Knowledge atom level knowledge graph learning main path planning method and system based on PSO algorithm
By using a knowledge atom-level knowledge graph learning master path planning method based on the PSO algorithm, the problems of coarse decision units and single optimization objectives in learning path planning are solved. This generates refined learning paths that are adapted to learners' needs and subject characteristics, thereby improving learning efficiency and the adaptability of path planning.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Chinese People's Liberation Army Cyberspace Force Information Engineering University
- Filing Date
- 2026-04-15
- Publication Date
- 2026-07-31
AI Technical Summary
Existing learning path planning methods lack focus on the smallest knowledge unit, the atom of knowledge. Their optimization goals are singular and cannot simultaneously meet learners' multiple demands for learning time, difficulty matching, and knowledge coverage. Furthermore, teachers of different subjects have different focuses in learning path design, and existing methods cannot support personalized weight adjustments.
A knowledge graph learning master path planning method based on the knowledge atom level of the PSO algorithm is adopted. By selecting the target knowledge atom set, a multi-objective weighted optimization model is constructed. Constraints are set in combination with the logical connections between knowledge atoms, and an improved PSO algorithm is designed to solve the problem. The arrangement sequence of knowledge atoms is used as particle encoding. Combined with the logical constraint optimization algorithm iterative strategy, the optimal learning path is generated.
It achieves a precise match between learning paths and learners' knowledge acquisition needs, improves learning efficiency, adapts to the core teaching points of different subjects, reduces cognitive load, quickly generates optimal learning paths, and meets the real-time recommendation needs in online education scenarios.
Smart Images
Figure CN122491445A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the fields of online education, knowledge graphs, and intelligent optimization algorithms, and in particular to a knowledge atom-level knowledge graph learning master path planning method and system based on the PSO algorithm, which is applicable to the design of personalized learning paths for different disciplines such as professional courses and mathematics courses. Background Technology
[0002] With the popularization of online education, the contradiction between the vast amount of learning resources and learners' limited learning time and personalized needs is becoming increasingly prominent. In the development of online education and intelligent teaching systems, the level of refinement in learning path planning directly affects learners' learning efficiency and effectiveness. Knowledge graphs, as an important tool for representing domain knowledge, can accurately depict the granularity and relationships of subject knowledge through their smallest units, "knowledge atoms," providing a foundation for refined learning path planning. Traditional learning paths are mostly designed manually by experts, resulting in severe homogenization and an inability to adapt to individual learner differences. While existing learning path planning methods based on knowledge graphs can effectively represent the dependencies between knowledge points, they do not fully consider multi-dimensional optimization objectives. Methods based on single-objective optimization struggle to balance multiple key requirements such as learning efficiency, difficulty suitability, and knowledge coverage.
[0003] Existing technologies suffer from the following shortcomings: First, current learning path planning methods often use knowledge points or course modules as decision-making units, lacking focus on the smallest knowledge unit—the "knowledge atom"—making it difficult for the planned paths to adapt to learners' needs for refined knowledge mastery. Second, the optimization objectives are singular, failing to simultaneously meet learners' multiple demands for learning time, difficulty level, and knowledge completeness. Third, instructors from different disciplines have different emphases in designing learning paths; for example, instructors of professional courses focus more on the contribution of knowledge atoms to professional goals, while instructors of mathematics and physics courses emphasize the logical sequence between knowledge atoms, but existing methods fail to support such personalized weight adjustments. Therefore, there is an urgent need for a learning master path planning method that integrates the advantages of knowledge graphs with multi-objective optimization capabilities. Summary of the Invention
[0004] This invention addresses the problems of coarse decision-making unit granularity and lack of subject-specific adaptability in existing learning path planning, and proposes a knowledge atom-level knowledge graph learning master path planning method and system based on the PSO algorithm.
[0005] To achieve the above objectives, the present invention adopts the following technical solution:
[0006] The first aspect of this invention proposes a knowledge atom-level knowledge graph learning main path planning method based on the PSO algorithm, comprising:
[0007] Guided by learning questions, a set of target knowledge atoms is selected from the knowledge graph;
[0008] Using the knowledge atoms in the target knowledge atom set as decision variables, an optimization model is constructed with the goal of maximizing the weighted summation, and constraints are set based on the logical connections between knowledge atoms.
[0009] For the aforementioned multi-objective weighted optimization model, an improved PSO algorithm is designed to solve it. The arrangement sequence of knowledge atoms is used as particle encoding, and the learning main path is obtained by combining the logical constraint optimization algorithm iteration strategy of knowledge atoms.
[0010] Furthermore, the process of selecting the target knowledge atom set from the knowledge graph based on learning questions includes:
[0011] Transform learners' learning goals into structured learning questions and extract core knowledge requirements;
[0012] Based on the semantic retrieval capability of knowledge graphs, relevant knowledge atoms are matched from the global knowledge graph to form an initial candidate set;
[0013] The initial candidate set is classified according to learning needs, and redundant and irrelevant knowledge atoms are removed to obtain the final target knowledge atom set.
[0014] Furthermore, the optimization model includes:
[0015] MaxF=
[0016] In the formula,
[0017]
[0018]
[0019]
[0020]
[0021] Where F is the objective function, For weighted parameters, , , and These are optimization functions for the dimensions of knowledge atom contribution, logical sequence, resource richness, and learning duration, respectively. For knowledge atoms The contribution value, where m is the number of knowledge atoms in the path. For knowledge atoms and Logical matching degree For knowledge atoms The number of associated learning resources, Let n be the maximum number of learning resources for knowledge atoms in the target knowledge atom set, where n represents the number of knowledge atoms in the target knowledge atom set. To master the atom of knowledge The estimated duration, The total estimated duration for all knowledge atoms in the target knowledge atom set.
[0022] Furthermore, the aforementioned Adjustments should be made adaptively based on the characteristics of the discipline.
[0023] Furthermore, the constraints include:
[0024] All knowledge atoms appearing in the path must also have all their prerequisite knowledge atoms appearing in the path, and these prerequisite knowledge atoms must precede the knowledge atom.
[0025] Furthermore, for the aforementioned multi-objective weighted optimization model, an improved PSO algorithm is designed for solving the problem. Using the permutation sequence of knowledge atoms as particle encoding, and combining iterative strategies based on the logical constraints of knowledge atoms, the main learning path is obtained, including:
[0026] Map the sequence of knowledge atoms in the path to particles, where the dimension of each particle is equal to the number of knowledge atoms in the path;
[0027] The particle population is initialized based on the logical connections of knowledge atoms in the knowledge graph, and the population size and maximum number of iterations are set.
[0028] The objective function F is used as the fitness function of the particles, and the fitness value of each particle is calculated.
[0029] The speed and position update formulas of the standard PSO algorithm are adopted, and a logical constraint penalty term is introduced to optimize the update strategy;
[0030] When the number of iterations reaches the maximum number of iterations or the fitness value converges, the iteration stops, and the knowledge atom sequence corresponding to the globally optimal particle is output as the optimal learning master path.
[0031] Furthermore, if a particle violates a logical constraint, the logical constraint penalty term is non-negative; otherwise, it is 0.
[0032] A second aspect of this invention proposes a knowledge atom-level knowledge graph learning master path planning system based on the PSO algorithm, comprising:
[0033] The knowledge atom set filtering module is used to filter out target knowledge atom sets from the knowledge graph based on learning questions.
[0034] The multi-objective weighted optimization model construction module is used to construct an optimization model with the goal of maximizing the weighted summation, using knowledge atoms in the target knowledge atom set as decision variables, and setting constraints based on the logical connections between knowledge atoms;
[0035] The model solving module based on the PSO algorithm is used to design an improved PSO algorithm to solve the multi-objective weighted optimization model. It uses the arrangement sequence of knowledge atoms as particle encoding and combines the logical constraint optimization algorithm iteration strategy of knowledge atoms to obtain the main learning path.
[0036] A third aspect of the present invention provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the knowledge atom-level knowledge graph learning master path planning method based on the PSO algorithm as described in any one of the claims.
[0037] The fourth aspect of this invention provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, it implements the knowledge atom-level knowledge graph learning master path planning method based on the PSO algorithm described in any one of the claims.
[0038] Compared with the prior art, the present invention has the following beneficial effects:
[0039] The proposed method and system for planning the main learning path of a knowledge graph based on the PSO algorithm demonstrate multi-dimensional application value and practical advantages. At the level of refined path planning, using knowledge atoms as the core decision-making unit, the planned learning path precisely matches the learner's knowledge acquisition needs, effectively mitigating the problems of leaps and redundancy in knowledge transfer and helping learners gradually build a complete and refined knowledge system. Relying on an advanced dynamic parameter adjustment mechanism, instructors can optimize the weighted parameter configuration according to the characteristics of different disciplines, ensuring that the learning path aligns with the core teaching points of each discipline. For example, professional courses emphasize the weight allocation of knowledge contribution, while mathematics and physics courses strengthen the dominant position of logical order, thereby improving the adaptability of the learning path to the subject teaching. By mining the inherent logical connections between knowledge atoms and setting constraints, the planned learning path strictly follows the prerequisite and sequential relationships of knowledge, significantly reducing the learner's cognitive load and thus achieving a significant improvement in learning efficiency. The improved PSO algorithm introduces a logical constraint penalty term to eliminate invalid solutions, which significantly optimizes the algorithm's optimization efficiency and convergence rate. It can quickly generate the optimal learning path for learners and fully adapt to the need for real-time recommendation of learning paths in online education scenarios. Attached Figure Description
[0040] Figure 1This is a flowchart illustrating a knowledge atom-level knowledge graph learning main path planning method based on the PSO algorithm according to an embodiment of the present invention.
[0041] Figure 2 The flowchart of the PSO-based multi-objective learning path optimization algorithm provided in the embodiments of the present invention is as follows:
[0042] Figure 3 This is a schematic diagram of the architecture of a knowledge atom-level knowledge graph learning master path planning system based on the PSO algorithm, provided in an embodiment of the present invention. Detailed Implementation
[0043] For ease of understanding, the following explanations are provided for some of the terms used in the specific embodiments of this invention:
[0044] Knowledge atom: The smallest indivisible knowledge unit with independent semantics in the field of education, and the basic unit for building knowledge graphs.
[0045] PSO Algorithm: Particle Swarm Optimization, an optimization algorithm based on swarm intelligence that finds the best by simulating the foraging behavior of bird flocks.
[0046] Contribution: The degree to which a knowledge atom is important for learners to achieve specific learning goals.
[0047] Logical learning order: The logical dependencies such as prior and subsequent learning naturally exist between knowledge atoms in a knowledge graph.
[0048] Resource richness: The quantity and variety of learning resources (videos, texts, quizzes, etc.) associated with a single knowledge atom.
[0049] Learning duration: The estimated time required for a learner to master a particular knowledge atom.
[0050] The present invention will be further explained below with reference to the accompanying drawings and specific embodiments:
[0051] This invention proposes a knowledge atom-level knowledge graph learning main path planning method based on the PSO algorithm, such as... Figure 1 As shown, it includes:
[0052] Guided by learning questions, a set of target knowledge atoms is selected from the knowledge graph;
[0053] Using the knowledge atoms in the target knowledge atom set as decision variables, an optimization model is constructed with the goal of maximizing the weighted summation, and constraints are set based on the logical connections between knowledge atoms.
[0054] For the aforementioned multi-objective weighted optimization model, an improved PSO algorithm is designed to solve it. The arrangement sequence of knowledge atoms is used as particle encoding, and the learning main path is obtained by combining the logical constraint optimization algorithm iteration strategy of knowledge atoms.
[0055] This method uses knowledge atoms as the basic decision-making unit, selects the target knowledge atom set based on the learning problem, sets weighted optimization objectives in combination with the needs of different disciplines, constructs a model based on the logical constraints between knowledge atoms, and finally solves the optimal learning path by improving the PSO algorithm.
[0056] Furthermore, the method specifically includes:
[0057] 1. Filtering of knowledge atom sets
[0058] The knowledge atom set selection process is driven by the learner's specific learning questions. It selects target knowledge atom sets from the knowledge graph. First, the learner's learning objectives (e.g., "mastering gear transmission design in mechanical design" or "mastering the application of calculus in advanced mathematics") are transformed into structured learning questions, and core knowledge requirements are extracted. Then, based on the semantic retrieval capabilities of the knowledge graph, relevant knowledge atoms are matched from the global knowledge graph to form an initial candidate set. Finally, the initial candidate set is categorized according to learning needs (it can be selected as a pure professional knowledge atom set, such as containing only mechanical design-related knowledge atoms, or a set integrating mathematical knowledge atoms, such as containing both mechanical design and advanced mathematics knowledge atoms). Redundant and irrelevant knowledge atoms are removed to obtain the final target knowledge atom set. (n is the number of knowledge atoms).
[0059] 2. Construction of a multi-objective weighted optimization model
[0060] Using the knowledge atoms in the target knowledge atom set as decision variables, an optimization model is constructed with the goal of maximizing the weighted summation, and constraints are set based on the logical connections between knowledge atoms.
[0061] (1) Objective function design
[0062] The objective function is to maximize the weighted sum of the four optimization dimensions, as shown in the following formula:
[0063] MaxF=
[0064] Where F is the objective function, For weighted parameters (satisfying) The schedule can be adjusted by different instructors according to the characteristics of the subject (e.g., the schedule for professional course instructors). =0.4, set by the math and physics teacher =0.4); , , and These are optimization functions for the dimensions of knowledge atom contribution, logical sequence, resource richness, and learning duration, respectively, and are defined as follows:
[0065] Knowledge Atom Contribution ( The formula for measuring the importance of knowledge atoms to learning objectives is:
[0066]
[0067] in, For knowledge atoms As one possible implementation method, the contribution value of a knowledge atom is the number of times it is called by the learning path, where n represents the number of knowledge atoms in the target knowledge atom set.
[0068] Logical learning order ( ): This measures the degree of fit between the arrangement of knowledge atoms in the path and the inherent logical order of the knowledge graph. The calculation formula is:
[0069]
[0070] Where m is the number of knowledge atoms in the path. For knowledge atoms and Logical matching degree (if) yes The prerequisite knowledge atoms. Otherwise, it is 0).
[0071] Abundance of learning resources ( ): This measures the diversity and quantity of learning resources associated with a knowledge atom. The calculation formula is:
[0072]
[0073] in, For knowledge atoms The number of associated learning resources (including video, text, quizzes, etc.). The maximum number of learning resources for knowledge atoms in the target knowledge atom set, which takes a value of 0-1 after normalization.
[0074] Study time ( ): This measures the estimated time required to master the knowledge atoms. Since the learning time needs to be minimized, it exists as a subtractive term in the objective function, and the calculation formula is:
[0075]
[0076] in, To master the atom of knowledge The estimated duration, The total estimated duration of all knowledge atoms in the target knowledge atom set is normalized to a value between 0 and 1.
[0077] (2) Setting constraints
[0078] The constraints are set based on the inherent logical connections between knowledge atoms in the knowledge graph, and the core constraints are:
[0079] ∀ , ∈ path M, ≥0, meaning the logical matching degree of any adjacent knowledge atoms in the path is non-negative, if knowledge atoms If the prerequisite knowledge atom does not appear in the path, then It must not appear before the prerequisite knowledge atom, ensuring that the path conforms to the inherent logical order of knowledge. That is, for every knowledge atom that appears in the path, all its prerequisite knowledge atoms must also appear in the path, and must be located before the knowledge atom.
[0080] 3. Model Solving Based on PSO Algorithm
[0081] To address the aforementioned multi-objective weighted optimization model, an improved PSO algorithm is designed, namely, a multi-objective learning path optimization algorithm based on PSO. This algorithm uses the permutation sequence of knowledge atoms as particle encoding and combines iterative strategies based on the logical constraints of knowledge atoms. Figure 2 As shown, the specific steps are as follows:
[0082] (1) Particle encoding and population initialization
[0083] Particle encoding: Maps the sequence of knowledge atoms in a path to particles. The dimension of each particle is equal to the number of knowledge atoms in the path, and the dimension value is the unique identifier (ID) of the knowledge atom. For example, particle X = [ The learning path is represented by knowledge atoms. .
[0084] Population initialization: The particle population is initialized based on the logical connections of knowledge atoms in the knowledge graph, ensuring that the paths corresponding to the initial particles satisfy the "pre-fix-successor" constraint and avoiding invalid solutions. The population size is set to N=50, and the maximum number of iterations is G=100.
[0085] (2) Fitness calculation
[0086] The objective function F is used as the fitness function of the particles. The fitness value of each particle is calculated. The higher the fitness value, the better the corresponding learning path.
[0087] (3) Particle Iterative Update
[0088] The velocity and position update formulas of the standard PSO algorithm are adopted, and a logical constraint penalty term is introduced to optimize the update strategy, so as to avoid violating the logical connection of knowledge atoms after particle update:
[0089]
[0090]
[0091] Where ω is the inertia weight (values range from 0.4 to 0.9). , The learning factor (usually 2) is used. , λ is a random number between 0 and 1, pbest is the individual optimal position, and gbest is the global optimal position; λ is the penalty coefficient, and P(x(t)) is the logical constraint penalty term (if particle x(t) violates the logical constraint, P(x(t)) is non-negative, otherwise it is 0). v(t) represents the velocity vector of the particle at the t-th iteration, v(t+1) represents the velocity vector of the particle at the t+1-th iteration, x(t) represents the position vector of the particle at the t-th iteration, and x(t+1) represents the position vector of the particle at the t+1-th iteration.
[0092] (4) Termination conditions and optimal path output
[0093] When the number of iterations reaches the maximum number of iterations G or the fitness value converges (the fitness change is less than 1 in 10 consecutive iterations), the fitness value is considered to converge. -5 When the condition is met, stop the iteration. Output the knowledge atom sequence corresponding to the globally optimal particle gbest as the optimal learning master path.
[0094] This invention proposes a PSO-based multi-objective learning path optimization algorithm. In the initialization phase, it directly addresses the core requirement of "knowledge atom-level path planning," explicitly defining initialization parameters as including both algorithm parameters and particle parameters. Furthermore, particle initialization is forcibly bound to the "target knowledge atom set" and "knowledge logic constraints," avoiding invalid paths from the source and solving the problem of poor compatibility between traditional PSO algorithms and knowledge graphs. It clarifies that "position constraints" are the "range of knowledge atom ID sets," and "logic constraint detection" is "precedence-successor relationship verification," directly echoing the core requirement in learning path planning that "paths must conform to knowledge logic," thus solving the problem of traditional algorithms ignoring knowledge associations. In the fitness calculation phase, it fully presents the composition of the fitness function, emphasizing... "Custom weights for instructors" addresses the need for different disciplines (professional courses / mathematical courses) to have different focuses, solving the problem of lack of subject-specific adaptability in optimization objectives; Termination conditions and output stages: clearly define the quantitative standards for termination conditions, and the output results directly correspond to the "knowledge atom sequence", clearly implementing the core output of the "learning main path" and solving the problem of the disconnect between algorithms and actual application scenarios.
[0095] This invention uses knowledge atoms as decision variables, refining the decision unit of learning path planning to the knowledge atom level, thus achieving refined path planning and solving the problem of coarse granularity in traditional methods. This aligns with learners' needs for refined knowledge mastery. Furthermore, it constructs a customizable weighted optimization model: a multi-objective weighted model incorporating contribution, logical order, resource richness, and learning duration. The weighting parameters can be adjusted by different instructors according to the characteristics of their respective subjects, meeting the teaching needs of different disciplines such as professional courses and mathematics courses. This model also integrates knowledge logic constraint design: constraints are set based on the inherent logical connections of knowledge atoms in the knowledge graph, ensuring that the planned path conforms to the prior and subsequent relationships of knowledge, guaranteeing the logical rationality of the path. Finally, this invention proposes a solution strategy based on an improved PSO algorithm, designing particle initialization and iterative update strategies for the logical constraints of knowledge atoms, and introducing logical constraint penalty terms to avoid generating invalid solutions during the algorithm's optimization process, thereby improving the algorithm's optimization efficiency and path quality.
[0096] In terms of empowering educational and teaching practices, the learning paths generated based on this method can support intelligent teaching systems in providing personalized teaching services. The system can push customized learning content and advancement plans based on learners' knowledge mastery status and the logical connections between knowledge atoms, meeting the differentiated learning needs of different learners and effectively improving learning outcomes. For example, in engineering courses, the system can recommend suitable case studies and practical training based on the logical connections between engineering principles and practical applications in the knowledge graph, as well as learners' previous learning feedback, helping them to deeply master professional knowledge. Teachers can use this path planning method to systematically organize the subject knowledge system, clearly grasp the logical connections and weight distribution between knowledge points, and thus optimize the design of teaching links and content arrangement, improving the pertinence and effectiveness of classroom teaching. Teachers can also design teaching processes that better conform to cognitive laws based on the decomposition and correlation analysis of knowledge atoms, guiding students to gradually build a systematic knowledge framework. Furthermore, the knowledge atom decomposition and path planning logic upon which this method relies can provide quantitative data analysis support for educational research. Researchers can analyze the correlation between path planning characteristics and learning outcomes in different disciplines to uncover the optimization direction and patterns of subject teaching, providing a scientific basis for decision-making in educational reform and curriculum system construction.
[0097] Based on the above embodiments, such as Figure 3 As shown, this invention also proposes a knowledge atom-level knowledge graph learning master path planning system based on the PSO algorithm, comprising:
[0098] The knowledge atom set filtering module is used to filter out target knowledge atom sets from the knowledge graph based on learning questions.
[0099] The multi-objective weighted optimization model construction module is used to construct an optimization model with the goal of maximizing the weighted summation, using knowledge atoms in the target knowledge atom set as decision variables, and setting constraints based on the logical connections between knowledge atoms;
[0100] The model solving module based on the PSO algorithm is used to design an improved PSO algorithm to solve the multi-objective weighted optimization model. It uses the arrangement sequence of knowledge atoms as particle encoding and combines the logical constraint optimization algorithm iteration strategy of knowledge atoms to obtain the main learning path.
[0101] The present invention also proposes an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the knowledge atom-level knowledge graph learning master path planning method based on the PSO algorithm described in any one of the present invention.
[0102] The present invention also proposes a computer-readable storage medium storing a computer program thereon, wherein the computer program, when executed by a processor, implements the knowledge atom-level knowledge graph learning master path planning method based on the PSO algorithm described in any one of the claims.
[0103] The above description is only a preferred embodiment of the present invention. It should be noted that those skilled in the art can make several improvements and modifications without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.
Claims
1. A knowledge atom-level knowledge graph learning master path planning method based on the PSO algorithm, characterized in that, include: Guided by learning questions, a set of target knowledge atoms is selected from the knowledge graph; Using the knowledge atoms in the target knowledge atom set as decision variables, an optimization model is constructed with the goal of maximizing the weighted summation, and constraints are set based on the logical connections between knowledge atoms. For the aforementioned multi-objective weighted optimization model, an improved PSO algorithm is designed to solve it. The arrangement sequence of knowledge atoms is used as particle encoding, and the learning main path is obtained by combining the logical constraint optimization algorithm iteration strategy of knowledge atoms.
2. The knowledge atom-level knowledge graph learning main path planning method based on the PSO algorithm according to claim 1, characterized in that, The process of selecting the target knowledge atom set from the knowledge graph, guided by learning questions, includes: Transform learners' learning goals into structured learning questions and extract core knowledge requirements; Based on the semantic retrieval capability of knowledge graphs, relevant knowledge atoms are matched from the global knowledge graph to form an initial candidate set; The initial candidate set is classified according to learning needs, and redundant and irrelevant knowledge atoms are removed to obtain the final target knowledge atom set.
3. The knowledge atom-level knowledge graph learning main path planning method based on the PSO algorithm according to claim 1, characterized in that, The optimization model includes: MaxF= In the formula, Where F is the objective function, For weighted parameters, , , and These are optimization functions for the dimensions of knowledge atom contribution, logical sequence, resource richness, and learning duration, respectively. For knowledge atoms The contribution value, where m is the number of knowledge atoms in the path. For knowledge atoms and Logical matching degree For knowledge atoms The number of associated learning resources, Let n be the maximum number of learning resources for knowledge atoms in the target knowledge atom set, where n represents the number of knowledge atoms in the target knowledge atom set. To master the atom of knowledge The estimated duration, The total estimated duration for all knowledge atoms in the target knowledge atom set.
4. The knowledge atom-level knowledge graph learning main path planning method based on the PSO algorithm according to claim 3, characterized in that, The Adjustments should be made adaptively based on the characteristics of the discipline.
5. The knowledge atom-level knowledge graph learning main path planning method based on the PSO algorithm according to claim 1, characterized in that, The constraints include: All knowledge atoms appearing in the path must also have all their prerequisite knowledge atoms appearing in the path, and these prerequisite knowledge atoms must precede the knowledge atom.
6. The knowledge atom-level knowledge graph learning main path planning method based on the PSO algorithm according to claim 3, characterized in that, For the aforementioned multi-objective weighted optimization model, an improved PSO algorithm is designed for solving the problem. Using the sequence of knowledge atoms as particle codes and combining iterative strategies based on the logical constraints of knowledge atoms, the main learning path is obtained, including: Map the sequence of knowledge atoms in the path to particles, where the dimension of each particle is equal to the number of knowledge atoms in the path; The particle population is initialized based on the logical connections of knowledge atoms in the knowledge graph, and the population size and maximum number of iterations are set. The objective function F is used as the fitness function of the particles, and the fitness value of each particle is calculated. The speed and position update formulas of the standard PSO algorithm are adopted, and a logical constraint penalty term is introduced to optimize the update strategy; When the number of iterations reaches the maximum number of iterations or the fitness value converges, the iteration stops, and the knowledge atom sequence corresponding to the globally optimal particle is output as the optimal learning master path.
7. The knowledge atom-level knowledge graph learning main path planning method based on the PSO algorithm according to claim 6, characterized in that, If a particle violates a logical constraint, the logical constraint penalty term is non-negative; otherwise, it is 0.
8. A knowledge atom-level knowledge graph learning master path planning system based on the PSO algorithm, characterized in that, include: The knowledge atom set filtering module is used to filter out target knowledge atom sets from the knowledge graph based on learning questions. The multi-objective weighted optimization model construction module is used to construct an optimization model with the goal of maximizing the weighted summation, using knowledge atoms in the target knowledge atom set as decision variables, and setting constraints based on the logical connections between knowledge atoms; The model solving module based on the PSO algorithm is used to design an improved PSO algorithm to solve the multi-objective weighted optimization model. It uses the arrangement sequence of knowledge atoms as particle encoding and combines the logical constraint optimization algorithm iteration strategy of knowledge atoms to obtain the main learning path.
9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the knowledge atom-level knowledge graph learning master path planning method based on the PSO algorithm as described in any one of claims 1 to 7.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the knowledge atom-level knowledge graph learning master path planning method based on the PSO algorithm as described in any one of claims 1 to 7.