before and after the lesson.

[0102] For the A historical lecture segment, the sequence of student responses after the start of the lecture is denoted as follows: .Will After inputting the knowledge tracing model, we obtain the probability vector of knowledge point mastery at each answering time after the lecture:

[0103] ;

[0104] In the formula, Indicates the first In a segment of a history lesson, history students answer questions after the lesson. Knowledge points The probability of mastery. Arrange the above knowledge points into a mastery probability vector according to the time of answering, forming a knowledge state trajectory after instruction:

[0105] ;

[0106] In the formula, Indicates the first The starting point of a historical lesson segment. The post-lesson knowledge state trajectory is used to characterize the change in the probability of history students mastering the knowledge points after receiving the lesson.

[0107] Both the post-lecture knowledge state trajectory and the pre-lecture knowledge state trajectory are output by the same knowledge tracking model, and both use the same set of knowledge points as the output dimension. Therefore, the changes in mastery probability before and after the lecture can be aligned and compared along the same knowledge point dependency path.

[0108] Align the knowledge state trajectory before and after the teaching with the knowledge state trajectory after the teaching according to the knowledge point dependency path corresponding to the state transition segment, extract the range of knowledge points whose mastery probability changes from continuous decline to continuous increase, and calculate the change in state before and after the teaching for each knowledge point within the range of knowledge points.

[0109] The online education platform reads the knowledge point dependency path corresponding to the state transition segment and configures the pre-lecture knowledge state trajectory and post-lecture knowledge state trajectory of the historical teaching segment onto that knowledge point dependency path. For the knowledge point dependency path: ;

[0110] The online education platform reads the changes in the probability of mastery at each path position of the knowledge state trajectory before and after instruction. The pre-instruction state change indicates whether a history student's mastery probability along the knowledge point's dependency path showed a continuous decline before receiving instruction. The post-instruction state change indicates whether a history student's mastery probability within the same knowledge point range turned into a continuous increase after receiving instruction.

[0111] The online education platform reads the pre-lesson and post-lesson state change directions of each knowledge point along the knowledge point dependency path corresponding to the state transition segment. When a knowledge point on a continuous path position has a decreasing state change direction before lesson and an increasing state change direction after lesson, the platform extracts these continuous knowledge points as candidate knowledge point ranges. If there are more than two candidate knowledge point ranges, the platform first sorts them in descending order according to the number of knowledge points contained in the candidate knowledge point range; if the number of knowledge points is the same, it sorts them according to the path order of the starting point of the candidate knowledge point range in the state transition segment; the candidate knowledge point range that appears first in the list is read as the knowledge point range. This knowledge point range serves as an index for the teacher's teaching effectiveness parameter, indicating that the teacher's instruction can change the direction of the mastery probability within the corresponding knowledge point range from decreasing to increasing.

[0112] Using a linear function course as an example, if, before the lesson, students experienced a decrease in their mastery of linear function word problems, linear function graphs, and linear function expressions, but after the lesson, their mastery of all three knowledge points increased, then the online education platform extracts these three knowledge points as the scope of knowledge points. If, after the lesson, only the mastery of linear function word problems and linear function graphs increased, then the scope of knowledge points covers both linear function word problems and linear function graphs. This processing ensures that the teacher's teaching effectiveness parameter reflects the actual range of knowledge points improved by the teacher.

[0113] The calculation of the changes in the state of each knowledge point before and after instruction includes:

[0114] Calculate the teacher's answer time based on the pre-lecture answer period. Corresponding to the Key points in a historical lesson segment Changes in state before instruction ,in:

[0115] ;

[0116] In the formula, Indicates the first Key points in a historical lesson segment The probability of mastery at the start of the pre-lecture answering period. Indicates the first Key points in a historical lesson segment The probability of mastery at the end of the pre-lecture answering period;

[0117] The pre-lecture response period is the time before the start of the lecture and is related to the knowledge point dependency path corresponding to the state transition segment. The online education platform reads the knowledge point mastery probability at the start and end of this response period from the pre-lecture knowledge state trajectory. For teachers... Corresponding to the Key points from a historical lecture segment The platform is Indicates the start of the pre-lecture answering period, with This indicates the end point of the answering period before the lecture, from which the change in state before the lecture can be calculated. .

[0118] when When the value is negative, it indicates that the history student's understanding of the knowledge points was insufficient before the teacher's instruction. The probability of mastering the knowledge decreases. This value is used to characterize the magnitude of the state decline before instruction. If consecutive knowledge points in the same knowledge point dependency path all correspond to negative values ​​of state change before instruction, then the historical student experienced a continuous decline process before instruction, corresponding to the state transition segment of the target student.

[0119] Calculate the teacher's answer time according to the period after the lecture. Corresponding to the Key points in a historical lesson segment Post-lecture state change This forms the pre-teaching state change and post-teaching state change for each knowledge point within the scope of the stated knowledge points, where:

[0120] ;

[0121] In the formula, Indicates the first Key points in a historical lesson segment The probability of mastery at the start of the post-lecture answering period. Indicates the first Key points in a historical lesson segment The probability of mastery at the end of the answering period after the lecture.

[0122] The post-lecture response period is the time following the start of the lecture and is related to the knowledge point dependency path of the state transition segment. The online education platform reads the knowledge point mastery probability at the start and end of this response period from the post-lecture knowledge state trajectory. For teachers... Corresponding to the Key points from a historical lecture segment The platform is Indicates the start of the response period after the lecture, with This indicates the end of the response period after the lesson, from which the change in state after the lesson can be calculated. .

[0123] when When the value is positive, it indicates that history students have grasped the knowledge points after the teacher's instruction. The probability of mastering the knowledge increases. The platform saves the changes in the student's state before and after the lesson for each knowledge point within the knowledge point range, which are used to generate parameters for the teacher's teaching effectiveness.

[0124] As a supplementary explanation, if the first In the historical lesson segment, the probability of mastering the knowledge point "linear function graph" at the start of the pre-lesson question-and-answer session was: The probability of mastering the material at the end of the pre-lecture answering period is: The change in state before the lesson is: ;

[0125] If the probability of mastering this knowledge point is at the starting point of the answering period after the lesson is... The probability of mastering the material at the end of the answering period after the lesson is: Then the change in state after the lesson is: ;

[0126] This data indicates that the probability of mastering this knowledge point decreased before instruction and increased after instruction. This differential treatment ensures that the teacher's teaching effectiveness stems from changes in the direction and magnitude of their state before and after instruction, rather than from subjective evaluation information from the teacher.

[0127] Collect the changes in the teacher's state before and after the lesson within the scope of the knowledge points identified by the teacher's identifier. Take the median of the difference between the changes in the teacher's state before and after the lesson for the same teacher on the same knowledge point, and generate a teacher teaching effectiveness parameter indexed by the scope of the knowledge points.

[0128] Online education platforms compile historical lesson segments based on teacher identification. For the same teacher... and the same knowledge point The platform reads the teacher's state changes before and after each historical teaching segment within the knowledge point range, and calculates the state change difference for each historical teaching segment: ;

[0129] In the formula, Indicates the first Key points in a historical lesson segment The state change difference. The state change difference is used to represent the magnitude of the change in the state of a knowledge point from a declining state to an improving state before and after the lesson.

[0130] For teachers In the knowledge points The platform uses the median of the differences in the corresponding state changes from multiple historical teaching segments to generate teacher teaching effectiveness parameter components. ;

[0131] In the formula, Teacher In the knowledge points The teacher teaching effectiveness parameter components on the screen, Teacher In the knowledge points The corresponding collection of historical lecture segments, This represents the median function. The platform arranges the teacher teaching effectiveness parameter components corresponding to each knowledge point within the knowledge point range according to the knowledge point dependency path, generating teacher teaching effectiveness parameters indexed by the knowledge point range: ;

[0132] In the formula, to This represents knowledge points arranged continuously along the knowledge point dependency path within the knowledge point range. By taking the median of the differences in state changes for the same teacher on the same knowledge point, the impact of random fluctuations in a single historical teaching segment on the teacher's teaching effectiveness parameter can be reduced. This processing enables the teacher's teaching effectiveness parameter to stably represent the extent of improvement in the teacher's teaching when the state continuously declines within the corresponding knowledge point range.

[0133] S104: Based on the teacher teaching effectiveness parameters, perform a knowledge-point-by-knowledge-point offsetting calculation on the decrease in the mastery probability of each knowledge point in the state transition segment to obtain the candidate teacher matching value, and output the teacher matching result according to the candidate teacher matching value;

[0134] The online education platform applies teacher teaching effectiveness parameters to the state transition segments of target students. These state transition segments provide the decrease in the probability of mastery for each knowledge point. Teacher teaching effectiveness parameters provide the improvement in the teacher's instruction for the same knowledge point in sequence. The platform calculates the offsetting effect of teacher teaching effectiveness parameters on the decrease in the probability of mastery for each knowledge point and generates candidate teacher matching values ​​based on the remaining amount after offsetting. These candidate teacher matching values ​​are used to rank candidate teachers and output the teacher matching results.

[0135] The input data for this step includes target student state transition segments and teacher teaching effectiveness parameters. The output data is the teacher matching result. The teacher matching result can include candidate teacher identifiers, candidate teacher matching values, and candidate teacher ranking order. This process establishes a knowledge-point-by-knowledge-point correspondence between the final teacher recommendation result and the decrease in the target student's mastery probability on each knowledge point, avoiding the need to output recommendation results solely based on overall teacher evaluations or course tags.

[0136] Specifically, the process of obtaining the teacher matching results includes:

[0137] The target student state vector is formed according to the decrease in the mastery probability of each knowledge point in the state transition segment.

[0138] The online education platform reads the decrease in the mastery probability of each knowledge point in the state transition segment of the target student, and arranges them according to the order of the knowledge point dependency paths corresponding to the state transition segment to form the target student's state vector: ;

[0139] In the formula, Represents the knowledge points in the target student's state vector The component represents the decrease in the mastery probability of the corresponding knowledge point in the state transition segment. The target student's state vector is used to represent the magnitude of the decrease in the target student's current learning state along the knowledge point dependency path.

[0140] The teacher teaching effectiveness parameters are converted into a teacher effectiveness vector according to the knowledge point order of the target student state vector, and the teacher effectiveness vector is calculated according to the following formula. In the knowledge points On offset amount :

[0141] ;

[0142] In the formula, This represents the knowledge points in the target student's state vector. The amount, This represents the knowledge points in the teacher performance vector. The amount;

[0143] Online education platforms transform teacher teaching effectiveness parameters according to the knowledge point order of the target student's state vector. When a teacher teaching effectiveness parameter has a component at a knowledge point corresponding to the target student's state vector, that component is written into the teacher effectiveness vector. Other knowledge point positions are not included. This leads to the acquisition of teachers Corresponding teacher performance vector:

[0144] ;

[0145] In the formula, Teacher In the knowledge points The corresponding teacher performance vector component. This component represents the teacher's... The offsetting magnitude of the decrease in the probability of mastering the corresponding knowledge points in the target student's state vector.

[0146] Subsequently, the platform calculated the teacher's... In the knowledge points On offset amount This offset amount represents the portion of the teacher's performance vector that can cover the decrease in the target student's mastery probability for that knowledge point. The above formula determines the offset amount per knowledge point through absolute value calculation, which can avoid over-offsetting when the teacher's performance vector component exceeds the decrease in the target student's mastery probability.

[0147] The candidate teacher matching value is calculated using the following formula. The candidate teachers are sorted in ascending order according to their matching scores, and the teacher matching results are output; wherein:

[0148] ;

[0149] In the formula, the summation symbol traverses each knowledge point contained in the target student state vector.

[0150] The online education platform sums the remaining values ​​after deducting the corresponding offset from the decrease in the probability of mastering each knowledge point in the target student's state vector, and forms the candidate teacher matching value. The candidate teacher matching value represents the remaining decrease in the probability of mastery during the target student's state transition segment after offsetting the decrease in teacher teaching effectiveness parameters on a knowledge-point-by-knowledge-point basis. The online education platform sorts candidate teachers in ascending order according to their matching values ​​and outputs the teacher matching results.

[0151] As a supplementary explanation, if the target student's state vector is: ;

[0152] teacher The teacher effectiveness vector is: ;

[0153] Then the teacher The amount of offsetting across the three knowledge points is: ;

[0154] The candidate teacher matching value is: ;

[0155] If teacher The teacher effectiveness vector is: ;

[0156] Then the teacher The candidate teacher matching value is: ;

[0157] Online education platforms rank candidate teachers in ascending order of their matching scores. Arranged in the teachers Previously, the ranking result was determined by offsetting the decrease in the probability of mastering each knowledge point in the target student's state transition segment using the teacher's teaching effectiveness parameter. This ensured a direct correspondence between the teacher's recommendation and the target student's current knowledge state decline process.

[0158] When candidate teachers have the same matching value, the online education platform selects teachers with a higher value from the teacher performance vector. The components are sorted in descending order of quantity; if the quantity of components is the same, they are sorted in descending order of the number of historical teaching records of the teacher in the platform; if the number of records is the same, they are sorted in ascending order of teacher identifier. The teacher matching results are output according to the above sorting rules.

[0159] Example 2

[0160] like Figure 2 As shown in the example, the parts not detailed in this embodiment are as shown in Example 1. This embodiment discloses a teacher matching system based on big data analysis, including:

[0161] The tracking module 201 is used to input the target student's answer sequence into the knowledge tracking model to form the target student's knowledge state trajectory that records the mastery probability of each knowledge point at the time of answering, and to extract the state transition segment where the mastery probability continuously decreases from the subsequent knowledge point to the previous knowledge point along the knowledge point dependency path.

[0162] The retrieval module 202 is used to retrieve the teacher's historical teaching records in the descent direction represented by the state transition segment, and retain the historical teaching segments corresponding to the knowledge state trajectory before the teaching and the descent direction.

[0163] Evaluation module 203 is used to generate teacher teaching effectiveness parameters based on the knowledge state trajectory before and after the teaching of the historical teaching segment. The teacher teaching effectiveness parameters are indexed by the range of knowledge points whose mastery probability changes from continuous decline to continuous increase.

[0164] The matching module 204 is used to perform a knowledge-point-by-knowledge-point offsetting calculation on the decrease in the mastery probability of each knowledge point in the state transition segment based on the teacher teaching effectiveness parameters, to obtain the candidate teacher matching value, and to output the teacher matching result according to the candidate teacher matching value.

[0165] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A teacher matching method based on big data analysis, characterized in that, include: Input the target student's answer sequence into the knowledge tracking model to form the target student's knowledge state trajectory, which records the mastery probability of each knowledge point at the time of answering. Then, extract the state transition segments along the knowledge point dependency path, where the mastery probability continuously decreases from the subsequent knowledge point to the previous knowledge point. Retrieve the teacher's historical teaching records based on the descent direction represented by the state transition segment, and retain the historical teaching segments corresponding to the knowledge state trajectory before the teaching and the descent direction; Based on the knowledge state trajectory before and after the historical teaching segment, teacher teaching effectiveness parameters are generated, with the range of knowledge points whose mastery probability changes from continuous decline to continuous increase as the index. Based on the teacher teaching effectiveness parameters, the decrease in the mastery probability of each knowledge point in the state transition segment is calculated on a knowledge point-by-knowledge point basis to obtain the candidate teacher matching value, and the teacher matching result is output according to the candidate teacher matching value.

2. The method according to claim 1, characterized in that, The target student's answer sequence is indexed by the answer time, the answer time corresponds to the question identifier, the question identifier points to the knowledge point through the question knowledge point correspondence, and the knowledge point has a path position in the knowledge point dependency path.

3. The method according to claim 1, characterized in that, The process of extracting state transition segments along the knowledge point dependency path where the mastery probability continuously decreases from subsequent knowledge points to preceding knowledge points includes: Based on the knowledge point mastery probability read from the target student's knowledge state trajectory at the time of answering, a knowledge point mastery probability sequence is formed. Configure the knowledge point mastery probability sequence to the knowledge point dependency path corresponding to the target student's answer sequence to obtain the path mastery probability sequence; In the path mastery probability sequence, a continuously descending trajectory segment is extracted along the direction from the subsequent knowledge point to the preceding knowledge point to form the state transition segment.

4. The method according to claim 1, characterized in that, The teacher's historical teaching record is indexed by the start time of the teaching session. The start time of the teaching session corresponds to the teacher's identifier, the teacher's identifier corresponds to the teaching content sequence, the teaching content sequence corresponds to the historical student answer sequence, and the historical student answer sequence includes answer records before and after the start time of the teaching session.

5. The method according to claim 1, characterized in that, The historical teaching segments that retain the pre-teaching knowledge state trajectory and correspond to the descent direction include: Using the start time of the lecture in the teacher's historical lecture record as the boundary, the sequence of historical student answers before the start time of the lecture is input into the knowledge tracking model to form a knowledge state trajectory before the lecture. Configure the state transition segment and the pre-lecture knowledge state trajectory to the same knowledge point dependency path, and calculate the path difference between the pre-lecture knowledge state trajectory and the state transition segment; Using teacher identifiers as units, the corresponding teacher's historical teaching records are sorted in ascending order according to the path difference, and the teacher's historical teaching record at the top of the list is read and saved as the historical teaching segment.

6. The method according to claim 5, characterized in that, Calculate the path difference between the pre-lecture knowledge state trajectory and the state transition segment, including: Based on the knowledge point dependency path of the state transition segment, the decrease in the mastery probability of each knowledge point within the state transition segment is read to form a sequence of changes in the target student's mastery probability. and read the first The decrease in the probability of mastery of the same knowledge point within the knowledge state trajectory before the lesson is taught forms a historical sequence of changes in students' mastery probability. ; Calculate the first according to the following formula. Path difference between the pre-lesson knowledge state trajectory and the state transition segment : , In the formula, This represents the first difference of the probability change sequence mastered by the target student. This represents the first difference of the sequence of historical student mastery probability changes. express Norm.

7. The method according to claim 1, characterized in that, Based on the knowledge state trajectory before and after the aforementioned historical teaching segments, teacher teaching effectiveness parameters are generated, including: Input the sequence of historical students' answers after the start of the lesson in the historical teaching segment into the knowledge tracking model to form a knowledge state trajectory after the lesson. Align the knowledge state trajectory before and after the teaching with the knowledge state trajectory after the teaching according to the knowledge point dependency path corresponding to the state transition segment, extract the range of knowledge points whose mastery probability changes from continuous decline to continuous increase, and calculate the change in state before and after the teaching for each knowledge point within the range of knowledge points. Collect the changes in the teacher's state before and after the lesson within the scope of the knowledge points identified by the teacher's identifier. Take the median of the difference between the changes in the teacher's state before and after the lesson for the same teacher on the same knowledge point, and generate a teacher teaching effectiveness parameter indexed by the scope of the knowledge points.

8. The method according to claim 7, characterized in that, Calculate the changes in the state of each knowledge point before and after instruction, including: Calculate the teacher's answer time based on the pre-lecture answer period. Corresponding to the Key points in a historical lesson segment Changes in state before instruction ,in: , In the formula, Indicates the first Key points in a historical lesson segment The probability of mastery at the start of the pre-lecture answering period. Indicates the first Key points in a historical lesson segment The probability of mastery at the end of the pre-lecture answering period; Calculate the teacher's response time based on the period after the lecture. Corresponding to the Key points in a historical lesson segment Post-lecture state change This forms the pre-teaching state change and post-teaching state change for each knowledge point within the scope of the stated knowledge points, where: , In the formula, Indicates the first Key points in a historical lesson segment The probability of mastery at the start of the post-lecture answering period. Indicates the first Key points in a historical lesson segment The probability of mastery at the end of the answering period after the lecture.

9. The method according to claim 1, characterized in that, The process of obtaining the teacher matching results includes: The target student state vector is formed according to the decrease in the mastery probability of each knowledge point in the state transition segment. The teacher teaching effectiveness parameters are converted into a teacher effectiveness vector according to the knowledge point order of the target student state vector, and the teacher effectiveness vector is calculated according to the following formula. In the knowledge points On offset amount : , In the formula, This represents the knowledge points in the target student's state vector. The amount, This represents the knowledge points in the teacher performance vector. The amount; The candidate teacher matching value is calculated using the following formula. The candidate teachers are sorted in ascending order according to their matching scores, and the teacher matching results are output; wherein: , In the formula, the summation symbol traverses each knowledge point contained in the target student state vector.

10. A teacher matching system based on big data analysis, implemented using the method described in any one of claims 1-9, characterized in that, include: The tracking module is used to input the target student's answer sequence into the knowledge tracking model, form the target student's knowledge state trajectory by recording the mastery probability of each knowledge point at the time of answering, and extract the state transition segment where the mastery probability continuously decreases from the subsequent knowledge point to the previous knowledge point along the knowledge point dependency path. The retrieval module is used to retrieve the teacher's historical teaching records in the descent direction represented by the state transition segment, and retain the historical teaching segments corresponding to the knowledge state trajectory before the teaching and the descent direction. The evaluation module is used to generate teacher teaching effectiveness parameters based on the knowledge status trajectory before and after the teaching of the historical teaching segments. The teacher teaching effectiveness parameters are indexed by the range of knowledge points whose mastery probability changes from continuous decline to continuous increase. The matching module is used to perform a knowledge-point-by-knowledge-point offsetting calculation on the decrease in the mastery probability of each knowledge point in the state transition segment based on the teacher teaching effectiveness parameters, obtain the candidate teacher matching value, and output the teacher matching result according to the candidate teacher matching value.

Images