CN112825147A - Learning path planning method, device, equipment and storage medium - Google Patents

Abstract

The embodiment of the invention provides a learning path planning method, a device, equipment and a storage medium, wherein the learning path planning method comprises the following steps: acquiring a knowledge point set to be subjected to path planning, learning behavior data of students to be planned and mastery degree data of the students to be planned at each knowledge point of the knowledge point set; acquiring at least one of a learning duration prediction vector and a learning interest prediction vector and a learning degree variation prediction vector according to the learning behavior data and the mastering degree data of each knowledge point; performing vector fusion on at least one of the learning duration prediction vector and the learning interest prediction vector and the grasping degree variation prediction vector to obtain an integrated knowledge point evaluation vector; the knowledge points are sequenced according to the size of each dimension of the comprehensive knowledge point evaluation vector to obtain the learning planning path of each knowledge point, so that the path planning accuracy and comprehensiveness can be improved, and the learning efficiency and learning interest of students can be better improved.

Description

Learning path planning method, device, equipment and storage medium

Technical Field

The embodiment of the invention relates to the technical field of information, in particular to a learning path planning method, a learning path planning device, learning path planning equipment and a storage medium.

Background

In the intelligent education, how to achieve the purpose of practical education, the learning and planning are the key. That is, it is necessary to plan an individualized learning path that can comprehensively consider factors of each learning party based on the learning condition, learning ability and learning preference of each student, so as to improve the learning efficiency and learning interest of the students.

However, comprehensive research shows that the prior art has the defects of low comprehensive degree, weak model generalization capability and the like during path planning, so that the planning result of the learned path planning is not accurate enough.

Therefore, there are great challenges and difficulties in improving the accuracy of the learning path planning result.

Disclosure of Invention

The technical problem to be solved by the embodiment of the invention is to provide a learning path planning method to improve the accuracy of a learning path planning result.

In order to solve the above problem, an embodiment of the present invention provides a learned path planning method, including:

acquiring a knowledge point set to be subjected to path planning, learning behavior data of students to be planned and mastery degree data of the students to be planned at each knowledge point of the knowledge point set;

acquiring at least one of a learning duration prediction vector and a learning interest prediction vector and a learning degree variation prediction vector according to the learning behavior data and the mastering degree data of each knowledge point, wherein each dimension of the learning duration prediction vector represents the learning duration of each knowledge point, each dimension of the learning interest prediction vector represents the learning interest of each knowledge point, and each dimension of the learning degree variation prediction vector represents the mastering degree variation of each knowledge point;

performing vector fusion on at least one of the learning duration prediction vector and the learning interest prediction vector and the mastering degree variation prediction vector to obtain an integrated knowledge point evaluation vector;

and sequencing the knowledge points according to the size of each dimension of the comprehensive knowledge point evaluation vector to obtain the learning planning path of each knowledge point.

In order to solve the above problem, an embodiment of the present invention further provides a learning path planning apparatus, including:

the learning data acquisition module is suitable for acquiring a knowledge point set to be subjected to path planning, learning behavior data of students to be planned and mastery degree data of the students to be planned at all knowledge points in the knowledge point set;

the prediction vector acquisition module is suitable for acquiring at least one of a learning duration prediction vector and a learning interest prediction vector and a learning degree variation prediction vector according to the learning behavior data and the mastering degree data of each knowledge point, wherein each dimension of the learning duration prediction vector represents the learning duration of each knowledge point, each dimension of the learning interest prediction vector represents the learning interest of each knowledge point, and each dimension of the mastering degree variation prediction vector represents the variation of the mastering degree of each knowledge point;

the comprehensive knowledge point evaluation vector acquisition module is suitable for carrying out vector fusion on at least one of the learning duration prediction vector and the learning interest prediction vector and the mastering degree variation prediction vector to acquire a comprehensive knowledge point evaluation vector;

and the learning path planning generation module is suitable for sequencing the knowledge points according to the size of each dimension of the comprehensive knowledge point evaluation vector to obtain the learning planning path of each knowledge point.

To solve the above problem, an embodiment of the present invention further provides an apparatus, including at least one memory and at least one processor; the memory stores a program, and the processor calls the program to execute the learned path planning method.

In order to solve the above problem, an embodiment of the present invention further provides a storage medium storing a program suitable for learning path planning, so as to implement the learning path planning method.

Compared with the prior art, the technical scheme of the embodiment of the invention has the following advantages:

by adopting the learning path planning method provided by the embodiment of the invention, firstly, a knowledge point set to be path planned, learning behavior data of students to be planned and mastery degree data of the students to be planned at each knowledge point of the knowledge point set are obtained, then a learning degree variation prediction vector and at least one of a learning duration prediction vector and a learning interest prediction vector are obtained according to the data, and at least one of the learning duration prediction vector and the learning interest prediction vector and the mastering degree variation prediction vector are subjected to vector fusion to obtain a comprehensive knowledge point evaluation vector; because the comprehensive knowledge point evaluation vector is fused with at least two dimension vectors, the prediction result of the comprehensive knowledge point evaluation vector can be more comprehensive and accurate, and therefore, the knowledge points are sequenced according to the dimension of each dimension of the comprehensive knowledge point evaluation vector, and the planning result of the learning path planning is more suitable for the learning efficiency and the learning interest of students to be planned. Therefore, the learning path planning method provided by the embodiment of the invention can comprehensively consider at least two factors influencing the learning of students, thereby improving the accuracy and the comprehensiveness of path planning, ensuring that the finally planned learning path has the advantages of high comprehensiveness, strong pertinence, expected accuracy and the like, and better improving the learning efficiency and the learning interest of the students.

In an alternative, the step of obtaining learning behavior data of a student to be planned includes: acquiring the learning behavior original data of the student to be planned; and carrying out statistical analysis on the learning behavior original data to obtain the learning behavior data of the student to be planned. Because the original data of the learning behaviors are normalized in advance, the learning behavior data obtained through statistical analysis is clean and has understandable statistical characteristics, and the normalized data with the added statistical characteristics can more accurately and comprehensively reflect the learning ability, learning interest and learning habit of students, so that the standardization degree of the data and the accuracy of the result are improved, the accuracy of final prediction is improved, and the accuracy and the operation efficiency of path planning are improved.

Drawings

Fig. 1 is a schematic flow chart of a learning path planning method according to an embodiment of the present invention;

fig. 2 is a schematic structural diagram of a trained deep neural network of the learned path planning method according to the embodiment of the present invention;

fig. 3 is a schematic diagram of a training process of a deep neural network of the learned path planning method according to the embodiment of the present invention;

FIG. 4 is a partial flowchart of a learned path planning method according to an embodiment of the present invention;

fig. 5 is a block diagram of a learned route planning apparatus according to an embodiment of the present invention.

Detailed Description

As can be seen from the background art, the planning result of the learned path planning in the prior art is not accurate enough.

Through analysis, the prior art scheme considers a plurality of planning targets to be integrated into a single model less, and most of the planning targets establish a single learning model aiming at each target, which brings two disadvantages: one is that a single target learning process faces the problems of large noise, weak generalization capability and the like, and the performance on new data is difficult to achieve the expectation; the other is that the learning process of a single target is independently carried out, the information of other targets cannot be well integrated to help the current target to be trained and learned better, and the accuracy and the intelligence degree are insufficient.

In order to improve the accuracy of the learning path planning result, the invention provides a learning path planning method, and the technical solution in the embodiment of the invention will be clearly and completely described below with reference to the drawings in the embodiment of the 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.

Fig. 1 is a flow chart of a learning path planning method according to an embodiment of the present invention.

As shown in fig. 1, the learned path planning method according to the embodiment of the present invention may include the following steps:

step S1: acquiring a knowledge point set to be subjected to path planning, learning behavior data of students to be planned and mastery degree data of the students to be planned at each knowledge point of the knowledge point set;

the knowledge point set to be planned for the path may include each specific knowledge point and knowledge point basic information mainly used for measuring basic attributes of the current knowledge points to be planned.

In order to improve the accuracy of the knowledge points, in an embodiment, the basic information of the knowledge points may include difficulty of the knowledge points and capability level of investigation of the knowledge points, i.e., a knowledge point set (denoted by od) for path planning₃) Each knowledge point in (1) slave knowledge point difficulty (denoted as od)_3，1) And level of knowledge point investigation (denoted od)_3，2) Two aspects are defined:

for this purpose, the set of knowledge points to be path-planned may be represented by a vector as: od₂＝(od_2，1，od_2，2). Wherein od_3，1＝(od_3，1，1，od_3，1，2，...，od_3，1，k，...，od_3，1，K)，od_3，1，kRepresenting the difficulty of the k-th knowledge point, wherein the value is from 0 to 1, and the larger the value is, the higher the difficulty of learning the knowledge point is; od_3，2＝(od_3，2，1，od_3，2，2，...，od_3，2，k，...，od_3，2，K)，od_3，2，kAnd the capability level of the k-th knowledge point investigation is represented, the value range is one of {1, 2, 3 and 4}, 1 represents memorization, 2 represents understanding, 3 represents application, and 4 represents synthesis. The larger the value, the higher the level of capability that represents the investigation. In addition, K represents the total number of knowledge points to be planned; of course, this is merely an example, and in other embodiments, a person skilled in the art may define the capability level of the knowledge point investigation according to the actual situation.

It is easy to understand that the learning behavior data of the student to be planned refers to behavior data generated by the student in learning each knowledge point, and may include total learning duration, number of questions to be made, course completion condition, single learning time, activity level (measured based on interaction with resources and other students), activity level (based on the active behavior of the student on the information system, such as searching for questions, asking questions, etc.), and the like.

In order to analyze the learning behavior of the student from multiple dimensions, in one embodiment, the learning behavior data of the student to be planned may include learning process data (denoted by od)₂) And learning preference data (denoted od)₅) Said learning process data od₂Representing process data generated by students to be planned in the process of learning each knowledge point, the learning preference data od₅And the data represent the learning preference and interest data of the students to be planned in learning each knowledge point.

Wherein the learning process data od₂Can reflect the learning habits, personal characters and psychological characteristics of students, for example, students with high initiative degree can actively take the initiative, and relatively speaking, the learning ability is stronger, because the students can actively search resources and seek help. For another example, the duration of a single study can reflect the concentration degree of students, and some students leave the system to a certain extent in a short time each time of study, which indicates that the concentration degree of the students is low.

Learning preference data od₅Mainly comes from the learning condition and the active operation behavior of the students to be planned. The learning condition refers to the final mastery degree of students on different knowledge, and knowledge points with high mastery degree are relatively large in learning preference and good at the knowledge points; the active operation behavior mainly refers to the behavior of actively searching for questions or actively learning more courseware and the like on the knowledge point, which all reflect the learning preference and learning interest of students, record and store the data and serve as learning preference data od₅。

And the mastery degree data (recorded as od) of the student to be planned at each knowledge point of the knowledge point set_4，1) The method is mainly used as a direct reference for the path planning of the study, and is the most direct guide for measuring the current study condition of the students.

Of course, since the previous learning situation can reflect the study to a large extentLearning ability and preference of students, in order to make the path planning result more accurate, in a specific embodiment, historical knowledge point mastery degree data (recorded as od) of students to be planned can be obtained_4，2) The historical knowledge point mastering degree data is mainly used as an indirect reference for the learning path planning.

In order to analyze students from more dimensions to obtain more accurate learning planning paths, in one embodiment, basic information data (denoted by od) of the students to be planned can be obtained₁)，

Basic information data od of student₁The data can reflect the learning ability and the local teaching habit of the students to a certain degree, and has certain influence on the learning interest portrayal of the students and the learning of knowledge.

In a specific embodiment, the basic information data representing the student by the vector may be: od₁-(od_1，1，od_1，2，od_1，3，od_1，4) Wherein od_1，1Indicating sex, od_1，2Indicating age, od_1，3Indicating nationality, od_1，4Representing the location of the school.

It should be noted that the corresponding features need to be digitized. For example, gender is represented by 0, 1, 2, 0 is unknown, 1 is male, 2 is female; the age is directly obtained by using actual age figures; the places of the student nationality and the school adopt a mode similar to gender numeralization to process so as to form a better utilized input data format. By standardizing the basic information data of students, the problems of large noise and weak generalization ability can be avoided in the data input process.

It can be seen that the above data respectively reflect the situations of students in learning from different aspects, and these aspects all affect the learning effect of the learning path obtained by planning on the students to be planned.

Step S2: and acquiring at least one of a learning duration prediction vector and a learning interest prediction vector and a learning degree variation prediction vector according to the learning behavior data and the mastering degree data of each knowledge point, wherein each dimension of the learning duration prediction vector represents the learning duration of each knowledge point, each dimension of the learning interest prediction vector represents the learning interest of each knowledge point, and each dimension of the learning degree variation prediction vector represents the mastering degree variation of each knowledge point.

In one embodiment, in order to obtain the learning planning path more accurately, the learning degree of knowledge points of the student may be predicted from multiple aspects, for example, a learning duration prediction vector, a learning interest prediction vector, and a learning degree variation prediction vector may be obtained based on the learning behavior data and the learning degree data of each knowledge point.

Of course, in another specific embodiment, a learning planning path may also be generated by obtaining a learning duration prediction vector and a learning degree variation prediction vector of the student according to the learning behavior data and the mastering degree data of each knowledge point; in other embodiments, the learning planned path may also be generated by obtaining a learning interest prediction vector and a degree of grasp variation prediction vector.

The learning behavior data and the mastery degree data of each knowledge point respectively reflect the conditions of students in learning from different aspects, and can embody the learning behaviors such as the mastery degree of the knowledge points, the learning interest, the learning duration and the like of the students to be planned, so that at least one of the learning duration prediction vector and the learning interest prediction vector and the learning degree variation prediction vector can be obtained according to the learning behavior data and the mastery degree data of each knowledge point.

In an embodiment, in order to improve review efficiency and achieve the purpose of rapidly mastering knowledge points, the step of obtaining the learning duration prediction vector includes; firstly, acquiring an initial prediction vector of learning duration; and then reversely normalizing the learning duration initial prediction vector to obtain the learning duration prediction vector. Because the larger the value is, the larger,the longer the learning time is, the later learning is performed as much as possible, and the conversion formula is as follows:

wherein,

as a slave vector

The selected maximum value.

It is easily understood that, when there is a case where at least one of the learning duration prediction vector and the learning interest prediction vector, and the grasping degree variation amount prediction vector is acquired, the basic information data od of the student to be planned is acquired₁Then according to the basic information data od of the student to be planned₁Learning behavior data and mastery degree data od of each knowledge point₃At least one of the learning duration prediction vector and the learning interest prediction vector, and the grasping degree variation prediction vector are acquired.

In one embodiment, in order to make the path planning result more accurate, personal learning data of the student to be planned may be input into the deep neural network to output at least one of the learning duration prediction vector and the learning interest prediction vector and the mastery degree variation prediction vector. Of course, in other embodiments, the prediction vectors may be obtained by inputting the data into a common model.

Inputting the obtained personal learning data of the students to be planned into a trained deep neural network to obtain a learning interest prediction vector,

the degree of knowledge of the knowledge points changes the prediction vector,

and learning the duration prediction vector,

when training a deep neural network, firstly, acquiring neural network training data, namely learning behavior data training data, mastery degree training data of knowledge points, learning duration vector training data, learning interest vector training data and mastery degree variation vector training data; respectively as input and output of the deep neural network.

And training the neural network based on the neural network training data until the neural network converges to obtain a deep neural network.

Because the output of a given deep neural network is required when training the deep neural network. Since the deep neural network is capable of outputting at least two different predicted targets: at least one of the interest prediction and the learning duration prediction and the knowledge point grasping degree change prediction, so that three different neural network outputs are constructed. Specifically, the method comprises the following steps:

learning interest vector training data (denoted as y)₁)，y₁＝(y_1，1，...，y_1，k，...，y_1，K). Wherein, y_1，kRepresenting the number of times of active search of students on the k-th knowledge point;

mastery degree variation vector training data (denoted as y)₂)，y₂＝(y_2，1，...，y_2，k，...，y_2，K). Wherein, y_2，kThe difference value of the mastery degree of the knowledge points of the student after learning on the kth knowledge point and the mastery degree of the knowledge points before learning is represented, and the larger the value is, the more obvious the representation is improved;

learning duration vector training data (denoted as y)₃)，y₃＝(y_3，1，...，y_3，k，...，y_3，R). Wherein, y_3，kThe total learning time length of the student at the k-th knowledge point is represented, and the larger the value is, the longer the learning time is represented.

Specifically, the learning behavior training data and the mastery degree training data of the knowledge points are input, and the output result of the deep neural network is trained based on the learning duration vector training data, the learning interest vector training data and the mastery degree variation vector training data until the deep neural network converges to obtain the trained deep neural network.

With continued reference to fig. 2, the deep neural network includes a data sharing layer, a knowledge point mastery degree variation vector output layer 100, a knowledge point interest vector output layer 200, and a knowledge point learning duration vector output layer 300. In other embodiments, the deep neural network may also add corresponding output layers according to the target task.

Because the deep neural network comprises the data sharing layer and all data sharing layers of the deep neural network, each target can incorporate information of other targets in the learning process, and the result of a single target is more accurate and more comprehensive. Moreover, because a single target has certain noise to a certain extent, a more generalized representation can be obtained when a plurality of targets are learned simultaneously through different noise modes of different targets, and further the generalization degree and the accuracy of the comprehensive multi-target learning path are improved; furthermore, all models during learning are more concerned about learning important features, because multi-target learning requires 2 or more target tasks to be completed simultaneously, and because targets complement each other, it is more beneficial to learn common core features of multiple targets.

Next, with reference to fig. 2 and 3, a specific construction and learning process of the multitask neural network will be described:

step S21: constructing a deep neural network, as shown in fig. 2, the deep neural network includes: n data sharing layers; the system comprises a knowledge point mastering degree variable quantity vector output layer, a knowledge point interest vector output layer and a knowledge point learning time vector output layer. The number of the neurons of each vector output layer is K, which is the number of the knowledge points. The knowledge point interest vector output layer represents a prediction learning interest task, the knowledge point mastery degree variable quantity vector output layer represents the variation of the knowledge point mastery degree of a prediction student, and the knowledge point learning time duration vector output layer represents the prediction student learning time duration;

step S22: and inputting learning behavior data training data and learning degree variation training data of the knowledge points into the deep neural network.

The learning behavior training data includes learning process training data and learning preference training data. When the variable quantity training data of the mastery degree of the knowledge points comprises the data od of the mastery degree of the student to be planned at each knowledge point in the knowledge point set_4，1And historical knowledge point mastery degree data od of students to be planned_4，2In time, it can be recorded as od₄＝(od_4，1，od_4，2). Of course, in order to make the training result more accurate, the basic information training data of the student to be planned can also be input into the deep neural network, i.e. OD ═ (OD)₁，od₂，od₃，od₄，od₅)。

It is easy to understand that the dimension of the training student data can be selected according to actual conditions, but the more the dimension is, the more accurate the result is.

Step S23: training data y based on learning interest vectors₁Training and learning the deep neural network to obtain the weight W of the data sharing layer_shareAnd learning interest vector output layer weights W₁；

Step S24: training data y based on learning degree variable quantity vector₂Carrying out a new round of training and learning on the deep neural network to obtain new data sharing layer weight W_shareAnd knowledge point mastery degree change prediction vector output layer weight W₂；

Step S25: training data y based on learning time length vector₃Training and learning the deep neural network again to obtain updated data sharing layer weight W_shareAnd knowledge point learning duration prediction vector output layer weight W₃；

Step S26: judging whether the deep neural network is converged, if so, obtaining the trained deep neural network, otherwise, repeating the steps S23 to S26 until the model is converged, and obtaining the final network model parameters of the deep neural network model: w_share，W₁，W₂，W₃Therefore, the finally learned network model parameters can synthesize the information of a plurality of target tasks, and the generalization capability and accuracy of the deep neural network model are improved.

It should be noted that, in the process of training and learning the deep neural network, the sequence of steps S23 to S25 is not limited, but step S23 to step S25 need to be traversed each time of a training loop, that is, the training sequence of three target tasks is not limited, but each training needs to ensure that the three target tasks are trained once, so as to ensure the accuracy of the training result.

By adopting a plurality of target tasks to share a data sharing layer and simultaneously having respective corresponding output layers, a deep neural network is constructed, and the generalization degree and accuracy of the model are obviously improved. More importantly, by simultaneously training a learning interest prediction target, a knowledge point mastering degree change prediction target and a learning duration prediction target related to learning path planning, each target task can incorporate information of other target tasks in the learning process, the result of a single target is more accurate and more comprehensive, and the robustness of the model is improved.

Step S3: performing vector fusion on at least one of the learning duration prediction vector and the learning interest prediction vector and the mastering degree variation prediction vector to obtain an integrated knowledge point evaluation vector;

deriving learning interest prediction vectors

Knowledge point mastery degree change prediction vector

Learning duration prediction vector

Then, vector fusion is carried out on the comprehensive knowledge point vector so as to obtain more comprehensive and accurate comprehensive knowledge point vector

An individualized learning path l capable of comprehensively considering a plurality of factors (learning duration, learning interest and knowledge point mastering degree variation) is planned for the target students_uThe learning efficiency and the learning interest of the final students are ensured, and the accuracy and the comprehensiveness of the path planning are obviously improved.

In particular, the comprehensive knowledge point evaluation vector of the student to be planned

Can be calculated by the following formula:

wherein, beta₁Represents the degree of importance of the learning interest in the final path planning, β₂Represents the degree of importance of the variable quantity of the knowledge point mastery degree in the final path planning, beta₃Indicating how important the learning duration is in the final path planning.

In addition, β is₁，β₂And beta₃The weight of (b) is determined by the actual situation, and if learning interest is more concerned, then β can be set₁Set to a value of beta₂And beta₃Large; similarly, assuming more attention is paid to learning duration, β will be₂Set to a value of beta₁And beta₃Is large.

Step S4: and sequencing the knowledge points according to the size of each dimension of the comprehensive knowledge point evaluation vector to obtain the learning planning path of each knowledge point.

In a specific embodiment, the result is obtained based on step S3

The values can be sorted from big to small, and the sorted knowledge points are the learning path l planned for the student u_u。

Thus, one strip can be obtainedPersonalized learning path l capable of comprehensively considering multiple factors (learning duration, learning interest and knowledge point promotion degree)_uThe learning efficiency and the learning interest of students are ensured, and the accuracy and the comprehensiveness of path planning are obviously improved.

Therefore, the learning path planning method provided by the embodiment of the invention can comprehensively consider at least two factors influencing the review effect of the students, thereby ensuring the final learning efficiency and learning interest of the students, improving the accuracy and comprehensiveness of the path planning, and better improving the learning efficiency and learning interest of the students.

Referring to fig. 4, in an embodiment, the step of acquiring learning behavior data of a student to be planned may include:

step S11: acquiring the learning behavior original data of the student to be planned;

the learning behavior is disordered in original data, the quality is uneven, the data size is large, the model is difficult to directly calculate, if the model is directly used, the calculation amount of the algorithm is increased, and the precision of the model cannot be guaranteed. Thus, the raw chaotic data needs to be standardized. Through the conversion of statistical characteristics, the original data can be standardized, and the application range of the data is greatly expanded.

Step S12: and carrying out statistical analysis on the learning behavior original data to obtain the learning behavior data of the student to be planned.

By carrying out statistical processing on the learning behavior original data, the data volume is reduced, and the calculation efficiency of the model can be improved.

Because the original data of the learning behaviors are normalized in advance, the learning behavior data obtained through statistical analysis is clean and has understandable statistical characteristics, and the normalized data with the added statistical characteristics can more accurately and comprehensively reflect the learning ability, learning interest and learning habit of students, so that the standardization degree of the data and the accuracy of the result are improved, the accuracy of final prediction is improved, and the accuracy and the operation efficiency of path planning are improved.

In a specific embodiment, at least one feature quantity of the maximum value, the minimum value, the mean value, the median, the variance, the standard value, the 25% quantile and the 75% quantile of the raw data of the learning behavior can be counted to obtain a statistical feature vector;

since it is difficult to intuitively understand the actual learning condition of the student from the confused raw data, it is difficult to accurately understand the actual learning process and learning state of the student based on the raw data. The original learning situation data are difficult to directly reflect the learning ability of students, and the learning process of the students can be more comprehensively and accurately evaluated and measured. The learning condition of the student cannot be reflected by the statistical characteristics, such as the learning duration of the student on each knowledge point. For example, the variance can intuitively understand the learning stability of the student. If the variance is large, the fluctuation of the students is large and unstable in the learning process. Also like the mean value, if the mean value is smaller, the student has better learning ability, and can learn a knowledge point faster. The stability and the overall level of the student learning can be intuitively known through the variance and the mean.

And then carrying out vector splicing on each statistical characteristic vector to obtain the learning behavior data of the student to be planned.

When the learning behavior data of the student to be planned comprises learning process data (marked as od)₂) And learning preference data (denoted od)₅) First, the learning process data od is₂Statistical analysis was performed.

Data od based on student learning process₂And calculating corresponding statistical characteristics by adopting a conventional statistical method. Specifically, the maximum value, the minimum value, the mean value, the median, the variance, the standard deviation, the 25% quantile, and the 75% quantile are calculated for the total learning duration at each knowledge point. And respectively counting the statistical characteristics of other characteristics such as the number of questions, the course completion degree, the single learning time, the activity degree and the initiative degree by adopting the same method: maximum, minimum, mean, median, variance, standard deviation, 25% quantile, 75% quantile. Then, the statistical features counted on the features are spliced to form learningProcess integration data vector cd₂＝(cd_2，1，cd_2，2，cd_2，2，cd_2，4，cd_2，5，cd_2，6). Wherein cd_2，1Statistical feature vector, cd, representing total learning duration_2，2Statistical feature vector, cd, representing number of questions made_2，3Statistical feature vector, cd, representing the degree of completion of a course_2，4Statistical feature vector, cd, representing a single learning time_2，5Statistical feature vector, cd, representing the degree of activity_2，6A statistical feature vector representing the degree of aggressiveness. Lifting cd_2，1For example, the statistical feature vector is embodied in the form of: CD (compact disc)_2，1＝(cd_2，1，1，cd_2，1，2，cd_2，1，3，cd_2，1，4，cd_2，1，5，cd_2，1，6，cd_2，1，7，cd_2，1，8). Wherein cd_2，1，1Denotes the maximum value, cd_2，1，2Denotes the minimum value, cd_2，1，3Denotes mean value, cd_2，1，4Denotes the median, cd_2，1，5Represents variance, cd_2，1，6Denotes standard deviation, cd_2，1，7Denotes 25% quantile, cd_2，1，8Representing a 75% quantile.

In the same way, based on student learning preference data od₅Calculating corresponding statistical characteristics, od, by conventional statistical method₅＝cd₅＝(cd_5，1，cd_5，2)。

Wherein cd_5，1Comprehensive data vector, cd, representing learning conditions_5，2A synthetic data vector representing the behavior of the active operation. Specifically, the method comprises the following steps:

learning context integration data vector, cd_5，1-(cd_5，1，1，cd_5，1，2). The vector is mainly solved based on the prior mastery degree of the students on each knowledge point, because the learning ability and the learning preference of the students have large difference. Some students are good at learning the class knowledge points, and some students are good at applying the class knowledge points. Also, some students prefer knowledge points with low difficulty, while some tend to challenge the difficultyHigh knowledge points. Based on the two observations, two statistical feature vectors based on knowledge point capability level and knowledge point difficulty are respectively constructed.

Statistical feature vector, cd, based on knowledge point capability hierarchy_5，1，1. The specific solving method is to divide the knowledge points based on the capability level (i.e. memorization, understanding, application and synthesis), and then to solve the corresponding for each level by adopting a statistical method for the divided knowledge points: maximum, minimum, mean, median, variance, standard deviation, 25% quantile, 75% quantile. Finally, a 32-dimensional (4 capability levels × 8 statistical features per capability level, i.e., 32-dimensional) statistical feature vector cd can be obtained_5，1，1；

Statistical feature vector, cd, based on knowledge point difficulty_5，1，2Firstly, the difficulty of knowledge point is set to [0, 1]]The continuous value between the two values becomes a discrete value from 1 to 5, and the specific mapping method is as follows: [0,0.2]Corresponding difficulty 1, [0.2, 0.4 ]]The corresponding difficulty is 2, [0.4, 0.6 ]]The corresponding difficulty is 3, [0.6, 0.8 ]]The corresponding difficulty is 4, [0.8, 1]The correspondence difficulty is 5. Then, a vector acquisition method similar to the 'statistical feature vector based on knowledge point capability hierarchy' is adopted to respectively solve corresponding problems for each difficulty level: maximum, minimum, mean, median, variance, standard deviation, 25% quantile, 75% quantile. Finally, a 40-dimensional (5 difficulties x 8 statistical features per difficulty, i.e., 40-dimensional) statistical feature vector cd can be obtained_5，1，2。

It should be noted that, in this embodiment, the difficulty level is a discrete value of 1 to 5, and in other embodiments, the difficulty level may be defined according to a specific situation, for example, a continuous value between [0, 1] and [ 10 ] of the difficulty of the knowledge point may be changed into a discrete value of 1 to 6 or 1 to 10.

Active operation behavior integrated data vector, cd_5，2. Integrated data vector cd for active operation behavior_5，2The more times of the behaviors of the students actively operating on the specific knowledge points are, the more the behaviors are, the more the students actively search for questions or actively learn more courseware and the like in the learning process of the existing knowledge points are, the more the behaviors areIndicating that the student has a greater interest and initiative in learning the knowledge point.

Specifically, the proactive operation behavior integrated data vector cd is similar to the learning situation integrated vector_5，2Respectively counting the statistical characteristic vector cd based on the knowledge point capability level based on the times of active operation behaviors_5，2，1And statistical feature vector cd based on knowledge point difficulty_5，2，2I.e. cd_5，2＝(cd_5，2，1，cd_5，2，2)。

Therefore, the student learning preference data obtained based on the statistical characteristics standardizes the original chaotic data on one hand; on the other hand, the data volume is reduced, and the model operation efficiency is improved; and finally, the learning process of the students can be known more comprehensively and accurately.

Compared with the defects of disordered data, multiple dimensions, difficulty in accurately depicting the learning ability and the learning interest of students and the like in the data using process of the existing path planning, a large number of standardized statistical feature vectors are extracted from the learning process data and the learning preference data by adopting a statistical feature method: at least one of the maximum value, the minimum value, the mean value, the median, the variance, the standard deviation, the 25% quantile and the 75% quantile finally obtains a set of clean, low-noise learning behavior data which can be processed by a neural network and can visually reflect the learning ability, habits and preferences of students.

In a specific embodiment, the step of obtaining the historical knowledge point mastery degree data of the student to be planned includes:

acquiring historical knowledge point mastery degree original data od of the student to be planned_4，2；

Carrying out statistical analysis on the original data of the mastery degree of the historical knowledge points of the students to be planned to obtain the data cd of the mastery degree of the historical knowledge points of the students to be planned_4，2。

Because the students learn a lot of knowledge points before, the students are difficult to directly apply to the neural network model. Meanwhile, it is difficult to intuitively know the specific learning ability of the student based on the original, and the degree of mastering of the historical knowledge pointsData cd_4，2The statistical transformation is carried out based on the mastering degree of students on each knowledge point before, so that more visual statistical characteristics can be obtained through the statistical transformation.

Specifically, the step of performing statistical analysis on the raw data of the historical knowledge point mastery degree of the student to be planned to obtain the historical knowledge point mastery degree data of the student to be planned includes:

calculating one or more statistical characteristic quantities in the maximum value, the minimum value, the mean value, the median, the variance, the standard value, the 25% quantile and the 75% quantile of the original data of the grasping degree of the historical knowledge points to obtain statistical characteristic vectors of the historical knowledge points;

and carrying out vector splicing on the statistical characteristic vectors of the historical knowledge points to obtain the mastery degree data of the historical knowledge points of the students to be planned.

Wherein cd_4，2，1Denotes the maximum value, cd_4，2，2Denotes the minimum value, cd_4，2，3Denotes mean value, cd_4，2，4Denotes the median, cd_4，2，5Represents variance, cd_4，2，6Denotes standard deviation, d_4，2，7Denotes 25% quantile, cd_4，2，8Representing a 75% quantile. Namely cd_4，2＝(cd_4，2，1，cd_4，2，2，cd_4，2，3，cd_4，2，4，cd_4，2，5，cd_4，2，6，cd_4，2，7，cd_4，2，8). For example, based on the variance cd_4，2，5The stability of the learning condition of the students can be known, and the smaller the value is, the smaller fluctuation of the learning of the students is, and the more stable the learning is.

In the following, the learned route planning apparatus provided by the embodiment of the present invention is introduced, and the learned route planning apparatus described below may be considered as a functional module architecture that is required to be set by an electronic device (e.g., a PC) to respectively implement the learned route planning method provided by the embodiment of the present invention. The contents of the learned route planning apparatus described below may be referred to in correspondence with the contents of the learned route planning method described above, respectively.

Fig. 5 is a block diagram of a learned route planning apparatus according to an embodiment of the present invention, where the learned route planning apparatus is applicable to both a client and a server, and referring to fig. 5, the learned route planning apparatus may include:

a learning data acquisition module 10, configured to acquire a knowledge point set to be subjected to path planning, learning behavior data of students to be planned, and mastery degree data of the students to be planned at each knowledge point in the knowledge point set;

a prediction vector obtaining module 20, configured to obtain, according to the learning behavior data and the mastery degree data of each knowledge point, at least one of a learning duration prediction vector and a learning interest prediction vector, and a mastery degree variation prediction vector, where each dimension of the learning duration prediction vector represents a learning duration of each knowledge point, each dimension of the learning interest prediction vector represents a learning interest of each knowledge point, and each dimension of the mastering degree variation prediction vector represents a variation in the mastery degree of each knowledge point;

an integrated knowledge point estimation vector obtaining module 30, configured to perform vector fusion on the learning duration prediction vector and the learning interest prediction vector, and the learning degree variation prediction vector, to obtain an integrated knowledge point estimation vector;

and the learning path planning generation module 40 is configured to sort the knowledge points according to the size of each dimension of the comprehensive knowledge point evaluation vector, so as to obtain a learning planning path of each knowledge point.

Firstly, acquiring a knowledge point set to be path planned, learning behavior data of students to be planned and mastery degree data of the students to be planned at each knowledge point of the knowledge point set according to a learning data acquisition module 10;

the knowledge point set to be planned for the path may include each specific knowledge point and knowledge point basic information mainly used for measuring basic attributes of the current knowledge points to be planned.

To improveThe accuracy of the knowledge points, in one embodiment, the basic information of the knowledge points may include difficulty of the knowledge points and capability level of investigation of the knowledge points, i.e. a set of knowledge points (denoted by od) for path planning₃) Each knowledge point in (1) slave knowledge point difficulty (denoted as od)_3，1) And level of knowledge point investigation (denoted od)_3，2) Two aspects are defined:

for this purpose, the set of knowledge points to be path-planned may be represented by a vector as: od₃＝(od_3，1，od_3，2). Wherein od_3，1＝(od_3，1，1，od_3，1，2，...，od_3，1，k，...，od_3，1，K)，od_3，1，kRepresenting the difficulty of the k-th knowledge point, wherein the value is from 0 to 1, and the larger the value is, the higher the difficulty of learning the knowledge point is; od_3，2＝(od_3，2，1，od_3，2，2，...，od_3，2，k，...，od_3，2，K)，od_3，2，kAnd the capability level of the k-th knowledge point investigation is represented, the value range is one of {1, 2, 3 and 4}, 1 represents memorization, 2 represents understanding, 3 represents application, and 4 represents synthesis. The larger the value, the higher the level of capability that represents the investigation. In addition, K represents the total number of knowledge points to be planned; of course, this is merely an example, and in other embodiments, a person skilled in the art may define the capability level of the knowledge point investigation according to the actual situation.

It is easy to understand that the learning behavior data of the student to be planned refers to behavior data generated by the student in learning each knowledge point, and may include total learning duration, number of questions to be made, course completion condition, single learning time, activity level (measured based on interaction with resources and other students), activity level (based on the active behavior of the student on the information system, such as searching for questions, asking questions, etc.), and the like.

In order to analyze the learning behavior of the student from multiple dimensions, in one embodiment, the learning behavior data of the student to be planned may include learning process data (denoted by od)₂) And learning preference data (denoted od)₅) The learning process dataod₂Representing process data generated by students to be planned in the process of learning each knowledge point, the learning preference data od₅And the data represent the learning preference and interest data of the students to be planned in learning each knowledge point.

Wherein the learning process data od₂Can reflect the learning habits, personal characters and psychological characteristics of students, for example, students with high initiative degree can actively take the initiative, and relatively speaking, the learning ability is stronger, because the students can actively search resources and seek help. For another example, the duration of a single study can reflect the concentration degree of students, and some students leave the system to a certain extent in a short time each time of study, which indicates that the concentration degree of the students is low.

Learning preference data od₅Mainly comes from the learning condition and the active operation behavior of the students to be planned. The learning condition refers to the final mastery degree of students on different knowledge, and knowledge points with high mastery degree are relatively large in learning preference and good at the knowledge points; the active operation behavior mainly refers to the behavior of actively searching for questions or actively learning more courseware and the like on the knowledge point, which all reflect the learning preference and learning interest of students, record and store the data and serve as learning preference data od₅。

And the mastery degree data (recorded as od) of the student to be planned at each knowledge point of the knowledge point set_4，1) The method is mainly used as a direct reference for the path planning of the study, and is the most direct guide for measuring the current study condition of the students.

Of course, since the previous learning condition can reflect the learning ability and preference of the student to a greater extent, in order to make the path planning result more accurate, in a specific embodiment, the learning data acquisition module 10 is further adapted to acquire the historical knowledge point mastering degree data (denoted by od) of the student to be planned_4，2) The historical knowledge point mastering degree data is mainly used as an indirect reference for the learning path planning.

To analyze students from more dimensions to obtain more accurate learned planned paths, in one embodimentThe learning data acquisition module 10 may also acquire basic information data (denoted as od) of the student to be planned₁)，

Basic information data od of student₁The data can reflect the learning ability and the local teaching habit of the students to a certain degree, and has certain influence on the learning interest portrayal of the students and the learning of knowledge.

In a specific embodiment, the basic information data representing the student by the vector may be: od₁-(od_1，1，od_1，2，od_1，3，od_1，4) Wherein od_1，1Indicating sex, od_1，2Indicating age, od_1，3Indicating nationality, od_1，4Representing the location of the school.

It should be noted that the corresponding features need to be digitized. For example, gender is represented by 0, 1, 2, 0 is unknown, 1 is male, 2 is female; the age is directly obtained by using actual age figures; the places of the student nationality and the school adopt a mode similar to gender numeralization to process so as to form a better utilized input data format. By standardizing the basic information data of students, the problems of large noise and weak generalization ability can be avoided in the data input process.

It can be seen that, by using the learning data acquiring module 10 to acquire the above data, the above data respectively reflects the situation of the student in learning from different aspects, and these aspects all affect the learning effect of the learning path obtained by planning on the student to be planned.

Secondly, a prediction vector obtaining module 20 is adopted to obtain at least one of a learning duration prediction vector and a learning interest prediction vector and a learning degree variation prediction vector according to the learning behavior data and the mastering degree data of each knowledge point, wherein each dimension of the learning duration prediction vector represents the learning duration of each knowledge point, each dimension of the learning interest prediction vector represents the learning interest of each knowledge point, and each dimension of the mastering degree variation prediction vector represents the mastering degree variation of each knowledge point.

In one embodiment, in order to obtain the learning planning path more accurately, the prediction vector obtaining module 20 may predict the degree of mastery of the knowledge points of the student from multiple aspects, for example, may obtain a learning duration prediction vector, a learning interest prediction vector, and a degree of variation prediction vector according to the learning behavior data and the degree of mastery data of each knowledge point.

Of course, in another embodiment, the prediction vector obtaining module 20 may also obtain a learning duration prediction vector and a learning degree variation prediction vector of the student according to the learning behavior data and the mastering degree data of each knowledge point to generate a learning planning path; in other embodiments, the prediction vector obtaining module 20 may also generate the learning planned path by obtaining a learning interest prediction vector and a degree of mastery change prediction vector.

The learning behavior data and the mastery degree data of each knowledge point respectively reflect the conditions of students in learning from different aspects, and can embody the learning behaviors such as the mastery degree of the knowledge points, the learning interest, the learning duration and the like of the students to be planned, so that at least one of a learning duration prediction vector and a learning interest prediction vector and a learning degree variation prediction vector can be obtained according to the learning behavior data and the mastery degree data of each knowledge point.

In one embodiment, in order to improve review efficiency and achieve the purpose of fast mastering knowledge points, the prediction vector obtaining module 20 is adapted to first obtain an initial prediction vector of a learning duration; and then reversely normalizing the learning duration initial prediction vector to obtain the learning duration prediction vector. Because the larger the value, the longer the learning time is, and the later learning is tried to be put, the conversion formula is:

wherein,

as a slave vector

The selected maximum value.

It is easy to understand that when the learning data acquisition module 10 can acquire basic information data od of a student to be planned₁The prediction vector acquisition module 20 is adapted to obtain the basic information data od of the student to be planned₁Learning behavior data and mastery degree data od of each knowledge point₃At least one of the learning duration prediction vector and the learning interest prediction vector, and the grasping degree variation prediction vector are acquired.

In an embodiment, in order to make the path planning result more accurate, the learned path planning apparatus according to the embodiment of the present invention further includes a deep neural network module, and the deep neural network module obtains at least one of the output learning duration prediction vector and the learning interest prediction vector and the grasping degree variation prediction vector. Of course, in other embodiments, the prediction vectors may be obtained by inputting the data into a common model.

Inputting the obtained personal learning data of the students to be planned into a trained deep neural network to obtain a learning interest prediction vector,

the degree of knowledge of the knowledge points changes the prediction vector,

and learning the duration prediction vector,

because the deep neural network comprises the data sharing layer and all data sharing layers of the deep neural network, each target can incorporate information of other targets in the learning process, and the result of a single target is more accurate and more comprehensive. Moreover, because a single target has certain noise to a certain extent, a more generalized representation can be obtained when a plurality of targets are learned simultaneously through different noise modes of different targets, and further the generalization degree and the accuracy of the comprehensive multi-target learning path are improved; furthermore, all models during learning are more concerned about learning important features, because multi-target learning requires 2 or more target tasks to be completed simultaneously, and because targets complement each other, it is more beneficial to learn common core features of multiple targets.

By adopting a plurality of target tasks to share a data sharing layer and simultaneously having respective corresponding output layers, a deep neural network is constructed, and the generalization degree and accuracy of the model are obviously improved. More importantly, by simultaneously training a learning interest prediction target, a knowledge point mastering degree change prediction target and a learning duration prediction target related to learning path planning, each target task can incorporate information of other target tasks in the learning process, the result of a single target is more accurate and more comprehensive, and the robustness of the model is improved.

Performing vector fusion on at least one of the learning duration prediction vector and the learning interest prediction vector and the grasping degree variation prediction vector by using an integrated knowledge point evaluation vector acquisition module 30 to acquire an integrated knowledge point evaluation vector;

deriving learning interest prediction vectors

Knowledge point mastery degree change prediction vector

Learning duration prediction vector

Then, the comprehensive knowledge point evaluation vector acquisition module 30 is used for carrying out vector fusion on the comprehensive knowledge point evaluation vector acquisition module, so that more comprehensive and accurate comprehensive knowledge is obtainedKnowledge point vector

An individualized learning path l capable of comprehensively considering a plurality of factors (learning duration, learning interest and knowledge point mastering degree variation) is planned for the target students_uThe learning efficiency and the learning interest of the final students are ensured, and the accuracy and the comprehensiveness of the path planning are obviously improved.

In particular, the comprehensive knowledge point evaluation vector of the student to be planned

Can be calculated by the following formula:

wherein, beta₁Represents the degree of importance of the learning interest in the final path planning, β₂Represents the degree of importance of the variable quantity of the knowledge point mastery degree in the final path planning, beta₂Indicating how important the learning duration is in the final path planning.

In addition, β is₁，β₂And beta₃The weight of (b) is determined by the actual situation, and if learning interest is more concerned, then β can be set₁Set to a value of beta₂And beta₃Large; similarly, assuming more attention is paid to learning duration, β will be₂Set to a value of beta₁And beta₃Is large.

And sequencing the knowledge points according to the dimension of the comprehensive knowledge point evaluation vector through a learning path planning generation module 40 to obtain a learning planning path of each knowledge point.

In a specific embodiment, the result is obtained based on step S3

The values in the data can be sorted from big to small, and the sorted knowledge points areFor a learning path l planned for a student u_u。

Thus, an individual learning path l capable of comprehensively considering a plurality of factors (learning duration, learning interest and knowledge point promotion degree) can be obtained_uThe learning efficiency and the learning interest of students are ensured, and the accuracy and the comprehensiveness of path planning are obviously improved.

Therefore, the learning path planning device provided by the embodiment of the invention can comprehensively consider at least two factors influencing the review effect of the students, thereby ensuring the final learning efficiency and learning interest of the students, improving the accuracy and comprehensiveness of the path planning, and better improving the learning efficiency and learning interest of the students.

In a specific embodiment, the learning data obtaining module 10 is further adapted to obtain the learning behavior raw data of the student to be planned, and perform statistical analysis on the learning behavior raw data to obtain the learning behavior data of the student to be planned.

Because the original data of the learning behaviors are normalized in advance, the learning behavior data obtained through statistical analysis is clean and has understandable statistical characteristics, and the normalized data with the added statistical characteristics can more accurately and comprehensively reflect the learning ability, learning interest and learning habit of students, so that the standardization degree of the data and the accuracy of the result are improved, the accuracy of final prediction is improved, and the accuracy and the operation efficiency of path planning are improved.

In a specific embodiment, at least one feature quantity of the maximum value, the minimum value, the mean value, the median, the variance, the standard value, the 25% quantile and the 75% quantile of the raw data of the learning behavior can be counted to obtain a statistical feature vector;

since it is difficult to intuitively understand the actual learning condition of the student from the confused raw data, it is difficult to accurately understand the actual learning process and learning state of the student based on the raw data. The original learning situation data are difficult to directly reflect the learning ability of students, and the learning process of the students can be more comprehensively and accurately evaluated and measured. The learning condition of the student cannot be reflected by the statistical characteristics, such as the learning duration of the student on each knowledge point. For example, the variance can intuitively understand the learning stability of the student. If the variance is large, the fluctuation of the students is large and unstable in the learning process. Also like the mean value, if the mean value is smaller, the student has better learning ability, and can learn a knowledge point faster. The stability and the overall level of the student learning can be intuitively known through the variance and the mean.

And then carrying out vector splicing on each statistical characteristic vector to obtain the learning behavior data of the student to be planned.

When the learning behavior data of the student to be planned comprises learning process data (marked as od)₂) And learning preference data (denoted od)₅) First, the learning process data od is₂Statistical analysis was performed.

Data od based on student learning process₂And calculating corresponding statistical characteristics by adopting a conventional statistical method. Specifically, the maximum value, the minimum value, the mean value, the median, the variance, the standard deviation, the 25% quantile, and the 75% quantile are calculated for the total learning duration at each knowledge point. And respectively counting the statistical characteristics of other characteristics such as the number of questions, the course completion degree, the single learning time, the activity degree and the initiative degree by adopting the same method: maximum, minimum, mean, median, variance, standard deviation, 25% quantile, 75% quantile. Then, the statistical features counted on the features are spliced to form a learning process comprehensive data vector cd₂＝(cd_2，1，cd_2，2，cd_2，2，cd_2，4，cd_2，5，cd_2，6)。

In the same way, based on student learning preference data od₅Calculating corresponding statistical characteristics, od, by conventional statistical method₅＝cd₅＝(cd_5，1，cd_5，2)。

Wherein cd_5，1Comprehensive data vector, cd, representing learning conditions_5，2A synthetic data vector representing the behavior of the active operation. Specifically, the method comprises the following steps:

learning context integration data vector, cd_5，1＝(cd_5，1，1，cd_5，1，2). The vector is mainly solved based on the prior mastery degree of the students on each knowledge point, because the learning ability and the learning preference of the students have large difference. Some students are good at learning the class knowledge points, and some students are good at applying the class knowledge points. Also, some students prefer knowledge points with low difficulty, while some tend to challenge knowledge points with high difficulty. Based on the two observations, two statistical feature vectors based on knowledge point capability level and knowledge point difficulty are respectively constructed.

Statistical feature vector, cd, based on knowledge point capability hierarchy_5，1，1. The specific solving method is to divide the knowledge points based on the capability level (i.e. memorization, understanding, application and synthesis), and then to solve the corresponding for each level by adopting a statistical method for the divided knowledge points: maximum, minimum, mean, median, variance, standard deviation, 25% quantile, 75% quantile. Finally, a 32-dimensional (4 capability levels × 8 statistical features per capability level, i.e., 32-dimensional) statistical feature vector cd can be obtained_5，1，1；

Statistical feature vector, cd, based on knowledge point difficulty_5，1，2Firstly, the difficulty of knowledge point is set to [0, 1]]The continuous value between the two values becomes a discrete value from 1 to 5, and the specific mapping method is as follows: [0,0.2]Corresponding difficulty 1, [0.2, 0.4 ]]The corresponding difficulty is 2, [0.4, 0.6 ]]The corresponding difficulty is 3, [0.6, 0.8 ]]The corresponding difficulty is 4, [0.8, 1]The correspondence difficulty is 5. Then, a vector acquisition method similar to the 'statistical feature vector based on knowledge point capability hierarchy' is adopted to respectively solve corresponding problems for each difficulty level: maximum, minimum, mean, median, variance, standard deviation, 25% quantile, 75% quantile. Finally, a 40-dimensional (5 difficulties x 8 statistical features per difficulty, i.e., 40-dimensional) statistical feature vector cd can be obtained_5，1，2。

It should be noted that, in this embodiment, the difficulty level is a discrete value of 1 to 5, and in other embodiments, the difficulty level may be defined according to a specific situation, for example, a continuous value between [0, 1] and [ 10 ] of the difficulty of the knowledge point may be changed into a discrete value of 1 to 6 or 1 to 10.

Active operation behavior integrated data vector, cd_5，2. Integrated data vector cd for active operation behavior_5，2The learning method is mainly based on the behaviors of actively searching questions or actively learning more courseware and the like in the learning process of the existing knowledge points by students, the more times of the behaviors of the students actively operating on the specific knowledge points are, and the greater interest and initiative of the students in learning the knowledge points are shown to a certain extent.

Specifically, the proactive operation behavior integrated data vector cd is similar to the learning situation integrated vector_5，2Respectively counting the statistical characteristic vector cd based on the knowledge point capability level based on the times of active operation behaviors_5，2，1And statistical feature vector cd based on knowledge point difficulty_5，2，2I.e. cd_5，2＝(cd_5，2，1，cd_5，2，2)。

Therefore, the student learning preference data obtained based on the statistical characteristics standardizes the original chaotic data on one hand; on the other hand, the data volume is reduced, and the model operation efficiency is improved; and finally, the learning process of the students can be known more comprehensively and accurately.

Compared with the defects of disordered data, multiple dimensions, difficulty in accurately depicting the learning ability and the learning interest of students and the like in the data using process of the existing path planning, a large number of standardized statistical feature vectors are extracted from the learning process data and the learning preference data by adopting a statistical feature method: at least one of the maximum value, the minimum value, the mean value, the median, the variance, the standard deviation, the 25% quantile and the 75% quantile finally obtains a set of clean, low-noise learning behavior data which can be processed by a neural network and can visually reflect the learning ability, habits and preferences of students.

In one embodiment, the learning data acquisition module 10 is further adapted to acquire historical knowledge point mastery degree raw data od of the student to be planned_4，2(ii) a The historical knowledge point mastery degree original data of the students to be planned are inputPerforming statistical analysis to obtain historical knowledge point mastery degree data cd of the students to be planned_4，2。

Because the students learn a lot of knowledge points before, the students are difficult to directly apply to the neural network model. Meanwhile, there is also a problem that it is difficult to intuitively understand the concrete learning ability of the student based on the original, and the historical knowledge point mastery degree data cd_4，2The statistical transformation is carried out based on the mastering degree of students on each knowledge point before, so that more visual statistical characteristics can be obtained through the statistical transformation.

Specifically, the step of performing statistical analysis on the raw data of the historical knowledge point mastery degree of the student to be planned to obtain the historical knowledge point mastery degree data of the student to be planned includes:

calculating one or more statistical characteristic quantities in the maximum value, the minimum value, the mean value, the median, the variance, the standard value, the 25% quantile and the 75% quantile of the original data of the grasping degree of the historical knowledge points to obtain statistical characteristic vectors of the historical knowledge points;

for example, based on the variance cd_4，2，5The stability of the learning condition of the students can be known, and the smaller the value is, the smaller fluctuation of the learning of the students is, and the more stable the learning is.

And then, carrying out vector splicing on the statistical characteristic vectors of the historical knowledge points to obtain the mastery degree data of the historical knowledge points of the students to be planned.

The learning path planning device provided by the embodiment of the invention can comprehensively consider at least two factors influencing the learning of students, thereby improving the accuracy and the comprehensiveness of path planning, ensuring that the finally planned learning path has the advantages of high comprehensiveness, strong pertinence, expected accuracy and the like, and better improving the learning efficiency and the learning interest of the students.

The embodiment of the invention also provides equipment, which comprises at least one memory and at least one processor; the memory stores programs, and the processor calls the programs to execute the learning path planning method, so that the path planning accuracy and the path planning comprehensiveness can be improved, the finally planned learning path is ensured to have the advantages of high comprehensiveness, strong pertinence, expected accuracy and the like, and the learning efficiency and the learning interest of students are better improved.

The embodiment of the invention also provides a storage medium which stores a program suitable for learning path planning so as to realize the learning path planning method, can improve the accuracy and the comprehensiveness of path planning, ensures that the finally planned learning path has the advantages of high comprehensiveness, strong pertinence, accuracy and the like, and realizes better improvement of the learning efficiency and the learning interest of students.

The embodiments of the present invention described above are combinations of elements and features of the present invention. Unless otherwise mentioned, the elements or features may be considered optional. Each element or feature may be practiced without being combined with other elements or features. In addition, the embodiments of the present invention may be configured by combining some elements and/or features. The order of operations described in the embodiments of the present invention may be rearranged. Some configurations of any embodiment may be included in another embodiment, and may be replaced with corresponding configurations of the other embodiment. It is obvious to those skilled in the art that claims that are not explicitly cited in each other in the appended claims may be combined into an embodiment of the present invention or may be included as new claims in a modification after the filing of the present application.

Embodiments of the invention may be implemented by various means, such as hardware, firmware, software, or a combination thereof. In a hardware configuration, the method according to an exemplary embodiment of the present invention may be implemented by one or more Application Specific Integrated Circuits (ASICs), Digital Signal Processors (DSPs), Digital Signal Processing Devices (DSPDs), Programmable Logic Devices (PLDs), Field Programmable Gate Arrays (FPGAs), processors, controllers, micro-controllers, microprocessors, and the like.

In a firmware or software configuration, embodiments of the present invention may be implemented in the form of modules, procedures, functions, and the like. The software codes may be stored in memory units and executed by processors. The memory unit is located inside or outside the processor, and may transmit and receive data to and from the processor via various known means.

The previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the invention. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Although the embodiments of the present invention have been disclosed, the present invention is not limited thereto. Various changes and modifications may be effected therein by one skilled in the art without departing from the spirit and scope of the invention as defined in the appended claims.

Claims (14)

1. A method of learning path planning, comprising:

acquiring a knowledge point set to be subjected to path planning, learning behavior data of students to be planned and mastery degree data of the students to be planned at each knowledge point of the knowledge point set;

acquiring at least one of a learning duration prediction vector and a learning interest prediction vector and a learning degree variation prediction vector according to the learning behavior data and the mastering degree data of each knowledge point, wherein each dimension of the learning duration prediction vector represents the learning duration of each knowledge point, each dimension of the learning interest prediction vector represents the learning interest of each knowledge point, and each dimension of the learning degree variation prediction vector represents the mastering degree variation of each knowledge point;

performing vector fusion on at least one of the learning duration prediction vector and the learning interest prediction vector and the mastering degree variation prediction vector to obtain an integrated knowledge point evaluation vector;

and sequencing the knowledge points according to the size of each dimension of the comprehensive knowledge point evaluation vector to obtain the learning planning path of each knowledge point.

2. The learning path planning method according to claim 1, wherein the step of acquiring learning behavior data of the student to be planned includes:

acquiring the learning behavior original data of the student to be planned;

and carrying out statistical analysis on the learning behavior original data to obtain the learning behavior data of the student to be planned.

3. The method for planning a learning path according to claim 2, wherein the step of performing statistical analysis on the raw data of the learning behavior to obtain the learning behavior data of the student to be planned comprises:

counting at least one characteristic quantity in the maximum value, the minimum value, the mean value, the median, the variance, the standard value, the 25% quantile and the 75% quantile of the learning behavior original data to obtain a statistical characteristic vector;

and carrying out vector splicing on each statistical characteristic vector to obtain the learning behavior data of the student to be planned.

4. The learned path planning method of claim 1, further comprising:

acquiring historical knowledge point mastery degree data of the students to be planned;

the step of obtaining at least one of a learning duration prediction vector and a learning interest prediction vector according to the learning behavior data and the mastery degree data of each knowledge point, and the learning degree variation prediction vector includes:

and acquiring at least one of a learning duration prediction vector and a learning interest prediction vector and a learning degree variation prediction vector according to the learning behavior data, the historical knowledge point mastery degree data and the mastery degree data of each knowledge point.

5. The learned path planning method according to claim 4, wherein the step of obtaining the historical knowledge point mastery degree data of the student to be planned includes:

acquiring historical knowledge point mastery degree original data of the students to be planned;

and carrying out statistical analysis on the original data of the mastery degree of the historical knowledge points of the students to be planned to obtain the mastery degree data of the historical knowledge points of the students to be planned.

6. The learned path planning method according to claim 1, wherein the step of obtaining the learned duration prediction vector includes;

acquiring an initial prediction vector of learning duration;

and reversely normalizing the learning duration initial prediction vector to obtain the learning duration prediction vector.

7. The learned path planning method of claim 1, wherein the step of ranking the knowledge points according to the magnitude of the dimensions of the integrated knowledge point evaluation vector comprises:

and sequencing the knowledge points according to the descending order of the dimensions of the comprehensive knowledge point evaluation vector.

8. The learning path planning method according to claim 1, wherein the learning behavior data of the student to be planned includes learning process data and learning preference data, the learning process data represents process data generated by the student to be planned in the process of learning each knowledge point, and the learning preference data represents learning preference and interest data of the student to be planned in the process of learning each knowledge point.

9. The learned path planning method of claim 1, further comprising:

acquiring basic information data of students to be planned;

the step of obtaining at least one of a learning duration prediction vector and a learning interest prediction vector according to the learning behavior data and the mastery degree data of each knowledge point, and the learning degree variation prediction vector includes:

and acquiring at least one of a learning duration prediction vector and a learning interest prediction vector and a learning degree variation prediction vector according to the basic information data of the student to be planned, the learning behavior data and the mastering degree data of each knowledge point.

10. The learning path planning method according to any one of claims 1 to 9, wherein the step of obtaining at least one of a learning duration prediction vector and a learning interest prediction vector based on the learning behavior data and the degree of grasp data of each of the knowledge points, and the degree of grasp variation prediction vector includes:

acquiring at least one of a learning duration prediction vector and a learning interest prediction vector and a learning degree variation prediction vector according to the learning behavior data and the mastery degree data of each knowledge point by using a deep neural network, wherein the deep neural network is acquired by the following steps:

acquiring learning behavior data training data, mastery degree training data of knowledge points, learning duration vector training data, learning interest vector training data and mastery degree variable quantity vector training data;

and inputting the learning behavior training data and the mastery degree training data of the knowledge points, and training the deep neural network based on the learning duration vector training data, the learning interest vector training data and the mastery degree variable quantity vector training data until the deep neural network is converged to obtain the trained deep neural network.

11. The learned path planning method according to claim 10, wherein the deep neural network includes a data sharing layer, a knowledge point mastery degree variation vector output layer, and a knowledge point interest vector output layer and a knowledge point learning duration vector output layer.

12. A learned path planning apparatus, comprising:

the learning data acquisition module is suitable for acquiring a knowledge point set to be subjected to path planning, learning behavior data of students to be planned and mastery degree data of the students to be planned at all knowledge points in the knowledge point set;

the prediction vector acquisition module is suitable for acquiring at least one of a learning duration prediction vector and a learning interest prediction vector and a learning degree variation prediction vector according to the learning behavior data and the mastering degree data of each knowledge point, wherein each dimension of the learning duration prediction vector represents the learning duration of each knowledge point, each dimension of the learning interest prediction vector represents the learning interest of each knowledge point, and each dimension of the mastering degree variation prediction vector represents the variation of the mastering degree of each knowledge point;

the comprehensive knowledge point evaluation vector acquisition module is suitable for carrying out vector fusion on at least one of the learning duration prediction vector and the learning interest prediction vector and the mastering degree variation prediction vector to acquire a comprehensive knowledge point evaluation vector;

and the learning path planning generation module is suitable for sequencing the knowledge points according to the size of each dimension of the comprehensive knowledge point evaluation vector to obtain the learning planning path of each knowledge point.

13. An apparatus comprising at least one memory and at least one processor; the memory stores a program that the processor invokes to perform the learned path planning method of any of claims 1-11.

14. A storage medium in which a program adapted to learn path planning is stored to implement the learned path planning method according to any one of claims 1 to 11.

Images