CN110807469A - Knowledge tracking method and system integrating long-time memory and short-time memory with Bayesian network - Google Patents
Knowledge tracking method and system integrating long-time memory and short-time memory with Bayesian network Download PDFInfo
- Publication number
- CN110807469A CN110807469A CN201910889394.3A CN201910889394A CN110807469A CN 110807469 A CN110807469 A CN 110807469A CN 201910889394 A CN201910889394 A CN 201910889394A CN 110807469 A CN110807469 A CN 110807469A
- Authority
- CN
- China
- Prior art keywords
- current time
- learner
- knowledge
- long
- neural network
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/29—Graphical models, e.g. Bayesian networks
- G06F18/295—Markov models or related models, e.g. semi-Markov models; Markov random fields; Networks embedding Markov models
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/36—Creation of semantic tools, e.g. ontology or thesauri
- G06F16/367—Ontology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Computational Linguistics (AREA)
- Evolutionary Computation (AREA)
- Computing Systems (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Animal Behavior & Ethology (AREA)
- Databases & Information Systems (AREA)
- Electrically Operated Instructional Devices (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a knowledge tracking method and a knowledge tracking system integrating long-time memory and short-time memory and a Bayesian network, which calculate the parameter group of a Bayesian knowledge tracking model of a knowledge component corresponding to the current time sequence by establishing a cognitive data set comprising the time sequence and a long-time and short-time memory neural network, thereby utilizing the Bayesian knowledge tracking model to calculate the probability predicted value of correct answer of a learner to the question of the current time sequence, and obtaining the loss function of the long-time and short-time memory neural network model corresponding to the current time sequence by comparing the true value of the correct answer of the question of the current time sequence in the cognitive data set, thereby obtaining the optimized value of a weight parameter matrix and the optimized value of a deviation parameter matrix; traversing all time sequences of the cognitive data set to obtain an optimized value of a long-term and short-term memory neural network model weight parameter matrix and an optimal value of a deviation parameter matrix; therefore, the learner can predict the cognitive state of the test question to be tested, and the learning path planning and/or the construction of the knowledge map of the learner can be carried out according to the prediction of the cognitive state of the learner.
Description
Technical Field
The invention belongs to the field of knowledge tracking, and particularly relates to a knowledge tracking method and system integrating long-time memory and short-time memory with a Bayesian network.
Background
Knowledge tracking is to model the knowledge learning state of a learner so as to track the mastery degree of the learner on a knowledge point, and further predict the performance of the learner in the next answering. Knowledge tracking can capture the current real requirements of a learner, and is a core task in modeling of the learner. However, due to the diversity of knowledge and the complexity of human brain, the learning process of human is complex and variable, which is the reason why the knowledge tracking is very difficult.
The field of knowledge tracking has two classical solution models. One of the classical models is the Hidden Markov Model (Hidden Markov Model), represented by Bayesian Knowledge Tracking (BKT), which models two learning states of the learner for the Knowledge Component (Knowledge Component), i.e. (not mastered, mastered), and updates the learner's current Knowledge mastery with the prior probabilities associated with whether the learner correctly answers the question and the four parameter values that play an important role. Bayesian knowledge tracking is a first-order hidden markov model, i.e. the update of the current state depends on the previous knowledge grasping state. The model has the advantages that the representation of the learning state is clear, and the dependence existing in the learning process is simulated to a certain extent; however, it is questionable that the four parameter values that play a critical role in the model are randomly selected and do not utilize any information of the learner's data. Another classical Model is a Deep learning Model (Deep learning Model), represented by Deep Knowledge Tracking (DKT), which uses a Neural Network with a "memory" function, i.e., a Recurrent Neural Network (RNN), to simulate the learning process of a learner, thereby achieving a good prediction result. The model has the advantages that the recurrent neural network is easy to train and obtain better model prediction effect; the disadvantage is that the neural network always has the problem of 'black box', and clear semantic interpretation can not be provided for the model.
Disclosure of Invention
Aiming at the defects or improvement requirements of the prior art, the invention provides a knowledge tracking method based on fusion cognition calculation, a Bayes knowledge tracking method based on deep learning and a system thereof.
In order to achieve the above object, according to an aspect of the present invention, there is provided a knowledge tracking method fusing long-time memory and short-time memory with a bayesian network, comprising the following specific steps:
s1, acquiring a statistical cognitive data set of a learner on a plurality of knowledge component samples, wherein the cognitive data set comprises a plurality of time sequences, and one time sequence comprises questions and question response conditions of the learner on one knowledge component;
s2, establishing a long-short term memory neural network model, inputting the long-short term memory neural network model into a current time sequence in the cognitive data, outputting a parameter group of a Bayesian knowledge tracking model of a knowledge component corresponding to the current time sequence, setting initial parameter values of the long-short term memory neural network model, and setting initial values of a weight parameter matrix and an offset parameter matrix of the long-short term memory neural network model; inputting the current time sequence into a long-short term memory neural network model to obtain a parameter group of a knowledge component Bayesian knowledge tracking model corresponding to the current time sequence;
s3, inputting a parameter group of a Bayesian knowledge tracking model of a knowledge component corresponding to the current time sequence into the Bayesian knowledge tracking model, calculating a probability predicted value that the learner answers the question of the current time sequence correctly, traversing all the questions of the current time sequence to obtain the probability predicted value that the learner answers the question of the current time sequence correctly, comparing the true value of whether the learner answers the question of the current time sequence in the cognitive data set is correct or not, and obtaining a long-short term memory neural network model loss function corresponding to the current time sequence, wherein the independent variable of the loss function comprises a weight parameter matrix and a deviation parameter matrix of the long-short term memory neural network model; minimizing a long-short term memory neural network model loss function corresponding to the current time sequence to obtain an optimized value of a weight parameter matrix and an optimized value of a deviation parameter matrix; traversing all time sequences of the cognitive data set to obtain an optimized value of a long-term and short-term memory neural network model weight parameter matrix and an optimal value of a deviation parameter matrix;
s4, calculating by using the optimized long-short term memory neural network model to obtain parameter groups of the Bayes knowledge tracking model corresponding to the learner and all knowledge components one to one; and predicting the cognitive state of the test question by the learner by using the Bayesian knowledge tracking model corresponding to all the optimized knowledge components, and planning the learning path of the learner and/or constructing a knowledge map according to the prediction of the cognitive state.
As a further improvement of the invention, the initial parameter values of the long-short term memory neural network model comprise input and output dimensions, the number of network layers, the number of neurons in each layer and the sequence times.
As a further development of the invention, the cognitive data set comprises n time series D1,D2,...,DnThe number of questions corresponding to each time sequence is I1,I2,...,InThe current time sequence is Dc。
As a further improvement of the invention, the current time series corresponds to the s learner and the r knowledge component, and the current time series DcThe parameter set of the corresponding knowledge component comprises:
prs(L0): probability estimation of the s-th learner with the initial learning state for the r-th knowledge component;
prs(T): the s learner learns a probability estimate for the r knowledge component;
prs(G) the method comprises the following steps Estimating the probability of guessing when the nth learner does not master the nth knowledge component;
prs(S): and (4) carrying out probability estimation of question and mistake after the s-th learner grasps the r-th knowledge component.
As a further improvement of the invention, with the current time series DcThe corresponding long-short term memory neural network model loss function is specifically as follows:
wherein, l is cross entropy, w is weight parameter matrix of long-short term memory neural network model, b is deviation parameter matrix of long-short term memory neural network model, r represents number of knowledge component corresponding to current time sequence, s represents number of learner corresponding to current time sequence, IcRepresenting the current time series DcNumber of questions of, p (C)is) Representing the answer correct probability prediction value of the ith learner on the current time sequenceiAnd the real value represents whether the answer of the ith question of the current time sequence in the cognitive data set is correct or not.
As a further improvement of the invention, the answer correct probability predicted value p (C) of the ith question of the current time sequence is used by the s-th learneris) The expression of (a) is:
p(Cis)＝prs(Li-1)*[1-prs(S)]+[1-prs(Li-1)]*prs(G)
wherein p isrs(Li-1) Probability estimates for the r knowledge component after the answer for the ith learner for the ith-1 question of the current time series.
As a further improvement of the invention, the s-th learner is rightProbability estimate p for learning the r-th knowledge component after answer of the ith question of the previous time seriesrs(Li) Expressed as:
prs(Li)＝prs(Li-1|evidencei)+[1-prs(Li-1|evidencei)]*prs(T)
wherein the content of the first and second substances,
wherein p isrs(Li) Probability estimation of learning the r knowledge component after answer of the ith question of the current time series for the s learner, evaluationi1 indicates that the ith question of the s learner answering the current time series is correct, and evidencei0 indicates that the ith learner answered the ith question error for the current time series.
As a further improvement of the invention, the loss function of the long-short term memory neural network model corresponding to the current time sequence is minimized by utilizing a stochastic gradient descent method to obtain the optimized value of the weight parameter matrix and the optimized value of the deviation parameter matrix, specifically, the current time sequence D is established according to the loss functioncThe corresponding cost function is used as a function of the cost,
the partial derivatives of w and b are separately derived to obtain the gradients dw and db, i.e.
Weight parameter matrix w of updated long-short term memory neural network model*And a deviation parameter matrix b*Are respectively as
Wherein, λ is the learning rate of the long-short term memory neural network model.
To achieve the above object, according to another aspect of the present invention, there is provided a knowledge tracking system fusing long-time memory and bayesian network, which includes a memory and a processor, wherein the memory stores instructions, and the instructions are executed by the processor to achieve the above method.
Generally, compared with the prior art, the above technical solution conceived by the present invention has the following beneficial effects:
the invention provides a knowledge tracking method and a knowledge tracking system integrating long-time memory and short-time memory and a Bayesian network, which calculate the parameter group of a Bayesian knowledge tracking model of a knowledge component corresponding to the current time sequence by establishing a cognitive data set comprising the time sequence and a long-time and short-time memory neural network, thereby utilizing the Bayesian knowledge tracking model to calculate the probability predicted value of correct answer of a learner to the question of the current time sequence, and obtaining the loss function of the long-time and short-time memory neural network model corresponding to the current time sequence by comparing the true value of the correct answer of the question of the current time sequence in the cognitive data set, thereby obtaining the optimized value of a weight parameter matrix and the optimized value of a deviation parameter matrix; traversing all time sequences of the cognitive data set to obtain an optimized value of a long-term and short-term memory neural network model weight parameter matrix and an optimal value of a deviation parameter matrix; therefore, the method realizes the prediction of the learner on the cognitive state of the test question to be tested, carries out the learning path planning and/or the knowledge map construction of the learner according to the prediction of the cognitive state of the learner, is particularly suitable for a self-adaptive education system, and can be applied to the technology for carrying out accurate pushing, the path planning of the learner or the construction of the knowledge map.
The knowledge tracking method and system integrating long-time and short-time memory and the Bayesian network, provided by the invention, utilize the 'memory' characteristic of the long-time and short-time memory network in processing time series data, further inherit the characteristics implicit in a learner answer sequence, accord with the learning rule of a learner based on a knowledge component when answering, and the obtained parameter values accord with the learning rule of the learner better, are more suitable for predicting whether the learner correctly answers the question or not, and are also more suitable for calculating the probability estimation value of the knowledge component mastered by the learner after answering the question.
The invention provides a knowledge tracking method and a knowledge tracking system integrating long-time memory and Bayesian network, which are used for calculating a predicted value p (C) of correct answer probability of the ith learner on the current time sequenceis) The learning process is based on the updated value of the knowledge mastering state after the last answer, and the learning process of the learner is met, namely the dependency existing between the front learning state and the back learning state in the learning process can be truly simulated to a certain extent.
The invention provides a knowledge tracking method and a knowledge tracking system integrating long-time memory and short-time memory and Bayesian network, which utilize a hidden Markov model to obtain a long-short term memory neural network model loss function, utilize a stochastic gradient descent method to minimize the long-short term memory neural network model loss function corresponding to a current time sequence, obtain an optimized value of a weight parameter matrix and an optimized value of a deviation parameter matrix, and combine a depth model and the hidden Markov model, thereby not only developing the characteristic that more data characteristics can be developed and utilized when the depth model processes data, but also developing the interpretability when the hidden Markov model calculates an answer prediction result, and solving the question point existing in the Bayesian knowledge tracking model.
Drawings
FIG. 1 is a schematic diagram of a knowledge tracking method of a fused long-term memory and Bayesian network according to an embodiment of the present invention;
FIG. 2 is a schematic structural diagram of a long-term and short-term memory neural network model according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
In addition, the technical features involved in the embodiments of the present invention described below may be combined with each other as long as they do not conflict with each other. The present invention will be described in further detail with reference to specific embodiments.
Fig. 1 is a schematic diagram of a knowledge tracking method of a fused long-time memory and bayesian network according to an embodiment of the present invention. As shown in fig. 1, the method comprises the steps of:
s1, acquiring a statistical cognitive data set of a learner on a plurality of knowledge component samples, wherein the cognitive data set comprises a plurality of time sequences, and one time sequence comprises a test question and a test question answer condition of the learner on one knowledge component; specifically, a knowledge component corresponds to a plurality of questions related to the knowledge component, the learner generates an answer sequence of the knowledge component for the ordered answer conditions of the plurality of questions related to the knowledge component, and the cognitive data set comprises n time sequences D1,D2,...DnThe number of questions corresponding to each time sequence is I1,I2,...In；
S2, establishing a long-short term memory neural network model, wherein the input of the long-short term memory neural network model is a cognitive data set, the output of the long-short term memory neural network model is a parameter set of a Bayesian knowledge tracking model of each knowledge component, the initial parameter value of the network model is set, and the initial value of a weight parameter matrix and the initial value of a deviation parameter matrix of the long-short term memory neural network model are set; inputting the current time sequence into the long-short term memory neural network model to obtain a parameter group of the knowledge component corresponding to the current time sequence; specifically, the initial parameter values include input/output dimensions of the long-short term memory neural network model, the number of network layers, the number of neurons in each layer, and the sequence times, the current time sequence corresponds to the s-th learner and the r-th knowledge component, and the parameter set includes:
prs(L0): probability estimation of the s-th learner with the initial learning state for the r-th knowledge component;
prs(T): the s learner learns a probability estimate for the r knowledge component;
prs(G) the method comprises the following steps Estimating the probability of guessing when the nth learner does not master the nth knowledge component;
prs(S): the probability estimation of question and error is carried out after the s learner masters the r knowledge component;
specifically, parameter values of the long-term and short-term memory neural network model are set according to data of the cognitive data set; FIG. 2 is a schematic structural diagram of a long-term and short-term memory neural network model according to an embodiment of the present invention. As shown in fig. 2, the input dimension of the long-short term memory neural network model is set to k dimensions, the output dimension is set to 4 dimensions (corresponding to the parameter set of the knowledge component), the number of network layers is set to m, the number of neurons in each layer is d, and the number of times of network training is epoch. Of course, the setting in fig. 2 is only an example, and the parameter values of the long-short term memory neural network model may be set according to specific data of the cognitive data set.
S3, inputting the parameter group of the knowledge component corresponding to the current time sequence into a Bayesian knowledge tracking model, calculating a probability prediction value that the learner answers the question of the current time sequence correctly, traversing all the questions in the time sequence corresponding to the knowledge component to obtain the probability prediction value that the learner answers the question of the current time sequence correctly, comparing the true value of whether the answer of the question of the current time sequence in the cognitive data set is correct or not, and obtaining a long-short term memory neural network model loss function corresponding to the current time sequence, wherein the independent variable of the loss function comprises a weight parameter matrix and a deviation parameter matrix of the long-short term memory neural network model; minimizing a long-short term memory neural network model loss function corresponding to the current time sequence to obtain an optimized value of a weight parameter matrix and an optimized value of a deviation parameter matrix; traversing all time sequences of the cognitive data set to obtain an optimized value of a long-term and short-term memory neural network model weight parameter matrix and an optimal value of a deviation parameter matrix;
in particular, the amount of the solvent to be used,
whether the learner will return or not is calculated by the Bayesian knowledge tracking modelInputting the predicted value of correct answer and the real answer result of learner into loss function to obtain current time sequence DcThe corresponding long-short term memory neural network model loss function specifically comprises the following steps:
wherein, l is cross entropy, w is weight parameter matrix of long-short term memory neural network model, b is deviation parameter matrix of long-short term memory neural network model, r represents number of knowledge component corresponding to current time sequence, s represents number of learner corresponding to current time sequence, IcRepresenting the current time series DcNumber of questions of, p (C)is) Representing the answer correct probability prediction value of the ith learner on the current time sequenceiRepresenting the real value of whether the answer of the ith question of the current time sequence in the cognitive data set is correct or not;
p(Cis) The specific calculation process is as follows:
p(Cis)＝prs(Li-1)*[1-prs(S)]+[1-prs(Li-1)]*prs(G)
wherein p isrs(Li-1) Probability estimation for the r knowledge component after answer to the i-1 th question of the current time series for the s learner, prs(S) estimating probability of question and mistake for the S learner after mastering the r knowledge component, prs(G) Estimating the probability of guessing when the nth learner does not master the nth knowledge component;
prs(Li) The specific calculation process is as follows:
prs(Li)＝prs(Li-1|evidencei)+[1-prs(Li-1|evidencei)]*prs(T)
wherein the content of the first and second substances,
wherein p isrs(Li) Probability estimation of learning the r knowledge component after answer of the ith question of the current time series for the s learner, evaluationi1 indicates that the ith question of the s learner answering the current time series is correct, and evidencei0 denotes the ith question error, p, for the s-th learner to answer the current time seriesrs(T) probability estimates for the nth knowledge component learned by the s learner;
as a preferred embodiment, the method includes minimizing a loss function of the long-short term memory neural network model corresponding to the current time sequence by using a stochastic gradient descent method to obtain an optimized value of the weight parameter matrix and an optimized value of the deviation parameter matrix, specifically, establishing the current time sequence D according to the loss functioncThe corresponding cost function is used as a function of the cost,
the partial derivatives of w and b are separately derived to obtain the gradients dw and db, i.e.
Weight parameter matrix w of updated long-short term memory neural network model*And a deviation parameter matrix b*Are respectively as
Wherein, λ is the learning rate of the long-short term memory neural network model;
of course, the random gradient descent method is only an example, and the weight parameter matrix and the deviation parameter matrix may be updated according to a batch gradient descent method, a small batch gradient descent method, a momentum method, Adam, RMSprop, and adagard, and may be adjusted accordingly according to the calculation requirement.
S4, calculating by using the optimized long-short term memory neural network model to obtain parameter groups of the Bayes knowledge tracking model corresponding to the learner and all knowledge components one to one; and predicting the cognitive state of the test question by the learner by using the Bayesian knowledge tracking model corresponding to all the optimized knowledge components, and planning the learning path of the learner and/or constructing a knowledge map according to the predicted state of the learner.
The knowledge tracking system fusing the long-time memory and the short-time memory and the Bayesian network comprises a memory and a processor, wherein instructions are stored in the memory, and when the instructions are executed by the processor, the method is realized.
Table 1 shows the results of the knowledge tracking method of the present invention and the prediction of the learner's answer by the prior art method. As shown in Table 1, experiments are performed on four data sets, and the cognitive data of the four data sets, namely, the attributes 2009-2010, the attributes 2014-2015, the attributes-5 and the statistics 2010, are used for corresponding prediction, wherein, BKT represents a Bayesian knowledge tracking method*The best experimental result of the Bayesian knowledge tracking method in other researches is shown, DKT shows a depth knowledge tracking method, LSTM-BKT shows the knowledge tracking method of the invention, and the measurement index of the experimental result is Area Under Curve (AUC), which is defined as the area enclosed by the coordinate axes under ROC Curve. As can be seen from the table, the prediction result of the technical scheme of the embodiment of the present invention is superior to that of the prior art.
Table 1 shows the results of the knowledge tracking method of the present invention and the prediction of the learner's answer in the prior art method
| Assist2009 | Assist2015 | Stimulated-5 | Statics2010 | |
| BKT | 0.67 | 0.64 | 0.54 | 0.73 |
| BKT* | 0.79 | - | 0.76 | 0.75 |
| DKT | 0.81 | 0.70 | 0.75 | 0.74 |
| LSTM+BKT | 0.82 | 0.73 | 0.74 | 0.79 |
It will be understood by those skilled in the art that the foregoing is only a preferred embodiment of the present invention, and is not intended to limit the invention, and that any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the scope of the present invention.
Claims (9)
1. A knowledge tracking method integrating long-time memory and short-time memory and a Bayesian network is characterized by comprising the following specific steps:
s1, acquiring a statistical cognitive data set of a learner on a plurality of knowledge component samples, wherein the cognitive data set comprises a plurality of time sequences, and one time sequence comprises questions and question response conditions of the learner on one knowledge component;
s2, establishing a long-short term memory neural network model, wherein the input of the long-short term memory neural network model is a current time sequence in the cognitive data, the output of the long-short term memory neural network model is a parameter group of a Bayes knowledge tracking model of a knowledge component corresponding to the current time sequence, the initial parameter value of the long-short term memory neural network model is set, and the initial value of a weight parameter matrix and the initial value of a deviation parameter matrix of the long-short term memory neural network model are set; inputting the current time sequence into a long-short term memory neural network model to obtain a parameter group of a knowledge component Bayesian knowledge tracking model corresponding to the current time sequence;
s3, inputting a parameter group of a Bayesian knowledge tracking model of a knowledge component corresponding to the current time sequence into the Bayesian knowledge tracking model, calculating a probability predicted value that the learner answers the question of the current time sequence correctly, traversing all the questions of the current time sequence to obtain the probability predicted value that the learner answers the question of the current time sequence correctly, comparing the true value of whether the learner answers the question of the current time sequence in the cognitive data set is correct or not, and obtaining a long-short term memory neural network model loss function corresponding to the current time sequence, wherein the independent variable of the loss function comprises a weight parameter matrix and a deviation parameter matrix of the long-short term memory neural network model; minimizing a long-short term memory neural network model loss function corresponding to the current time sequence to obtain an optimized value of a weight parameter matrix and an optimized value of a deviation parameter matrix; traversing all time sequences of the cognitive data set to obtain an optimized value of a long-term and short-term memory neural network model weight parameter matrix and an optimal value of a deviation parameter matrix;
s4, calculating by using the optimized long-short term memory neural network model to obtain parameter groups of the Bayes knowledge tracking model corresponding to the learner and all knowledge components one to one; and predicting the cognitive state of the test question by the learner by using the Bayesian knowledge tracking model corresponding to all the optimized knowledge components, and planning the learning path of the learner and/or constructing a knowledge map according to the prediction of the cognitive state.
2. The method of claim 1, wherein the initial parameter values of the long-term and short-term memory neural network model include input-output dimensions, number of network layers, number of neurons in each layer, and sequence number.
3. The method of claim 1, wherein the cognitive data set comprises n time series D1,D2,...DnThe number of questions corresponding to each time sequence is I1,I2,...InThe current time sequence is Dc。
4. The method of claim 3, wherein the current time series corresponds to the s-th learner and the r-th knowledge component, and the current time series DcThe parameter set of the corresponding knowledge component comprises:
prs(L0): probability estimation of the s-th learner with the initial learning state for the r-th knowledge component;
prs(T): the s learner learns a probability estimate for the r knowledge component;
prs(G) the method comprises the following steps Estimating the probability of guessing when the nth learner does not master the nth knowledge component;
prs(S): first, theAnd (4) carrying out probability estimation of question and mistake after the s learners master the r-th knowledge component.
5. The method of claim 4, wherein the knowledge tracking method is combined with a current time series DcThe corresponding long-short term memory neural network model loss function is specifically as follows:
wherein, l is cross entropy, w is weight parameter matrix of long-short term memory neural network model, b is deviation parameter matrix of long-short term memory neural network model, r represents number of knowledge component corresponding to current time sequence, s represents number of learner corresponding to current time sequence, IcRepresenting the current time series DcNumber of questions of, p (C)is) Representing the answer correct probability prediction value of the ith learner on the current time sequenceiAnd the real value represents whether the answer of the ith question of the current time sequence in the cognitive data set is correct or not.
6. The knowledge tracking method combining long-time and short-time memory and Bayesian network as claimed in claim 5, wherein the answer correct probability prediction value p (C) for the ith question of the current time series is applied to the s-th learneris) The expression of (a) is:
p(Cis)＝prs(Li-1)*[1-prs(S)]+[1-prs(Li-1)]*prs(G)
wherein p isrs(Li-1) Probability estimates for the r knowledge component after the answer for the ith learner for the ith-1 question of the current time series.
7. The method of claim 6 wherein the s-th learner is interested in the current time seriesAfter answering of the ith question, the probability estimate p of the r-th knowledge componentrs(Li) Expressed as:
prs(Li)＝prs(Li-1|evidencei)+[1-prs(Li-1|evidencei)]*prs(T)
wherein the content of the first and second substances,
wherein p isrs(Li) Probability estimation of learning the r knowledge component after answer of the ith question of the current time series for the s learner, evaluationi1 indicates that the ith question of the s learner answering the current time series is correct, and evidencei0 indicates that the ith learner answered the ith question error for the current time series.
8. The knowledge tracking method integrating long-time and short-time memory and Bayesian network as claimed in any one of claims 5-7, wherein a random gradient descent method is used to minimize a long-short term memory neural network model loss function corresponding to a current time sequence, so as to obtain an optimized value of a weight parameter matrix and an optimized value of a bias parameter matrix, specifically, a current time sequence D is established according to the loss functioncThe corresponding cost function is used as a function of the cost,
the partial derivatives of w and b are separately derived to obtain the gradients dw and db, i.e.
Weight parameter matrix of updated long-short term memory neural network modelw*And a deviation parameter matrix b*Are respectively as
Wherein, λ is the learning rate of the long-short term memory neural network model.
9. A knowledge tracking system incorporating long-short term memory and a bayesian network, comprising a memory and a processor, the memory having stored therein instructions which, when executed by the processor, carry out the method according to any of claims 1-8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910889394.3A CN110807469B (en) | 2019-09-19 | 2019-09-19 | Knowledge tracking method and system integrating long-time memory and short-time memory with Bayesian network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910889394.3A CN110807469B (en) | 2019-09-19 | 2019-09-19 | Knowledge tracking method and system integrating long-time memory and short-time memory with Bayesian network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110807469A true CN110807469A (en) | 2020-02-18 |
| CN110807469B CN110807469B (en) | 2020-09-11 |
Family
ID=69487666
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910889394.3A Active CN110807469B (en) | 2019-09-19 | 2019-09-19 | Knowledge tracking method and system integrating long-time memory and short-time memory with Bayesian network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110807469B (en) |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112256858A (en) * | 2020-10-09 | 2021-01-22 | 华中师范大学 | Double-convolution knowledge tracking method and system fusing question mode and answer result |
| CN112766513A (en) * | 2021-01-29 | 2021-05-07 | 华中师范大学 | Knowledge tracking method and system with memory cooperation |
| CN112785039A (en) * | 2020-12-31 | 2021-05-11 | 科大讯飞股份有限公司 | Test question answering score prediction method and related device |
| CN112818100A (en) * | 2021-01-29 | 2021-05-18 | 华中师范大学 | Knowledge tracking method and system fusing question difficulty |
| CN112949935A (en) * | 2021-03-26 | 2021-06-11 | 华中师范大学 | Knowledge tracking method and system fusing student knowledge point question interaction information |
| CN112949929A (en) * | 2021-03-15 | 2021-06-11 | 华中师范大学 | Knowledge tracking method and system based on collaborative embedded enhanced topic representation |
| CN113283585A (en) * | 2021-05-24 | 2021-08-20 | 北京大学 | Knowledge tracking method and system based on self-attention mechanism |
| CN113361685A (en) * | 2021-05-25 | 2021-09-07 | 华中师范大学 | Knowledge tracking method and system based on learner knowledge state evolution expression |
| CN113704235A (en) * | 2021-08-03 | 2021-11-26 | 桂林电子科技大学 | Depth knowledge tracking model based on self-attention mechanism |
| CN114723590A (en) * | 2022-03-30 | 2022-07-08 | 华南师范大学 | Group-oriented knowledge tracking method, system, device and storage medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20160155346A1 (en) * | 2013-07-16 | 2016-06-02 | Jun Wang | System and method for learning recommendation simulation |
| CN107633111A (en) * | 2017-08-14 | 2018-01-26 | 苏州爱德克思智能科技有限公司 | A kind of situation simulation support system towards mathematical education |
| CN108921349A (en) * | 2018-07-04 | 2018-11-30 | 北京希子教育科技有限公司 | A method of topic errors present is done based on Bayesian network forecasting |
| CN109598995A (en) * | 2019-01-08 | 2019-04-09 | 上海健坤教育科技有限公司 | Intelligent tutoring system based on Bayes's knowledge trace model |
| CN110032651A (en) * | 2019-04-18 | 2019-07-19 | 江苏师范大学 | A kind of constructing method of the individualized learning characteristic model of knowledge based map |
-
2019
- 2019-09-19 CN CN201910889394.3A patent/CN110807469B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20160155346A1 (en) * | 2013-07-16 | 2016-06-02 | Jun Wang | System and method for learning recommendation simulation |
| CN107633111A (en) * | 2017-08-14 | 2018-01-26 | 苏州爱德克思智能科技有限公司 | A kind of situation simulation support system towards mathematical education |
| CN108921349A (en) * | 2018-07-04 | 2018-11-30 | 北京希子教育科技有限公司 | A method of topic errors present is done based on Bayesian network forecasting |
| CN109598995A (en) * | 2019-01-08 | 2019-04-09 | 上海健坤教育科技有限公司 | Intelligent tutoring system based on Bayes's knowledge trace model |
| CN110032651A (en) * | 2019-04-18 | 2019-07-19 | 江苏师范大学 | A kind of constructing method of the individualized learning characteristic model of knowledge based map |
Non-Patent Citations (2)
| Title |
|---|
| PENGHE CHEN ET AL: "Prerequisite-Driven Deep Knowledge Tracing", 《2018 IEEE INTERNATIONAL CONFERENCE ON DATA MINING》 * |
| 徐墨客 等: "多知识点知识追踪模型与可视化研究", 《学习环境与资源》 * |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112256858A (en) * | 2020-10-09 | 2021-01-22 | 华中师范大学 | Double-convolution knowledge tracking method and system fusing question mode and answer result |
| CN112785039A (en) * | 2020-12-31 | 2021-05-11 | 科大讯飞股份有限公司 | Test question answering score prediction method and related device |
| CN112785039B (en) * | 2020-12-31 | 2023-12-01 | 科大讯飞股份有限公司 | Prediction method and related device for answer score rate of test questions |
| CN112766513A (en) * | 2021-01-29 | 2021-05-07 | 华中师范大学 | Knowledge tracking method and system with memory cooperation |
| CN112818100A (en) * | 2021-01-29 | 2021-05-18 | 华中师范大学 | Knowledge tracking method and system fusing question difficulty |
| CN112818100B (en) * | 2021-01-29 | 2023-12-22 | 华中师范大学 | Knowledge tracking method and system for integrating question difficulty |
| CN112949929B (en) * | 2021-03-15 | 2022-06-07 | 华中师范大学 | Knowledge tracking method and system based on collaborative embedded enhanced topic representation |
| CN112949929A (en) * | 2021-03-15 | 2021-06-11 | 华中师范大学 | Knowledge tracking method and system based on collaborative embedded enhanced topic representation |
| CN112949935B (en) * | 2021-03-26 | 2022-06-17 | 华中师范大学 | Knowledge tracking method and system fusing student knowledge point question interaction information |
| CN112949935A (en) * | 2021-03-26 | 2021-06-11 | 华中师范大学 | Knowledge tracking method and system fusing student knowledge point question interaction information |
| CN113283585A (en) * | 2021-05-24 | 2021-08-20 | 北京大学 | Knowledge tracking method and system based on self-attention mechanism |
| CN113361685A (en) * | 2021-05-25 | 2021-09-07 | 华中师范大学 | Knowledge tracking method and system based on learner knowledge state evolution expression |
| CN113361685B (en) * | 2021-05-25 | 2022-08-19 | 华中师范大学 | Knowledge tracking method and system based on learner knowledge state evolution expression |
| CN113704235A (en) * | 2021-08-03 | 2021-11-26 | 桂林电子科技大学 | Depth knowledge tracking model based on self-attention mechanism |
| CN114723590A (en) * | 2022-03-30 | 2022-07-08 | 华南师范大学 | Group-oriented knowledge tracking method, system, device and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110807469B (en) | 2020-09-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110807469B (en) | Knowledge tracking method and system integrating long-time memory and short-time memory with Bayesian network | |
| Wilson et al. | Back to the basics: Bayesian extensions of IRT outperform neural networks for proficiency estimation | |
| CN110428010B (en) | Knowledge tracking method | |
| CN106409290B (en) | A method of child's intelligent sound education based on image analysis | |
| CN112085168A (en) | Knowledge tracking method and system based on dynamic key value gating circulation network | |
| Dorça et al. | A stochastic approach for automatic and dynamic modeling of students' learning styles in adaptive educational systems | |
| Guo et al. | Learning to query, reason, and answer questions on ambiguous texts | |
| CN114201684A (en) | Knowledge graph-based adaptive learning resource recommendation method and system | |
| CN114254127A (en) | Student ability portrayal method and learning resource recommendation method and device | |
| CN113724110A (en) | Interpretable depth knowledge tracking method and system and application thereof | |
| CN115545160B (en) | Knowledge tracking method and system for multi-learning behavior collaboration | |
| CN111126552A (en) | Intelligent learning content pushing method and system | |
| CN114861754A (en) | Knowledge tracking method and system based on external attention mechanism | |
| CN114429212A (en) | Intelligent learning knowledge ability tracking method, electronic device and storage medium | |
| Sha et al. | Neural knowledge tracing | |
| Kaptein et al. | Statistics for Data Scientists | |
| CN108229688A (en) | A kind of information processing method and device based on IRT | |
| KR20180061998A (en) | System and method for diagnosing mass attributes of leaner | |
| Yi | Research on English teaching reading quality evaluation method based on cognitive diagnostic evaluation | |
| Sprenger | Conditional degree of belief and Bayesian inference | |
| CN114117033B (en) | Knowledge tracking method and system | |
| CN113378581B (en) | Knowledge tracking method and system based on multivariate concept attention model | |
| CN114971066A (en) | Knowledge tracking method and system integrating forgetting factor and learning ability | |
| CN114971972A (en) | Deep knowledge tracking method integrating forgetting factor and exercise difficulty and intelligent terminal | |
| CN115131600A (en) | Detection model training method, detection method, device, equipment and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |