CN114742319A - Method, system and storage medium for predicting scores of law examination objective questions - Google Patents

Method, system and storage medium for predicting scores of law examination objective questions Download PDF

Info

Publication number
CN114742319A
CN114742319A CN202210528593.3A CN202210528593A CN114742319A CN 114742319 A CN114742319 A CN 114742319A CN 202210528593 A CN202210528593 A CN 202210528593A CN 114742319 A CN114742319 A CN 114742319A
Authority
CN
China
Prior art keywords
score
test
examination
value
training
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.)
Pending
Application number
CN202210528593.3A
Other languages
Chinese (zh)
Inventor
陈旭阳
杨旭川
刘琛
顾颃
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Chongqing Juexiao Technology Co ltd
Original Assignee
Chongqing Juexiao Technology Co ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Chongqing Juexiao Technology Co ltd filed Critical Chongqing Juexiao Technology Co ltd
Priority to CN202210528593.3A priority Critical patent/CN114742319A/en
Publication of CN114742319A publication Critical patent/CN114742319A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • 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
    • G06N3/048—Activation functions
    • 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
    • G06N3/06—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons
    • G06N3/061—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons using biological neurons, e.g. biological neurons connected to an integrated circuit
    • 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
    • G06N3/084—Backpropagation, e.g. using gradient descent
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
    • G06Q50/10—Services
    • G06Q50/20—Education
    • G06Q50/205—Education administration or guidance

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Business, Economics & Management (AREA)
  • Theoretical Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • General Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • General Health & Medical Sciences (AREA)
  • Strategic Management (AREA)
  • Molecular Biology (AREA)
  • Software Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • Mathematical Physics (AREA)
  • Computing Systems (AREA)
  • Data Mining & Analysis (AREA)
  • Computational Linguistics (AREA)
  • Human Resources & Organizations (AREA)
  • Artificial Intelligence (AREA)
  • Economics (AREA)
  • Tourism & Hospitality (AREA)
  • Marketing (AREA)
  • Educational Technology (AREA)
  • Educational Administration (AREA)
  • Neurology (AREA)
  • General Business, Economics & Management (AREA)
  • Microelectronics & Electronic Packaging (AREA)
  • Primary Health Care (AREA)
  • Development Economics (AREA)
  • Game Theory and Decision Science (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Operations Research (AREA)
  • Quality & Reliability (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention discloses a method, a system and a storage medium for predicting scores of legal examination objective questions, wherein the method comprises the following steps: the method comprises the steps of obtaining examination scores of m students preset in a previous year a, preprocessing the examination scores to obtain m score label vectors L, wherein a and m are positive integers; acquiring the average accuracy of each knowledge point of each student, the total class listening time and the total question making amount of each subject, and performing normalization processing to obtain m learning feature vectors F; constructing an examination result training sample according to the score label vector L and the learning characteristic vector F; training by using an examination result training sample to obtain an examination result prediction model, wherein the examination result prediction model is a pulse MLP model, LIF neurons are adopted for forward propagation, and sigmoid neurons are adopted for backward propagation; and acquiring the average correct rate of each knowledge point of the current student, the total length of the lecture listening time and the total quantity of the questions to be made of each subject, and predicting the examination score of the current student based on the examination score prediction model.

Description

Method, system and storage medium for predicting scores of law examination objective questions
Technical Field
The invention relates to the technical field of online education, in particular to a method, a system and a storage medium for predicting scores of law and examination objective questions.
Background
At present, the conventional method for predicting the achievement of objective questions in legal test mainly comprises the following steps:
the first method comprises the following steps: and (3) constructing a regression model by adopting a statistical learning method, and performing score fitting prediction. For example, a second-order response surface model is constructed by extracting four characteristics of the class concentration degree, the attendance rate, the online learning duration and the difficulty degree score of the test paper of the students to predict the score. The method is simple, and has the advantages of coarse feature granularity, low prediction accuracy and high mean square error.
And the second method comprises the following steps: by adopting a deep learning method, under the condition of known examination questions, the score of the current examination questions made by the student is predicted, but in the current annual examination, the student cannot know the examination questions in advance, so that the method has larger use limitation. If the method is used for predicting the scores of the students under the condition of unknown examination questions, the method has the defect of large calculation amount.
And the third is that: the method adopts a method combining machine learning and deep learning, such as a PCANet-BiGRU-based achievement prediction method, a processor, a readable storage medium and computer equipment, and adopts PCANet to extract features and input the features into a deep learning model BiGRU for prediction.
In summary, there is a need for a method for predicting the examination result of a trainee under the condition of unknown examination questions, and under the condition of ensuring accuracy, the method reduces the amount of calculation and energy consumption and improves the response speed.
Disclosure of Invention
Aiming at the technical defects in the prior art, the embodiment of the invention aims to provide a method, a system and a storage medium for predicting the score of an objective examination question of a legal examination, so as to accurately predict the score of a student taking an objective examination under the condition of unknown examination questions of the objective examination question of the legal examination.
In order to achieve the above object, in a first aspect, an embodiment of the present invention provides a method for predicting scores of legal test objective questions, including:
the method comprises the steps of obtaining examination scores of m students preset in a previous year a, preprocessing the examination scores to obtain m score label vectors L, wherein a and m are positive integers;
acquiring the average accuracy of each knowledge point of each student, the total class listening time and the total question making amount of each subject, and performing normalization processing to obtain m learning feature vectors F;
constructing an examination result training sample according to the score label vector L and the learning characteristic vector F;
training by using the test result training sample to obtain a test result prediction model, wherein the test result prediction model is a pulse MLP model, LIF neurons are adopted for forward propagation, and sigmoid neurons are adopted for backward propagation;
and acquiring the average accuracy of each knowledge point of the current student, the total length of the lectures attended by each subject and the total quantity of the questions asked, and predicting the examination score of the current student based on the examination score prediction model after normalization processing.
Optionally, the preprocessing the test scores includes:
labeling the test scores according to a preset score section;
and adopting one-hot coding to the corresponding label value to obtain a score label vector L.
Optionally, the labeling the test scores according to the preset score segment includes:
dividing the fraction into [0-M ]1)、[M1-M2)、……、[MN-1-MN)、[MN- ∞) N +1 stages;
if the test score belongs to the first score segment, setting the label of the test score to be 0;
if the test score belongs to the second score segment, setting the label of the test score to be 1;
by the way of analogy, the method can be used,
and if the test score belongs to the (N + 1) th score section, setting the label of the test score as N.
Optionally, an output layer of the test result prediction model has N +1 neurons, and an output dimension is N +1 dimension;
the updating mode of the LIF neuron value is as follows:
Vt＝Vt-1+[xt-(Vt-1-Vreset)]/tau；
wherein, VresetSetting the model voltage resetting parameter of the LIF neuron as 0; tau is a time constant parameter of the LIF neuron and is set to be 2; vt-1Taking the value of the LIF neuron at the previous moment, VtIs the current value of LIF neuron, xtThe input value of the LIF neuron at the current moment;
firing threshold V of LIF neuronsthresholdSet to 1, when the current value of LIF neuron is VtWhen 1 or more, the value transmitted to the next connected LIF neuron is 1, and the value of the LIF neuron is reset to 0; if the current value of LIF neuron is VtLess than 1, and a value of 0 is transmitted to the next connected LIF neuron.
Optionally, the training by using the test result training sample to obtain the test result prediction model includes:
s401: equally dividing the test result training samples into mutually exclusive K parts, and selecting K-1 parts of the mutually exclusive K parts for training, wherein K is more than or equal to 3;
s402: setting batch _ size as M, setting time step of each training of each batch sample as T, and setting iteration times as E, wherein M, T, E are positive integers;
s403: within each time step, carrying out Poisson coding on each examination result training sample to obtain a pulse training sample;
s404: inputting the pulse training sample into an initial examination result prediction model for training, and counting N +1 gods of an output layer of the initial examination result prediction modelProbability vector P ═ P for the average firing pulse of each neuron in the channel0，P1，P2，…，Pw，…，PN]Wherein, PwThe pulse number/T of the ith neuron in T time steps is given, w belongs to [0, N ∈]；
S405: selecting a loss function as MSE according to the probability vector P and the fraction label vector L, calculating the overall loss, performing back propagation, and adjusting the weight value, wherein,
the equation for the MSE loss function is as follows:
Figure BDA0003645618320000041
wherein L iskScore tag vector, P, representing the kth test achievement training samplekA probability vector representing a kth test achievement training sample;
s406: and (5) repeatedly executing the steps S403-S405, finishing training after all the test result training samples are subjected to iterative training for E times, and obtaining the test result prediction model.
Optionally, the predicting the examination result of the current student based on the examination result prediction model includes:
and the test result prediction model outputs a probability vector P, and a score segment represented by the maximum value in the probability vector P is selected as a prediction score segment of the current student.
Optionally, the average accuracy of each knowledge point of the learner is calculated by using the following formula:
Correct_ratei＝AVG(topic_correct_ratei)；
wherein the topic _ correct _ rateiCorrect _ Rate, the Correct rate of questions asked of the student that contain the ith knowledge pointiThe average accuracy of the student at the ith knowledge point is that i is more than or equal to 1.
Optionally, the following formula is adopted to perform normalization processing on the total lecture listening duration and the total subject making amount of each subject:
Figure BDA0003645618320000042
wherein x isjThe lecture listening duration or total question making value, x, of the jth subject of the studentjminThe minimum value of the lecture listening time or the minimum value of the total number of questions, x, of the jth subject in the m studentsjmaxMaximum value of the length of class listening time or maximum value of total number of questions, x, of jth subject in m studentsjnewThe value of the class time of the jth subject of the student or the value after the normalization of the total quantity of questions is given, and j is more than or equal to 1.
In a second aspect, an embodiment of the present invention provides a system for predicting a performance of a legal test objective, including a processor, an input device, an output device, and a memory, where the processor, the input device, the output device, and the memory are connected to each other, where the memory is used to store a computer program, and the computer program includes program instructions, and the processor is configured to call the program instructions to execute the method according to the first aspect.
In a third aspect, the present invention provides a computer-readable storage medium, in which a computer program is stored, the computer program including program instructions, which, when executed by a processor, cause the processor to execute the method according to the first aspect.
According to the method for predicting the performance of the legal test objective questions, provided by the embodiment of the invention, the calculation complexity of the model is reduced by adopting a mode of replacing neurons in the traditional MLP model with pulse neurons, so that the time for training and testing the model is reduced, and under the condition that a deep learning model is still adopted, the calculation complexity is reduced and the response speed is increased while the prediction accuracy is ensured. The problem of accurately predicting the score of the student in the objective examination under the condition of unknown legal examination objective examination questions is solved.
Drawings
In order to more clearly illustrate the detailed description of the invention or the technical solutions in the prior art, the drawings that are needed in the detailed description of the invention or the prior art will be briefly described below.
FIG. 1 is a schematic flow chart of a method for predicting the performance of a legal examination objective question provided in an embodiment of the present invention;
fig. 2 is a schematic structural diagram of a device for predicting the performance of a legal examination objective question provided in the embodiment of the present invention;
fig. 3 is a schematic structural diagram of a system for predicting the performance of a forensic objective test provided in the embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, not all, embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
It will be understood that the terms "comprises" and/or "comprising," when used in this specification and the appended claims, specify the presence of stated features, integers, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and/or groups thereof.
It is also to be understood that the terminology used in the description of the invention herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. As used in this specification and the appended claims, the singular forms "a", "an", and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise.
It should be further understood that the term "and/or" as used in this specification and the appended claims refers to any and all possible combinations of one or more of the associated listed items and includes such combinations.
As used in this specification and the appended claims, the term "if" may be interpreted contextually as "when", "upon" or "in response to a determination" or "in response to a detection". Similarly, the phrase "if it is determined" or "if a [ described condition or event ] is detected" may be interpreted contextually to mean "upon determining" or "in response to determining" or "upon detecting [ described condition or event ]" or "in response to detecting [ described condition or event ]".
It is to be noted that, unless otherwise specified, technical or scientific terms used herein shall have the ordinary meaning as understood by those skilled in the art to which the invention pertains.
The examination of the objective questions of the legal test totally comprises 18 subjects and 1394 knowledge points of criminal law, criminal complaint, constitution law, political law and administrative litigation law, law theory, international official law, legal occupation, legal history, civil law, civil complaint, business law, economic law, labor and social security law, international private law, international economic law, intellectual property law and environment and resource protection law.
After the students learn and test on the learning platform, the big data platform can count the total time length of listening to the lessons and the total quantity of doing questions of each student to each subject, and count the accuracy of the questions including the ith knowledge point made by each student. The big data platform also counts the scores of the students who have taken the examination of the objective questions of the legal examination in the previous year, the total time length of listening to the lessons and the total quantity of the questions to be taken of each subject and the accuracy rate of the questions containing the ith knowledge point to be taken by the students who have taken the examination of the objective questions of the legal examination. The server can retrieve all the data from the big data platform for analysis and use.
In a first aspect, as shown in fig. 1, a flowchart of a method for predicting the performance of an objective test of a legal test provided by an embodiment of the present invention may be used for a server to predict a score segment to which the performance of the objective test of the legal test of a student belongs without knowing the subject of the objective test of the legal test. The method may comprise the steps of:
s100: the method comprises the steps of obtaining examination scores of m preset students in a previous year a, preprocessing the examination scores and obtaining m score label vectors L, wherein a and m are positive integers.
The server extracts examination scores of m students preset in the previous a years from the big data platform, for example, the examination scores of 17892 students who participate in an objective examination in the previous 1 year may be extracted, that is, a is 1, and m is 17892. After the examination score is extracted, the examination score is preprocessed, and the preprocessing comprises the following steps: labeling the test scores according to a preset score section; and adopting one-hot coding to the corresponding label value to obtain a score label vector L. And finally, summarizing the score label vectors L of the m students to obtain m score label vectors L.
In this embodiment, the labeling the test scores according to the preset score segment includes:
dividing the fraction into [0-M ]1)、[M1-M2)、……、[MN-1-MN)、[MN- ∞) N +1 stages;
if the test score belongs to the first score segment, setting the label of the test score to be 0;
if the test score belongs to the second score segment, setting the label of the test score to be 1;
by the way of analogy, the method can be used,
and if the test score belongs to the (N + 1) th score section, setting the label of the test score as N.
For example, the student's test scores are labeled according to the following score segments:
if the test score is less than 140 points, the label is 0; if the test score is larger than or equal to 140 and smaller than 150, the label is 1; if the test score is more than or equal to 150 and less than 160, the label is 2; if the test score is more than or equal to 160 and less than 170, the label is 3; if the test score is greater than or equal to 170 and less than 180, the label is 4; if the test score is greater than or equal to 180 and less than 190, the label is 5; if the test score is more than or equal to 190 and less than 200, the label is 6; if the test score is greater than or equal to 200 and less than 210, the label is 7; if the test score is more than or equal to 210 and less than 220, the label is 8; if the test score is greater than or equal to 220, the label is 9. In the present example, N +1 is divided into 10 fractional segments, where N is 9.
The mark of the student score is greater than or equal to 5, the corresponding label value is coded by one-hot, for example, if the examination score of a student is 180 points, the score label vector L formed after the label is 5, one-hot coding is marked as [0,0,0,0,0,1,0,0,0,0, 0 ]. The score tag vectors L for 17892 students are summed to form a score tag vector dataset with dimensions 17892 x1 x 10.
S200: and acquiring the average correct rate of each knowledge point of each student, the total class listening time and the total question making amount of each subject, and performing normalization processing to obtain m learning feature vectors F.
Respectively counting the average correct rate of each knowledge point of each student from a big data platform, wherein the average correct rate of the knowledge points is counted according to the questions made by the students, and the statistical formula is as follows:
Correct_ratei＝AVG(topic_correct_ratei)；
wherein, the top _ correct _ rateiCorrect _ Rate, the Correct rate of questions including the ith knowledge Point for the studentiThe average accuracy of the student at the ith knowledge point is that i is more than or equal to 1. The above statistics form 1394 eigenvalues, which lie between 0 and 1.
Then, from the big data platform, the total length of class attending and total number of questions of each subject of each student are respectively counted, and 36 feature values are counted.
Because the two characteristics of the lecture attending duration and the question making total amount and the average accuracy rate characteristic dimension of the knowledge points are different, in order to eliminate the difference of the characteristics on the dimension, the lecture attending duration and the question making total amount are respectively normalized, and the normalization operation formula is as follows:
Figure BDA0003645618320000081
wherein x isjThe lecture listening duration or total question making value, x, of the jth subject of the studentjminThe minimum value of the lecture listening time or the minimum value of the total number of questions, x, of the jth subject in the m studentsjmaxMaximum lecture time length or maximum question making total amount of jth subject in m students, xjnewThe value of the student after the student's j subject listening time or the question total quantity normalizationIn this embodiment, j ∈ [1,18 ]]。
In this embodiment, the learning feature vector F of each learner includes 1430 feature values, and 17892 learning feature vectors F with 1430 dimensions are obtained.
S300: and constructing an examination result training sample according to the score label vector L and the learning characteristic vector F.
In this embodiment, the 17892 learning feature vectors F with 1430 dimensions are used as examination result training samples for training.
S400: training by using the test result training sample to obtain a test result prediction model, wherein the test result prediction model is a pulse MLP model, LIF neurons are adopted for forward propagation, and sigmoid neurons are adopted for backward propagation.
Since the conventional MLP model cannot reduce the amount of calculation, the pulse MLP model is adopted, but the pulse MLP model has a big problem: when a spiking neuron is used, model training cannot be performed using back-propagation of traditional neural networks because the spiking neuron is not conductive. In order to solve the problem, in the embodiment, the pulse MLP model uses the LIF neuron during forward propagation and uses the sigmoid neuron during backward propagation, and the sigmoid neuron is differentiable, so that the backward propagation can be adopted for training, after the training is completed, the weight is retained, and the test result prediction model only leaves the LIF neuron for reasoning. Because the calculation complexity of the LIF neuron is far lower than that of the traditional tanh neuron, sigmoid neuron or Gaussian error linear unit neuron, the method has great progress in reducing the calculation amount, improving the reasoning speed and reducing the response time.
In this embodiment, the output layer of the test result prediction model has N +1 neurons, and the output dimension is N +1 dimension, and corresponds to the label values 0 to N of the test scores respectively.
Specifically, in this embodiment, the network structure of the whole pulse MLP model is as follows:
an input layer: the dimension is 1430, 1430 neurons are used, corresponding to 1430 eigenvalues for each student.
Hidden layer 1: the dimension is 500, and the characteristic value of the student is subjected to dimension reduction processing.
Hidden layer 2: the dimension is 100, and the characteristic value of the student is subjected to further dimension reduction processing.
An output layer: the dimension is 10, 10 neurons are totally arranged, and the 10 neurons respectively correspond to the label values of the test scores from 0 to 9.
When the LIF neuron is used for forward propagation, when the input of the LIF neuron is attenuated along with time, the updating mode of the LIF neuron value is as follows:
Vt＝Vt-1+[xt-(Vt-1-Vreset)]/tau；
wherein, VresetSetting the model voltage resetting parameter of the LIF neuron as 0; tau is a time constant parameter of the LIF neuron and is set to be 2; vt-1Taking the value of the LIF neuron at the previous moment, VtIs the current value of LIF neuron, xtThe input value of the LIF neuron at the current moment is obtained;
firing threshold V of LIF neuronsthresholdSet to 1, when the current value of LIF neuron is VtWhen the current value is 1 or more, the value transmitted to the next connected LIF neuron is 1, that is, a pulse is issued to the next connected LIF neuron, and the value of the LIF neuron is reset to 0. If the current value of LIF neuron is VtAnd if the number is less than 1, the value transmitted to the next connected LIF neuron is 0, namely, a pulse is not sent to the next connected LIF neuron, so that part of neurons are not in an activated state, the calculated amount of the model is reduced, and the calculation speed is improved.
In the process of back propagation, sigmoid neurons are adopted, and the formula is as follows:
Figure BDA0003645618320000101
the derivative of sigmoid is derived from mathematical derivation as:
Figure BDA0003645618320000102
in this embodiment, the training by using the test result training sample to obtain the test result prediction model includes:
s401: and equally dividing the test result training samples into mutually exclusive K parts, and selecting K-1 parts of the mutually exclusive K parts for training, wherein K is more than or equal to 3.
In this embodiment, a 5-fold cross validation method is adopted to perform model training and testing, 17892 parts of the sample are equally divided into 5 parts, 4 parts are selected for training each time, and the remaining 1 part is validated.
S402: setting batch _ size as M, setting time step of each training of each batch sample as T, and setting iteration number as E, wherein M, T, E are positive integers.
In this embodiment, the batch _ size is set to 64, that is, M is set to 64; setting the time step T of each training of each batch sample to be 100; the number of training iterations E is 100, which means that training is performed 100 times by using 17892 samples, and after 100 iterations, the training is ended.
S403: and in each time step, performing Poisson coding on each test result training sample to obtain a pulse training sample.
In this embodiment, in each time step, first, 1430 feature values to be input are subjected to poisson coding, where the poisson coding includes:
s4031: randomly generating 1430 uniformly distributed values which are greater than or equal to 0 and less than 1 to form a vector X with 1430 dimensions;
s4032: comparing the size of a learning characteristic vector F consisting of the vector X and 1430 characteristic values of the student according to each dimension; if XI<＝FiSetting the input pulse value to be 1, otherwise, setting the input pulse value to be 0, forming a 1430-dimensional input pulse vector I consisting of 0 and 1, and taking the input pulse vector I as a pulse training sample.
So far, the learning characteristic vector F of each student is converted into an input pulse vector I, the distribution of the number of times of pulse issuing conforms to the Poisson process, and obviously, the characteristic value FiThe larger the number of pulses delivered over a given period of time.
S404: will be described inInputting a pulse training sample into an initial test result prediction model for training, and counting the probability vector P of average pulse transmission of each neuron in N +1 neurons of an output layer of the initial test result prediction model, wherein the probability vector P is [ P ═ P [ ]0，P1，P2，…，Pw，…，PN]Wherein, PwThe pulse number/T of the ith neuron in T time steps is given, w belongs to [0, N ∈]。
In this embodiment, a pulse training sample is input to an initial test result prediction model, and a probability vector P of an average release pulse for each neuron of 10 neurons in a statistical model output layer is [ P ═ P0，P1，P2，P3，P4，P5，P6，P7，P8，P9]. For example, if P is [0,0,0,0,0.2,0.8,0,0]At this time, P can be seen4＝0.2，P5When the pulse release probability of the fifth neuron of the output layer is 0.2 and the pulse release probability of the sixth neuron is 0.8, the sixth neuron is more active in the case of input pulse, and the final test score of the student is predicted to be 180 segments because the sixth neuron represents 180 segments.
S405: selecting a loss function as MSE according to the probability vector P and the fraction label vector L, calculating the overall loss, performing back propagation, and adjusting the weight value, wherein,
the calculation formula of the MSE loss function is as follows:
Figure BDA0003645618320000111
wherein L iskScore tag vector, P, representing the kth test achievement training samplekIs a probability vector representing the kth test achievement training sample.
S406: and (5) repeatedly executing the steps S403-S405, finishing training after all the test result training samples are subjected to iterative training for E times, and obtaining the test result prediction model.
In this embodiment, steps S403 to S405 are repeatedly executed, iterative training is performed 100 times using 17892 samples, and after 100 iterations, the training is ended, and the examination result prediction model is obtained.
S500: and acquiring the average correct rate of each knowledge point of the current student, the total length of the lecture listening time and the total quantity of the questions to be asked, normalizing, and predicting the examination score of the current student based on the examination score prediction model.
The average accuracy of each knowledge point, the total length of the lecture listening time and the total quantity of the questions of each subject of the current student are extracted from a big data platform, normalization processing is carried out, the obtained data of the current student are input into an examination score prediction model, the examination score prediction model outputs a corresponding prediction score section, and a prediction result is given to the examination score of the current student.
For example: the average accuracy of 1394 knowledge points of 18 subjects acquired to the trainee is:
x1＝[0.4,0.2,0.3,……,0.4,0.8]；
the class attending time of 18 subjects is respectively as follows:
[20,32,21,18,10,4,3,5,2,8,9,2,3,8,9,10,11,12],
after the normalization treatment, the data are processed,
x2＝[0.4,0.64,0.42,0.36,0.2,0.08,0.06,0.1,0.04,0.16,0.18,0.04,0.06,0.16,0.18,0.2,0.22,0.24]；
the questions were:
[100,200,1100,1200,2000,300,400,20,70,80,200,210,230,320,443,234,532,234]，
after the normalization treatment is carried out,
x3＝[0.04,0.08,0.44,0.48,0.8,0.12,0.16,0.008,0.028,0.032,0.08,0.084,0.092,0.128,0.1772,0.0936,0.2128,0.0936]；
setting T to be 100, splicing x1, x2 and x3, performing pulse coding on input by adopting the method of the step S403, inputting the input into a score prediction model, performing nonlinear mapping on an implicit layer and an output layer, outputting 10 neurons of the layer, counting the pulse transmission frequency of each neuron in T to be 100 time steps to obtain a probability vector P, and selecting a score segment represented by the maximum value in the probability vector P as a prediction score segment of a student.
According to the method for predicting the performance of the legal test objective questions, provided by the embodiment of the invention, the calculation complexity of the model is reduced by adopting a mode of replacing neurons in the traditional MLP model with pulse neurons, so that the time for training and testing the model is reduced, and under the condition that a deep learning model is still adopted, the calculation complexity is reduced and the response speed is increased while the prediction accuracy is ensured. The problem of accurately predicting the score of the student in the objective examination under the condition of unknown legal examination objective examination questions is solved.
In a second aspect, based on the same inventive concept, an embodiment of the present invention provides a device for predicting a performance of a legal test objective problem. As shown in fig. 2, the apparatus may include:
the score label vector generation module 201 is configured to acquire examination scores of m students preset in a year before, and preprocess the examination scores to acquire m score label vectors L, where a and m are positive integers;
a learning feature vector generation module 202, configured to obtain an average accuracy of each knowledge point of each learner, a total length of lectures attended by each subject, and a total number of questions asked by each subject, and perform normalization processing to obtain m learning feature vectors F;
a training sample construction module 203, configured to construct an examination result training sample according to the score label vector L and the learning feature vector F;
the prediction model generation module 204 is used for training by using the test result training sample to obtain a test result prediction model, wherein the test result prediction model is a pulse MLP model, LIF neurons are adopted for forward propagation, and sigmoid neurons are adopted for backward propagation;
the examination result prediction module 205 is configured to obtain an average accuracy of each knowledge point of the current student, a total length of lectures and a total amount of questions asked of each subject, and predict an examination result of the current student based on the examination result prediction model.
In a third aspect, the embodiment of the invention provides a system for predicting the performance of a legal test objective subject. As shown in fig. 3, the system may include: one or more processors 101, one or more input devices 102, one or more output devices 103, and memory 104, the processors 101, input devices 102, output devices 103, and memory 104 being interconnected via a bus 105. The memory 104 is used for storing a computer program comprising program instructions, and the processor 101 is configured to call the program instructions to execute the method of the embodiment of the method for predicting the performance of the legal test objective.
It should be understood that, in the embodiment of the present invention, the Processor 101 may be a Central Processing Unit (CPU), and the Processor may also be other general-purpose processors, Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), Field-Programmable Gate arrays (FPGAs) or other Programmable logic devices, discrete Gate or transistor logic devices, discrete hardware components, and the like. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.
The input device 102 may include a keyboard or the like, and the output device 103 may include a display (LCD or the like), a speaker, or the like.
The memory 104 may include read-only memory and random access memory, and provides instructions and data to the processor 101. A portion of the memory 104 may also include non-volatile random access memory. For example, the memory 104 may also store device type information.
In a specific implementation, the processor 101, the input device 102, and the output device 103 described in the embodiment of the present invention may execute the implementation manner described in the embodiment of the legal test objective performance prediction method provided in the embodiment of the present invention, and are not described herein again.
It should be noted that, with respect to the specific workflow of the system for predicting the performance of the legal test objective problem, reference may be made to the foregoing method embodiment, and details are not described herein again.
Further, an embodiment of the present invention also provides a readable storage medium, in which a computer program is stored, where the computer program includes program instructions, and the program instructions, when executed by a processor, implement: the method for predicting the scores of the objective questions in the test.
The computer readable storage medium may be an internal storage unit of the background server described in the foregoing embodiment, for example, a hard disk or a memory of the system. The computer readable storage medium may also be an external storage device of the system, such as a plug-in hard drive, Smart Media Card (SMC), Secure Digital (SD) Card, Flash memory Card (Flash Card), etc. provided on the system. Further, the computer readable storage medium may also include both an internal storage unit and an external storage device of the system. The computer-readable storage medium is used for storing the computer program and other programs and data required by the system. The computer readable storage medium may also be used to temporarily store data that has been output or is to be output.
Those of ordinary skill in the art will appreciate that the elements and algorithm steps of the examples described in connection with the embodiments disclosed herein may be embodied in electronic hardware, computer software, or combinations of both, and that the components and steps of the examples have been described in a functional general in the foregoing description for the purpose of illustrating clearly the interchangeability of hardware and software. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the implementation. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present invention.
In addition, functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit can be realized in a form of hardware, and can also be realized in a form of a software functional unit.
The integrated unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present invention essentially or partly contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk, or an optical disk, and various media capable of storing program codes.
While the invention has been described with reference to specific embodiments, the invention is not limited thereto, and various equivalent modifications and substitutions can be easily made by those skilled in the art within the technical scope of the invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims (10)

1. A method for predicting scores of legal test objective questions is characterized by comprising the following steps:
the method comprises the steps of obtaining examination scores of m students preset in a previous year a, preprocessing the examination scores to obtain m score label vectors L, wherein a and m are positive integers;
acquiring the average accuracy of each knowledge point of each student, the total class listening time and the total question making amount of each subject, and performing normalization processing to obtain m learning feature vectors F;
constructing an examination result training sample according to the score label vector L and the learning characteristic vector F;
training by using the test result training sample to obtain a test result prediction model, wherein the test result prediction model is a pulse MLP model, LIF neurons are adopted for forward propagation, and sigmoid neurons are adopted for backward propagation;
and acquiring the average correct rate of each knowledge point of the current student, the total length of the lecture listening time and the total quantity of the questions to be asked, normalizing, and predicting the examination score of the current student based on the examination score prediction model.
2. The method of claim 1, wherein the pre-processing the test scores comprises:
labeling the test scores according to a preset score section;
and adopting one-hot coding to the corresponding label value to obtain a score label vector L.
3. The method of claim 2, wherein said labeling test scores according to a preset score segment comprises:
dividing the fractional segments into [0-M1)、[M1-M2)、……、[MN-1-MN)、[MN- ∞) N +1 stages;
if the test score belongs to the first score segment, setting the label of the test score to be 0;
if the test score belongs to the second score segment, setting the label of the test score to be 1;
by the way of analogy, the method can be used,
and if the test score belongs to the (N + 1) th score section, setting the label of the test score as N.
4. The method of claim 3, wherein an output layer of the test achievement prediction model has N +1 neurons, an output dimension is N +1 dimension;
the updating mode of the LIF neuron value is as follows:
Vt＝Vt-1+[xt-(Vt-1-Vreset)]/tau；
wherein, VresetSetting the model voltage resetting parameter of the LIF neuron as 0; tau is a time constant parameter of the LIF neuron and is set to be 2; vt-1Value V of the LIF neuron at the previous momenttIs the current value of LIF neuron, xtThe input value of the LIF neuron at the current moment;
firing threshold V of LIF neuronsthresholdSet to 1, when the current value of LIF neuron is VtWhen the current value is greater than or equal to 1, the value transmitted to the next connected LIF neuron is 1, and the value of the LIF neuron is reset to 0; if the current value of LIF neuron is VtLess than 1, and a value of 0 is transmitted to the next connected LIF neuron.
5. The method of claim 4, wherein training with the test achievement training sample to obtain a test achievement prediction model comprises:
s401: equally dividing the test result training samples into mutually exclusive K parts, and selecting K-1 parts of the mutually exclusive K parts for training, wherein K is more than or equal to 3;
s402: setting batch _ size as M, setting time step of each training of each batch sample as T, and setting iteration times as E, wherein M, T, E are positive integers;
s403: within each time step, carrying out Poisson coding on each examination result training sample to obtain a pulse training sample;
s404: inputting the pulse training sample into an initial test result prediction model for training, and counting the probability vector P of average pulse transmission of each neuron in N +1 neurons of an output layer of the initial test result prediction model [ P ═ P [ ]0，P1，P2，…，Pw，…，PN]Wherein, PwThe pulse number/T of the ith neuron in T time steps is given, w belongs to [0, N ∈]；
S405: selecting a loss function as MSE according to the probability vector P and the fraction label vector L, calculating the overall loss, performing back propagation, and adjusting the weight value, wherein,
the equation for the MSE loss function is as follows:
Figure FDA0003645618310000021
wherein L iskScore tag vector, P, representing the kth test achievement training samplekA probability vector representing a kth test achievement training sample;
s406: and (5) repeatedly executing the steps S403-S405, finishing training after all the test result training samples are subjected to iterative training for E times, and obtaining the test result prediction model.
6. The method of claim 5, wherein: the predicting the examination result of the current student based on the examination result prediction model comprises the following steps:
and the test result prediction model outputs a probability vector P, and a score segment represented by the maximum value in the probability vector P is selected as a prediction score segment of the current student.
7. The method of claim 1 wherein the average correct rate for each knowledge point of the trainee is calculated using the formula:
Correct_ratei＝AVG(topic_correct_ratei)；
wherein the topic _ correct _ rateiCorrect _ Rate, the Correct rate of questions asked of the student that contain the ith knowledge pointiThe average accuracy of the student at the ith knowledge point is that i is more than or equal to 1.
8. The method of claim 1, wherein the following formula is used to normalize the total length of lectures and total number of questions asked for each subject:
Figure FDA0003645618310000031
wherein x isjThe lecture listening duration or total question making value, x, of the jth subject of the studentjminThe minimum value of the lecture listening time or the minimum value of the total number of questions, x, of the jth subject in the m studentsjmaxMaximum length of class listening time of jth subject in m students orMaximum value of total amount of questions, xjnewThe value of the class time of the jth subject of the student or the value after the normalization of the total quantity of questions is given, and j is more than or equal to 1.
9. A forensic objective performance prediction system comprising a processor, an input device, an output device and a memory, the processor, the input device, the output device and the memory being interconnected, wherein the memory is adapted to store a computer program comprising program instructions, the processor being configured to invoke the program instructions to perform the method of any of claims 1 to 8.
10. A computer-readable storage medium, characterized in that the computer-readable storage medium stores a computer program comprising program instructions that, when executed by a processor, cause the processor to carry out the method according to any one of claims 1-8.
CN202210528593.3A 2022-05-16 2022-05-16 Method, system and storage medium for predicting scores of law examination objective questions Pending CN114742319A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210528593.3A CN114742319A (en) 2022-05-16 2022-05-16 Method, system and storage medium for predicting scores of law examination objective questions

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210528593.3A CN114742319A (en) 2022-05-16 2022-05-16 Method, system and storage medium for predicting scores of law examination objective questions

Publications (1)

Publication Number Publication Date
CN114742319A true CN114742319A (en) 2022-07-12

Family

ID=82285077

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210528593.3A Pending CN114742319A (en) 2022-05-16 2022-05-16 Method, system and storage medium for predicting scores of law examination objective questions

Country Status (1)

Country Link
CN (1) CN114742319A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117556381A (en) * 2024-01-04 2024-02-13 华中师范大学 Knowledge level depth mining method and system for cross-disciplinary subjective test questions
CN117689025A (en) * 2023-12-07 2024-03-12 上海交通大学 Quick large model reasoning service method and system suitable for consumer display card

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117689025A (en) * 2023-12-07 2024-03-12 上海交通大学 Quick large model reasoning service method and system suitable for consumer display card
CN117556381A (en) * 2024-01-04 2024-02-13 华中师范大学 Knowledge level depth mining method and system for cross-disciplinary subjective test questions
CN117556381B (en) * 2024-01-04 2024-04-02 华中师范大学 Knowledge level depth mining method and system for cross-disciplinary subjective test questions

Similar Documents

Publication Publication Date Title
US10628731B1 (en) Deep convolutional neural networks for automated scoring of constructed responses
CN108171358B (en) Score prediction method and device, storage medium and electronic device
EP3872716A1 (en) Image classification neural network system with batch normalization layers
CN108280757B (en) User credit evaluation method and device
CN112508334B (en) Personalized paper grouping method and system integrating cognition characteristics and test question text information
CN114742319A (en) Method, system and storage medium for predicting scores of law examination objective questions
CN110263328B (en) Discipline capability type labeling method and device, storage medium and terminal equipment
CN112632351B (en) Classification model training method, classification method, device and equipment
CN113240510B (en) Abnormal user prediction method, device, equipment and storage medium
US20220108149A1 (en) Neural networks with pre-normalized layers or regularization normalization layers
CN110852071A (en) Knowledge point detection method, device, equipment and readable storage medium
CN111079433A (en) Event extraction method and device and electronic equipment
WO2021147405A1 (en) Customer-service statement quality detection method and related device
CN113053395A (en) Pronunciation error correction learning method and device, storage medium and electronic equipment
CN113705092B (en) Disease prediction method and device based on machine learning
CN113239075B (en) Construction data self-checking method and system
CN115660060A (en) Model training method, detection method, device, equipment and storage medium
Lim et al. Score tests for zero-inflation and overdispersion in two-level count data
CN112085584A (en) Enterprise credit default probability calculation method and system
CN114911911A (en) Multi-turn conversation method and device and electronic equipment
Zamanzadeh et al. Autopopulus: a novel framework for autoencoder imputation on large clinical datasets
CN111325387A (en) Interpretable law automatic decision prediction method and device
US20230368920A1 (en) Learning apparatus, mental state sequence prediction apparatus, learning method, mental state sequence prediction method and program
US20240119366A1 (en) Online training of machine learning models using bayesian inference over noise
CN117236900B (en) Individual tax data processing method and system based on flow automation

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination