Personalized online learning path planning and optimizing method

Abstract

The invention discloses a personalized online learning path planning and optimizing method, which comprises the following steps of: obtaining a learning target through user interaction, and generating a dynamic priority matrix; selecting knowledge nodes related to the target from the target map, and generating a knowledge unit set in combination with a user learning background; a modular learning path is constructed, and nonlinear jump between the paths is realized through dynamic adjustment; in the path execution process, collecting user behavior data, dynamically adjusting node content and sequence, and introducing related background knowledge and practice tasks; low-priority nodes are removed through pruning, and the mastering effect of core nodes is enhanced; after the current path is completed, migrating the key knowledge node to a new path framework; and displaying a learning path structure and an adjustment result through a visual interface to form a closed-loop optimization mechanism of path planning. According to the method, personalized construction and optimization of the learning path are realized, and the pertinence and efficiency of online learning are improved.

Description

Technical Field

[0001] The present invention relates to the field of online education technology, and in particular to a personalized online learning path planning and optimization method. Background Art

[0002] At present, online education, as an important part of modern education, has been continuously expanding its application scope. However, in the face of users' growing personalized learning needs, the existing online learning path planning methods still have many shortcomings in technical implementation, which limits the improvement of learning effects. The design of traditional online learning paths is often based on a preset course structure or a fixed sequence of knowledge points. This static planning method is difficult to dynamically adjust to the changing needs and status of users in the learning process. For example, users may need to re-plan their learning paths due to changes in interests, adjustments in learning goals, or differences in the degree of mastery of knowledge points. However, existing methods lack flexibility in dealing with these dynamic needs and cannot effectively adapt to the personalized needs of learners.

[0003] Current learning path planning has great limitations in goal analysis. Most methods can only simply classify or stratify learning goals, lack in-depth correlation analysis and multi-level analysis capabilities, resulting in the inability to effectively reflect the connection between the main goal, secondary goal and extended goal. This single analysis mode makes it difficult for the learning path to support complex learning needs during execution, reducing learning efficiency. At the same time, the low matching between knowledge content recommendations and user learning scenarios is also a common problem of existing methods. Existing methods usually select content based on static knowledge graphs without fully considering the user's learning background, learning preferences and learning scenarios. This static matching mechanism leads to a clear gap between learning content and user needs, making it difficult to meet users' requirements for content adaptability in the actual learning process.

[0004] In terms of path optimization, existing technical methods are too simplistic and are often limited to the simple logical relationship optimization of path nodes, or a path evaluation model centered on learning completion rate, while ignoring the deep optimization of path content and the nonlinear design of dynamic jump logic. Especially in complex learning scenarios, traditional path optimization methods cannot effectively deal with learning bottlenecks, resulting in a decline in user learning experience. In addition, the path planning and optimization of most online learning systems lack a closed-loop feedback mechanism, and user behavior data and learning path execution data are difficult to feed back into the planning model, limiting the iterative ability of path optimization. Existing methods are unable to dynamically optimize target sorting and path adjustment rules, and are unable to achieve continuous improvement.

[0005] Therefore, how to provide a personalized online learning path planning and optimization method is a problem that technical personnel in this field urgently need to solve. Summary of the invention

[0006] One object of the present invention is to propose a personalized online learning path planning and optimization method. The present invention constructs a dynamic priority matrix, combines knowledge graphs with user learning scenario characteristics, modularly designs and optimizes the learning path in real time, and describes in detail the whole process from target demand analysis to closed-loop optimization of the learning path, including target priority calculation, knowledge unit generation and path planning, path dynamic adjustment and pruning optimization, and multi-scenario path adaptation and closed-loop feedback mechanism implementation. The method makes full use of user behavior data and learning scenario characteristics, dynamically adjusts the content and order of path nodes, and strengthens the mastery of key knowledge points. It has the advantages of high personalized adaptation ability, dynamic optimization ability and improved learning efficiency.

[0007] A personalized online learning path planning and optimization method according to an embodiment of the present invention includes the following steps:

[0008] S1. Obtain short-term goals, long-term goals and implicit learning needs through user interaction, analyze the goals into primary goals, secondary goals and extended goals, build a goal map, and generate a dynamic priority matrix based on the time priority, relevance and interest changes of the goals;

[0009] S2. According to the dynamic priority matrix, select knowledge nodes related to the target from the knowledge graph, perform semantic decomposition and reconstruction on the selected nodes, and generate a set of knowledge units based on the user's learning background and scenario characteristics;

[0010] S3. Construct a modular learning path based on a set of knowledge units. The path includes a main path, an auxiliary path, and a flexible path. The modular path structure is used to realize dynamic association and nonlinear logic jumps between paths. The path design dynamically adjusts the order and content layout of target nodes according to a dynamic priority matrix.

[0011] S4. During the execution of the learning path, collect user behavior data, including node dwell time, completion rate, and feedback results, adjust the content and order of the path nodes according to the behavior data, and introduce background knowledge, predictive knowledge, and practical tasks related to the current node through the expansion mechanism to update the content depth and logical order of the path;

[0012] S5. Conduct real-time value assessment of nodes in the modular learning path, prune according to node completion rate, importance weight and path logic, remove low-priority nodes, and add verification tasks to the core nodes of the main path to dynamically strengthen the mastery of key knowledge points and optimize the focus and coherence of the path structure;

[0013] S6. After completing the current path, based on the dynamic priority matrix and pruning optimization results, analyze the relevance of the core knowledge units in the current path with the new goal, migrate and integrate the core nodes into the new path framework, generate a multi-scenario learning path, and support the learning needs of the new goal;

[0014] S7. Display the path structure, learning progress and adjustment results through a visual interface, input user behavior data and path execution data into the dynamic priority matrix update model, optimize the target sorting and path planning rules, and form a closed-loop optimization mechanism for path planning.

[0015] Optionally, the S1 specifically includes:

[0016] S11, collect the user's learning demand information through user interaction, the learning demand information includes short-term goals, long-term goals and implicit goals, and construct a goal set G = {g1, g2, ..., g n}, where g i is the i-th target node in the target set;

[0017] S12. Analyze each target node in the target set, decompose the target into primary target, secondary target and extended target, and construct a target map. The target map represents the target nodes and the relationship between them. The relationship is defined as R ij =f(g i ,g j ), where R ij Represents the target node g i and g j The strength of the association between them, f represents the functional form of the dependent characteristics;

[0018] S13. Calculate the target priority weight W based on the target node characteristics in the target graph i :

[0019]

[0020] Among them, W i is the target node g i The priority weight, P t (g i ) is the time urgency of the target node, P r (g i ) is the relevance weight of the target node, P h (g i ) is the interest weight of the target node, α t , α r , α h are the moderators of time, relevance, and interest weights, respectively;

[0021] S14, based on the calculated priority weight W i , sort the target nodes and generate a dynamic priority matrix M = [W1, W2, ..., W n ];

[0022] S15. During the user learning process, collect user behavior feedback data, including completion rate, dwell time and interest changes of target nodes, and dynamically adjust the target priority weight according to the feedback data;

[0023] S16. Combined with the dynamically updated priority matrix M, adjust the association relationship and sorting rules of the target nodes in the target graph to generate an optimized target graph.

[0024] Optionally, the S2 specifically includes:

[0025] S21. According to the target weight in the dynamic priority matrix, a set of knowledge nodes related to the target node is selected from the optimized target map. The selection of knowledge nodes is based on the correlation strength R between the target node and the knowledge node. ij :

[0026]

[0027] Among them, R ij Represents the target node g i With knowledge node k j The strength of association, d(g i ,k k ) is the target node g i With knowledge node k k The shortest path length, P(k j ) represents knowledge node k j The set of neighboring nodes, ρ(k j ) is the semantic relevance of knowledge nodes, γ and δ are the adjustment factors of path length and semantic strength, respectively, and max(d) is the longest path length in the graph;

[0028] S22, perform semantic decomposition on the selected knowledge nodes, and decompose each knowledge node into core knowledge units and extended knowledge units, where the core knowledge unit C ij Directly supports target learning needs and expands knowledge unit E ij To supplement background information or practice tasks, the semantic decomposition rules are dynamically adjusted according to the association weights between nodes and targets;

[0029] S23. Combine the user's learning background and learning scenario characteristics to analyze the decomposed core knowledge unit C ij and Extended Knowledge Unit E ij Perform dynamic reconstruction to generate a knowledge unit set K' i ={k′i1 ,k' i2 ,…,k′ ip};

[0030] S24, for the dynamically reconstructed knowledge unit set, according to the priority weight W of the target node i The correlation strength R of the knowledge node ij To perform priority scoring:

[0031] S k′ =W i α+R ij (1-α);

[0032] Among them, S k· Score the comprehensive priority of the knowledge unit, W i is the priority weight of the target node, R ij is the association strength between the target and the knowledge node, α is the weight distribution coefficient;

[0033] S25. According to the scoring results, filter the knowledge units with higher priorities and construct the final set of knowledge units.

[0034] Optionally, the S3 specifically includes:

[0035] S31, based on the final knowledge unit set and the dynamic priority matrix, the knowledge units are divided into a modular path structure according to the priority weight of the target node, and the modular path includes a main path, an auxiliary path and a flexible path;

[0036] S32, the main path is composed of core knowledge units, the auxiliary path is composed of extended knowledge units, the elastic path is used to connect the main path and the auxiliary path, and the dynamic association between the path modules is established through non-linear jump logic;

[0037] S33, according to the weight of the target node in the dynamic priority matrix, the nodes in the modular path are prioritized to optimize the path layout;

[0038] S34, design nonlinear jump logic in modular paths, through the jump probability matrix T = [t ij ] Determine the jump relationship between path nodes, where t ij Represents the path node p i Jump to node p j Probability:

[0039]

[0040] Among them, P rank (p i ) represents the path node p i The priority score, R ijRepresents the path node p i and p j The strength of logical association;

[0041] S35, dynamically adjust the generated modular path in combination with the user learning behavior data, including the user's stay time, completion rate and feedback data, adjust the order and jump relationship of the path nodes, and form a preliminary optimized modular path set P′={P1,P2,…,P q}.

[0042] Optionally, the S4 specifically includes:

[0043] S41, in the formed modular path set P′={P1,P2,…,P q}, perform learning path planning, collect user behavior data, and construct a behavior data set D = {d1, d2, …, d n}, where d i Include path node p i The residence time t(p i ), completion rate c(p i ) and user feedback f(p i );

[0044] S42, using the behavior data set D, dynamically adjust the priority weights of the path nodes, and adjust the priority weights P rank (p i ) to update and recalculate the comprehensive priority of the path nodes:

[0045] P rank ′(p i )=P rank (p i )+γ1·t(p i )+γ2·c(p i )+γ3·f(p i );

[0046] Among them, P rank ′(p i ) is the updated priority weight, P rank (p i ) is the initial priority weight, γ1, γ2, γ3 are the adjustment factors of residence time, completion rate and user feedback respectively;

[0047] S43, according to the updated priority weight P rank ′(p i ), dynamically expand the path node content, optimize the content depth and logical structure of the path node by introducing background knowledge, predictive knowledge and practical tasks, and generate the expanded path node set p i ′;

[0048] S44, based on the expanded path node set p i ′ and the updated priority weight P rank ′(p i ), optimize the jump logic of the modular path, and update the jump probability matrix T′=[t ij ′], where the jump probability t ij ′Dynamically adjust through node priority and logical association strength;

[0049] S45, integrating the expanded and updated path node content and the dynamic jump logic, and feeding back to the modular path structure, generating the final optimized modular path set P″={P″1,P″2,…,P″ q}, the path set P″ contains the updated priority weight P rank ′(p i ), node jump logic T′ and user behavior data feedback.

[0050] Optionally, the S5 specifically includes:

[0051] S51, generating the final optimized modular path set P″={P″1,P″2,…,P″ q}, the learning completion rate, importance weight and path logic of the path nodes are evaluated in real time, and the node value evaluation set V = {v1, v2, …, v n}:

[0052]

[0053] Among them, v i Represents the path node p i The comprehensive value of i ) is the node p i The learning completion rate, W′(p i ) is the priority weight of the node, is the association strength of the node in the path logic, α and β are weight factors, which dynamically adjust the contribution ratio of each dimension when evaluating the node value;

[0054] S52: According to the node value evaluation set V, prune the low-value nodes, remove the nodes whose comprehensive value is lower than the pruning threshold θ, and generate the pruned modular path set P″ pruned ;

[0055] S53, the modular path set P″ after pruning pruned In the process, the core node of the main path is selected to design the verification task. The verification task is associated with the knowledge point of the core node. The completion rate R(p i ) Update the priority weight of the core node;

[0056] S54. According to the core node priority and pruning results fed back by the verification task, the node order and jump logic in the modular path are optimized, the nodes are rearranged, the content focus and logical coherence of the path are optimized, and the enhanced modular path set P″′={P″′1,P″′2,…,P″′ q}.

[0057] Optionally, the S6 specifically includes:

[0058] S61. Based on the pruning optimization result and the dynamic priority matrix, extract the current path set P″′={P″′1, P″′2, …, P″′ q}, analyze the core knowledge unit k i Relevance to the new learning goal set G′:

[0059] R′(k i ,g′ j )=(α·A(k i ,g′ j )+β·C(k i ,g′ j ))+γ·W′(k i );

[0060] Among them, R′(k i ,g′ j ) represents node k i With the new target g′ j The priority association degree, A(k i ,g′ j ) represents the semantic association strength, C(k i ,g′ j ) represents context adaptability, W′(k i ) is the importance weight of the node, α, β, γ are adjustment factors;

[0061] S62, according to the priority association R′(k i ,g′ j ), sort the core knowledge unit set K, give priority to the nodes with high correlation, and migrate them to the new path framework to form the core knowledge unit set K′ that adapts to the new goal;

[0062] S63. Based on the core knowledge unit set K′, the extended knowledge unit and the context knowledge unit are combined to construct a multi-scenario adaptation path set P scene ;

[0063] S64: Adapting the path set P for multiple scenarios sceneThe content layout and logical structure are optimized, the order and jump relationship of the path nodes are adjusted according to the dynamic priority matrix, and the target adaptation path set is generated.

[0064] Optionally, the S7 specifically includes:

[0065] S71. In the generated target adaptation path set Based on the above, the path structure is displayed through a visual interface, including the priority score of the path node, the logical jump relationship, the learning progress and the adjustment results;

[0066] S72, collect user learning behavior data, and construct a behavior data set D = {d1, d2, ..., d v}, where d i Include path node p i The residence time t(p i ), completion rate c(p i ), feedback result f(p i ) and node association behavior B(p i );

[0067] S73, input the behavior data set D and the path execution data into the dynamic priority matrix update model, and recalculate the priority score P′ of the path node rank (p i ), update the path planning rules;

[0068] S74. Score P′ based on the updated priority rank (p i ), adjust the content layout and logical order of the path nodes, rearrange the path nodes according to the priority weights, and generate an optimized path set

[0069] S75: Adapt the generated target path set P adapt and the optimized path set P final The integration is fed back into the dynamic priority matrix update model to optimize the target sorting and path planning rules, forming a closed-loop path optimization mechanism.

[0070] The beneficial effects of the present invention are:

[0071] The present invention utilizes a dynamic priority matrix, and through a comprehensive analysis of the user's short-term goals, long-term goals and implicit needs, decomposes the goals into primary goals, secondary goals and extended goals, and generates an optimized target map in combination with time priority, relevance and interest changes, to ensure the diversity and accuracy of learning goals. By selecting knowledge nodes with strong relevance from the optimized target map, a dynamically adjusted set of knowledge units is constructed. The present invention effectively solves the problem of fixed existing learning content and realizes the dynamic decomposition and reconstruction of knowledge units. On this basis, a modular path structure including a main path, an auxiliary path and a flexible path is designed. The flexible relevance between paths is enhanced through non-linear jump logic to support learning needs of multi-level goals. The design of modular paths is combined with user behavior data, so that the learning path can dynamically adjust the content and sequence according to data such as residence time, completion rate and feedback results during execution, thereby optimizing the depth and logic of the path.

[0072] The present invention improves the focus and logical coherence of the learning path by performing real-time value evaluation and pruning optimization on path nodes. At the same time, it strengthens the mastery of knowledge points by introducing verification tasks in core nodes. Combined with the migration and integration mechanism of core knowledge units, the present invention can generate multi-scenario adaptation paths according to new target requirements to meet the ever-changing needs of user learning goals, reflecting a high degree of flexibility and scalability. Through a closed-loop path optimization mechanism, user behavior data and path execution data are fed back to the dynamic priority matrix update model, and path planning rules are continuously optimized to ensure continuous improvement of path planning and optimization. BRIEF DESCRIPTION OF THE DRAWINGS

[0073] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0074] Figure 1 This is a flow chart of a personalized online learning path planning and optimization method proposed by the present invention. DETAILED DESCRIPTION

[0075] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, which only illustrate the basic structure of the present invention in a schematic manner, and therefore only show the components related to the present invention.

[0076] refer to Figure 1 , a personalized online learning path planning and optimization method, comprising the following steps:

[0077] S1. Obtain short-term goals, long-term goals and implicit learning needs through user interaction, analyze the goals into primary goals, secondary goals and extended goals, build a goal map, and generate a dynamic priority matrix based on the time priority, relevance and interest changes of the goals;

[0078] S2. According to the dynamic priority matrix, select knowledge nodes related to the target from the knowledge graph, perform semantic decomposition and reconstruction on the selected nodes, and generate a set of knowledge units based on the user's learning background and scenario characteristics;

[0079] S3. Construct a modular learning path based on a set of knowledge units. The path includes a main path, an auxiliary path, and a flexible path. The modular path structure is used to realize dynamic association and nonlinear logic jumps between paths. The path design dynamically adjusts the order and content layout of target nodes according to a dynamic priority matrix.

[0080] S4. During the execution of the learning path, collect user behavior data, including node dwell time, completion rate, and feedback results, adjust the content and order of the path nodes according to the behavior data, and introduce background knowledge, predictive knowledge, and practical tasks related to the current node through the expansion mechanism to update the content depth and logical order of the path;

[0081] S5. Conduct real-time value assessment of nodes in the modular learning path, prune according to node completion rate, importance weight and path logic, remove low-priority nodes, and add verification tasks to the core nodes of the main path to dynamically strengthen the mastery of key knowledge points and optimize the focus and coherence of the path structure;

[0082] S6. After completing the current path, based on the dynamic priority matrix and pruning optimization results, analyze the relevance of the core knowledge units in the current path with the new goal, migrate and integrate the core nodes into the new path framework, generate a multi-scenario learning path, and support the learning needs of the new goal;

[0083] S7. Display the path structure, learning progress and adjustment results through a visual interface, input user behavior data and path execution data into the dynamic priority matrix update model, optimize the target sorting and path planning rules, and form a closed-loop optimization mechanism for path planning.

[0084] In this implementation manner, the S1 specifically includes:

[0085] S11, collect the user's learning demand information through user interaction, the learning demand information includes short-term goals, long-term goals and implicit goals, and construct a goal set G = {g1, g2, ..., g n}, where g i is the i-th target node in the target set;

[0086] S12. Analyze each target node in the target set, decompose the target into primary target, secondary target and extended target, and construct a target map. The target map represents the target nodes and the relationship between them. The relationship is defined as R ij =f(g i, g j ), where R ij Represents the target node g i and g j The strength of the association between f The functional form of the dependent characteristics;

[0087] S13. Calculate the target priority weight W based on the target node characteristics in the target graph i :

[0088]

[0089] Among them, W i is the target node g i The priority weight, P t (g i ) is the time urgency of the target node, P r (g i ) is the relevance weight of the target node, P h (g i ) is the interest weight of the target node, α t , α r , α h are the moderators of time, relevance, and interest weights, respectively;

[0090] S14, based on the calculated priority weight W i , sort the target nodes and generate a dynamic priority matrix M = [W1, W2, ..., W n ];

[0091] S15. During the user learning process, collect user behavior feedback data, including completion rate, dwell time and interest changes of target nodes, and dynamically adjust the target priority weight according to the feedback data;

[0092] S16. Combined with the dynamically updated priority matrix M, adjust the association relationship and sorting rules of the target nodes in the target graph to generate an optimized target graph.

[0093] In this implementation, S2 specifically includes:

[0094] S21. According to the target weight in the dynamic priority matrix, a set of knowledge nodes related to the target node is selected from the optimized target map. The selection of knowledge nodes is based on the correlation strength R between the target node and the knowledge node.ij :

[0095]

[0096] Among them, R ij Represents the target node g i With knowledge node k j The strength of association, d(g i ,k k ) is the target node g i With knowledge node k k The shortest path length, P(k j ) represents knowledge node k j The set of neighboring nodes, ρ(k j ) is the semantic relevance of knowledge nodes, γ and δ are the adjustment factors of path length and semantic strength, respectively, and max(d) is the longest path length in the graph;

[0097] S22, perform semantic decomposition on the selected knowledge nodes, and decompose each knowledge node into core knowledge units and extended knowledge units, where the core knowledge unit C ij Directly supports target learning needs and expands knowledge unit E ij To supplement background information or practice tasks, the semantic decomposition rules are dynamically adjusted according to the association weights between nodes and targets;

[0098] S23. Combine the user's learning background and learning scenario characteristics to analyze the decomposed core knowledge unit C ij and Extended Knowledge Unit E ij Perform dynamic reconstruction to generate a knowledge unit set K' i ={k′ i1 ,k' i2 ,…,k′ ip};

[0099] S24, for the dynamically reconstructed knowledge unit set, according to the priority weight W of the target node i The correlation strength R of the knowledge node ij To perform priority scoring:

[0100] S k′ =W i α+R ij (1-α);

[0101] Among them, S k′ Score the comprehensive priority of the knowledge unit, W i is the priority weight of the target node, R ij is the association strength between the target and the knowledge node, α is the weight distribution coefficient;

[0102] S25. According to the scoring results, filter the knowledge units with higher priorities and construct the final set of knowledge units.

[0103] In this implementation, S3 specifically includes:

[0104] S31, based on the final knowledge unit set and the dynamic priority matrix, the knowledge units are divided into a modular path structure according to the priority weight of the target node, and the modular path includes a main path, an auxiliary path and a flexible path;

[0105] S32, the main path is composed of core knowledge units, the auxiliary path is composed of extended knowledge units, the elastic path is used to connect the main path and the auxiliary path, and the dynamic association between the path modules is established through non-linear jump logic;

[0106] S33, according to the weight of the target node in the dynamic priority matrix, the nodes in the modular path are prioritized to optimize the path layout;

[0107] S34, design nonlinear jump logic in modular paths, through the jump probability matrix T = [t ij ] Determine the jump relationship between path nodes, where t ij Represents the path node p i Jump to node p j Probability:

[0108]

[0109] Among them, P rank (p i ) represents the path node p i The priority score, R ij Represents the path node p i and p j The strength of logical association;

[0110] S35, dynamically adjust the generated modular path in combination with the user learning behavior data, including the user's stay time, completion rate and feedback data, adjust the order and jump relationship of the path nodes, and form a preliminary optimized modular path set P′={P1,P2,…,P q}.

[0111] In this implementation manner, the S4 specifically includes:

[0112] S41, in the formed modular path set P′={P1,P2,…,P q}, perform learning path planning, collect user behavior data, and construct a behavior data set D = {d1, d2, …, d n}, where d i Include path node pi The residence time t(p i ), completion rate c(p i ) and user feedback f(p i );

[0113] S42, using the behavior data set D, dynamically adjust the priority weights of the path nodes, and adjust the priority weights P rank (p i ) to update and recalculate the comprehensive priority of the path nodes:

[0114] P rank ′(p i )=P rank (p i )+γ1·t(p i )+γ2·c(p i )+γ3·f(p i );

[0115] Among them, P rank ′(p i ) is the updated priority weight, P rank (p i ) is the initial priority weight, γ1, γ2, γ3 are the adjustment factors of residence time, completion rate and user feedback respectively;

[0116] S43, according to the updated priority weight P rank ′(p i ), dynamically expand the path node content, optimize the content depth and logical structure of the path node by introducing background knowledge, predictive knowledge and practical tasks, and generate the expanded path node set p i ′;

[0117] S44, based on the expanded path node set p i ′ and the updated priority weight P rank ′(p i ), optimize the jump logic of the modular path, and update the jump probability matrix T'=[t ij '], where the jump probability t ij 'Dynamically adjust through node priority and logical association strength;

[0118] S45, integrating the expanded and updated path node content and the dynamic jump logic, and feeding back to the modular path structure, generating the final optimized modular path set P″={P″1,P″2,…,P″ q}, the path set P″ contains the updated priority weight P rank '(p i ), node jump logic T' and user behavior data feedback.

[0119] In this implementation manner, S5 specifically includes:

[0120] S51, generating the final optimized modular path set P″={P″1,P″2,…,P″ q}, the learning completion rate, importance weight and path logic of the path nodes are evaluated in real time, and the node value evaluation set V = {v1, v2, …, v n}:

[0121]

[0122] Among them, v i Represents the path node p i The comprehensive value of i ) is the node p i The learning completion rate, W'(p i ) is the priority weight of the node, is the association strength of the node in the path logic, α and β are weight factors, which dynamically adjust the contribution ratio of each dimension when evaluating the node value;

[0123] S52: According to the node value evaluation set V, prune the low-value nodes, remove the nodes whose comprehensive value is lower than the pruning threshold θ, and generate the pruned modular path set P″ pruned ;

[0124] S53, the modular path set P″ after pruning pruned In the process, the core node of the main path is selected to design the verification task. The verification task is associated with the knowledge point of the core node. The completion rate R(p i ) Update the priority weight of the core node;

[0125] S54. According to the core node priority and pruning results fed back by the verification task, the node order and jump logic in the modular path are optimized, the nodes are rearranged, the content focus and logical coherence of the path are optimized, and the enhanced modular path set P″′={P″′1,P″′2,…,P″′ q}.

[0126] In this implementation manner, S6 specifically includes:

[0127] S61. Based on the pruning optimization result and the dynamic priority matrix, extract the current path set P″′={P″′1, P″′2, …, P″′ q}, analyze the core knowledge unit k i Relevance to the new learning goal set G′:

[0128] R′(ki ,g′ j )=(α·A(k i ,g′ j )+β·C(k i ,g′ j ))+γ·W′(k i );

[0129] Among them, R′(k i ,g′ j ) represents node k i With the new target g j ′’s priority association, A(k i ,g′ j ) represents the semantic association strength, C(k i ,g′ j ) represents context adaptability, W′(k i ) is the importance weight of the node, α, β, γ are adjustment factors;

[0130] S62, according to the priority association R′(k i ,g′ j ), sort the core knowledge unit set K, give priority to the nodes with high correlation, and migrate them to the new path framework to form the core knowledge unit set K′ that adapts to the new goal;

[0131] S63. Based on the core knowledge unit set K′, the extended knowledge unit and the context knowledge unit are combined to construct a multi-scenario adaptation path set P scene ;

[0132] S64: Adapting the path set P for multiple scenarios scene The content layout and logical structure are optimized, the order and jump relationship of the path nodes are adjusted according to the dynamic priority matrix, and the target adaptation path set is generated.

[0133] In this implementation manner, the S7 specifically includes:

[0134] S71. In the generated target adaptation path set Based on the above, the path structure is displayed through a visual interface, including the priority score of the path node, the logical jump relationship, the learning progress and the adjustment results;

[0135] S72, collect user learning behavior data, and construct a behavior data set D = {d1, d2, ..., d v}, where d i Include path node p i The residence time t(p i ), completion rate c(p i), feedback result f(p i ) and node association behavior B(p i );

[0136] S73, input the behavior data set D and the path execution data into the dynamic priority matrix update model, and recalculate the priority score P′ of the path node rank (p i ), update the path planning rules;

[0137] S74. Score P′ based on the updated priority rank (p i ), adjust the content layout and logical order of the path nodes, rearrange the path nodes according to the priority weights, and generate an optimized path set

[0138] S75: Adapt the generated target path set P adapt and the optimized path set P final The integration is fed back into the dynamic priority matrix update model to optimize the target sorting and path planning rules, forming a closed-loop path optimization mechanism.

[0139] Embodiment 1:

[0140] In order to verify the feasibility of the present invention in implementation, the present invention is applied to the field of intelligent education. The demand for personalized learning is becoming increasingly urgent. An online learning platform aims to provide personalized course tutoring services for junior and senior high school students. However, the existing course recommendation system can only recommend fixed learning content based on the user's subject selection and lacks the ability to dynamically adjust. As a result, it is difficult to accurately cover students' learning goals, the path planning efficiency is low, and the learning effect is not ideal. To solve these problems, the platform introduced the personalized online learning path planning and optimization method of the present invention, and applied and verified it in real scenarios.

[0141] A high school student named Li needs to improve his math scores. He also plans to consolidate his geometry and function knowledge in a short period of time and strengthen the parts that are easy to lose points in the exam. Li's long-term goal is to fully master high school math knowledge, and his implicit need is to improve his interest in math through interesting learning methods. According to the method of the present invention, the platform collects Li's learning goals and needs through interaction with him, and generates a dynamic priority matrix through time urgency, relevance and interest changes. The platform analyzes these goals and decomposes them into the main goal of "geometry reinforcement", the secondary goal of "function consolidation" and the extended goal of "special training for lost points in the exam".

[0142] Based on the target analysis results and the existing knowledge graph, the platform further screens out core knowledge nodes related to geometry and functions, such as "determination of similar triangles" and "properties of quadratic functions". Through dynamic priority matrix adjustment, these knowledge nodes are decomposed into core knowledge units and extended knowledge units. Combined with Li's learning background and preferences, they are dynamically reconstructed into multiple knowledge unit sets such as "geometry basics reinforcement unit", "function question type training unit" and "wrong question special unit".

[0143] In actual application, the platform built a modular learning path for Li. The main path consists of core knowledge units to help Li quickly master the necessary knowledge; the auxiliary path contains extended knowledge units to supplement the background knowledge of geometry and functions; the flexible path is mainly based on wrong question training and practical application tasks, providing Li with a flexible learning method. During the execution of the learning path, the system collects Li's behavioral data in real time, including residence time, completion rate and learning feedback. For example, at the "Determination of Similar Triangles" node, Li's completion rate is 80%, and the feedback is "medium difficulty in understanding"; while at the "Properties of Quadratic Functions" node, the completion rate is only 40%, and the feedback is "complex question type, additional practice is needed".

[0144] Based on the above data, the platform dynamically adjusted the priority weights and content layout of the path nodes. For example, "Determination of Similar Triangles" was marked as completed and jumped directly to the next knowledge point; and the "Properties of Quadratic Functions" node was expanded into two sub-nodes, "Function Property Analysis" and "Typical Question Analysis", adding more background knowledge and practice tasks. In addition, to verify the mastery of knowledge points, the system embedded special verification tasks in the core nodes, such as exam simulations of function questions. Li's completion rate gradually increased from 50% in the first time to 85%.

[0145] After completing the initial learning path, the system analyzes the correlation between the core knowledge units in the current path and the next stage goal of "special training for lost marks in exams" based on the pruning optimization results and the dynamic priority matrix. Through quantitative calculation of the correlation, the system migrates the highly correlated core knowledge units to the new path framework, such as migrating the "properties of quadratic functions" to the "special topic for lost marks in exams", and integrates relevant background knowledge and practical tasks to generate a multi-scenario learning path.

[0146] The platform showed Li the structure of the learning path, learning progress and optimization results through a visual interface. For example, the completion rate of the main path was 90%, and the completion rate of the auxiliary path was 70%. The system input the user behavior data and the path execution data into the dynamic priority matrix update model, continuously optimized the target sorting and path rules, and formed a closed-loop optimization mechanism. The data showed that after the path planning and optimization method of the present invention, Li's math score increased from 62 points in the mock exam to 85 points, and his learning efficiency increased by 30%. At the same time, his learning interest increased from 3 points (out of 5 points) in the initial survey to 4.5 points, which fully demonstrated the significant advantages of the method of the present invention in personalized learning.

[0147] Table 1 Learning objectives and path node optimization table

[0148]

[0149] Table 2 Student behavior and verification task results

[0150]

[0151] From the above tables, we can see that Table 1 shows the division of short-term goals, long-term goals and implicit goals, and combines the weights of each goal in the priority matrix to generate a dynamic learning path, reflecting the accuracy and dynamic adjustment ability of the method in goal management. The change of priority weights intuitively reflects the dynamic changes of user needs and its optimization effect on the learning path; Table 2 focuses on reflecting the actual effect of optimizing the learning path through behavioral feedback data. The real-time collection and analysis of residence time, completion rate and user feedback directly affect the priority adjustment and content optimization of path nodes. By comparing the path efficiency data before and after optimization, the effectiveness of the pruning and reinforcement strategies is clearly verified. The optimized path finally generated not only improves students' completion rate and learning efficiency, but also significantly improves users' satisfaction with content relevance and path rationality.

[0152] In summary, the present invention realizes the personalization, dynamism and efficiency of the learning path through a dynamic priority matrix, modular path planning and closed-loop optimization mechanism. It not only solves the problem of single path and difficulty in adjustment in traditional methods, but also significantly improves the learning effect, providing a high-value application solution for online education platforms.

[0153] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A personalized online learning path planning and optimization method, characterized in that: The steps include: S1. Obtain short-term goals, long-term goals and implicit learning needs through user interaction, analyze the goals into primary goals, secondary goals and extended goals, build a goal map, and generate a dynamic priority matrix based on the time priority, relevance and interest changes of the goals; S2. According to the dynamic priority matrix, select knowledge nodes related to the target from the target graph, perform semantic decomposition and reconstruction on the selected nodes, and generate a set of knowledge units based on the user's learning background and scenario characteristics; S3. Construct a modular learning path based on a set of knowledge units. The path includes a main path, an auxiliary path, and a flexible path. The modular path structure is used to realize dynamic association and nonlinear logic jumps between paths. The path design dynamically adjusts the order and content layout of target nodes according to a dynamic priority matrix. S4. During the execution of the learning path, collect user behavior data, including node dwell time, completion rate, and feedback results, adjust the content and order of the path nodes according to the behavior data, and introduce background knowledge, predictive knowledge, and practical tasks related to the current node through the expansion mechanism to update the content depth and logical order of the path; S5. Conduct real-time value assessment of nodes in the modular learning path, prune according to node completion rate, importance weight and path logic, remove low-priority nodes, and add verification tasks to the core nodes of the main path to dynamically strengthen the mastery of key knowledge points and optimize the focus and coherence of the path structure; S6. After completing the current path, based on the dynamic priority matrix and pruning optimization results, analyze the relevance of the core knowledge units in the current path with the new goal, migrate and integrate the core nodes into the new path framework, generate a multi-scenario learning path, and support the learning needs of the new goal; S7. Display the path structure, learning progress and adjustment results through a visual interface, input user behavior data and path execution data into the dynamic priority matrix update model, optimize the target sorting and path planning rules, and form a closed-loop optimization mechanism for path planning.

2. A personalized online learning path planning and optimization method according to claim 1, characterized in that: The S1 specifically includes: S11. Collect the user's learning demand information through user interaction, wherein the learning demand information includes short-term goals, long-term goals and implicit goals, and construct a target set G = g1, g2, ..., g n }, where g i is the i-th target node in the target set; S12. Analyze each target node in the target set, decompose the target into primary target, secondary target and extended target, and construct a target map. The target map represents the target nodes and the relationship between them. The relationship is defined as R ij =f(g i ,g j ), where R ij Represents the target node g i and g j The strength of association between them, f represents the functional form of the dependent characteristics; S13. Calculate the target priority weight W based on the target node characteristics in the target graph i : Among them, W i is the target node g i The priority weight, P t (g i ) is the time urgency of the target node, P r (g i is the relevance weight of the target node, P h (g i ) is the interest weight of the target node, α t , α r , α h are the moderators of time, relevance, and interest weights, respectively; S14, based on the calculated priority weight W i , sort the target nodes and generate a dynamic priority matrix M = [W1, W2, ..., W n 】; S15. During the user learning process, collect user behavior feedback data, including completion rate, dwell time and interest change of target nodes, and dynamically adjust the target priority weight according to the feedback data; S16. Combined with the dynamically updated priority matrix M, adjust the association relationship and sorting rules of the target nodes in the target graph to generate an optimized target graph.

3. A personalized online learning path planning and optimization method according to claim 1, characterized in that: The S2 specifically includes: S21. According to the target weight in the dynamic priority matrix, a set of knowledge nodes related to the target node is selected from the optimized target map. The selection of knowledge nodes is based on the correlation strength R between the target node and the knowledge node. ij : Among them, R ij Represents the target node g i With knowledge node k j The strength of association, d(g i ,k k ) is the target node g i With knowledge node k k The shortest path length, P(k j ) represents knowledge node k j The set of neighboring nodes, ρ(k j ) is the semantic relevance of knowledge nodes, γ and δ are the adjustment factors of path length and semantic strength, respectively, and max(d) is the longest path length in the graph; S22, perform semantic decomposition on the selected knowledge nodes, and decompose each knowledge node into core knowledge units and extended knowledge units, where the core knowledge unit C ij Directly supports target learning needs and expands knowledge unit E ij To supplement background information or practical tasks, the semantic decomposition rules are dynamically adjusted according to the association weights between nodes and targets; S23. Combine the user's learning background and learning scenario characteristics to analyze the decomposed core knowledge unit C ij and Extended Knowledge Unit E ij Perform dynamic reconstruction to generate a knowledge unit set K' i ={k′ i1 ,k' i2 ,…,k′ ip }; S24, for the dynamically reconstructed knowledge unit set, according to the priority weight W of the target node i The correlation strength R of the knowledge node ij To perform priority scoring: S k′ =W i ·a+R ij ·(1-a); Among them, S k′ Score the comprehensive priority of the knowledge unit, W i is the priority weight of the target node, R ij is the association strength between the target and the knowledge node, α is the weight distribution coefficient; S25. According to the scoring results, filter the knowledge units with higher priorities and construct the final set of knowledge units.

4. A personalized online learning path planning and optimization method according to claim 1, characterized in that: The S3 specifically includes: S31, based on the final knowledge unit set and the dynamic priority matrix, the knowledge units are divided into a modular path structure according to the priority weight of the target node, and the modular path includes a main path, an auxiliary path and a flexible path; S32, the main path is composed of core knowledge units, the auxiliary path is composed of extended knowledge units, the elastic path is used to connect the main path and the auxiliary path, and the dynamic association between the path modules is established through non-linear jump logic; S33, according to the weight of the target node in the dynamic priority matrix, the nodes in the modular path are prioritized to optimize the path layout; S34, design nonlinear jump logic in modular paths, through the jump probability matrix T = [t ij ] Determine the jump relationship between path nodes, where t ij Represents the path node p i Jump to node p j Probability: Among them, P rank (p i ) represents the path node p i The priority score, R ij Represents the path node p i and p j The strength of logical association; S35, dynamically adjust the generated modular path in combination with the user learning behavior data, including the user's stay time, completion rate and feedback data, adjust the order and jump relationship of the path nodes, and form a preliminary optimized modular path set P′={P1,P2,…,P q }.

5. A personalized online learning path planning and optimization method according to claim 1, characterized in that: The S4 specifically includes: S41, in the formed modular path set P′={P1,P2,…,P q }, perform learning path planning, collect user behavior data, and construct a behavior data set D = {d1, d2, …, d n }, where d i Include path node p i The residence time t(p i ), completion rate c(p i ) and user feedback f(p i ); S42, using the behavior data set D, dynamically adjust the priority weights of the path nodes, and adjust the priority weights P rank (p i ) to update and recalculate the comprehensive priority of the path nodes: P rank ′(pi)=P rank (p i )+γ1·t(p i )+γ2·c(p i )+γ3·f(p i ); Among them, P rank ′(p i ) is the updated priority weight, P rank (p i ) is the initial priority weight, γ1, γ2, γ3 are the adjustment factors of residence time, completion rate and user feedback respectively; S43, according to the updated priority weight P rank ′(p i ), dynamically expand the path node content, optimize the content depth and logical structure of the path node by introducing background knowledge, predictive knowledge and practical tasks, and generate the expanded path node set p i ′; S44, based on the expanded path node set p i ′ and the updated priority weight P rank ′(p i ), optimize the jump logic of the modular path, and update the jump probability matrix T′=[t ij ′], where the jump probability t ij ′Dynamically adjust through node priority and logical association strength; S45, integrating the expanded and updated path node content and dynamic jump logic, and feeding back to the modular path structure to generate the final optimized modular path set P ″ = {P ″ 1,P ″ 2,…,P ″ q }, path set P ″ Contains the updated priority weight P rank ′(p i ), node jump logic T′ and user behavior data feedback.

6. A personalized online learning path planning and optimization method according to claim 1, characterized in that: The S5 specifically includes: S51, in the generated final optimized modular path set P ″ = {P ″ 1,P ″ 2,…,P ″ q }, the learning completion rate, importance weight and path logic of the path nodes are evaluated in real time, and the node value evaluation set V = {v1, v2, …, v n }: Among them, v i Represents the path node p i The comprehensive value of i ) is the node p i The learning completion rate, W′(p i ) is the priority weight of the node, is the association strength of the node in the path logic, α and β are weight factors, which dynamically adjust the contribution ratio of each dimension when evaluating the node value; S52: According to the node value evaluation set V, prune the low-value nodes, remove the nodes whose comprehensive value is lower than the pruning threshold θ, and generate the pruned modular path set P″ pruned ; S53, the modular path set P″ after pruning pruned In the process, the core node of the main path is selected to design the verification task. The verification task is associated with the knowledge point of the core node. The completion rate R(p i ) Update the priority weight of the core node; S54. According to the core node priority and pruning results fed back by the verification task, the node order and jump logic in the modular path are optimized, the nodes are rearranged, the content focus and logical coherence of the path are optimized, and the enhanced modular path set P″′={P″′1,P″′2,…,P″′ q }.

7. A personalized online learning path planning and optimization method according to claim 1, characterized in that: The S6 specifically includes: S61. Based on the pruning optimization result and the dynamic priority matrix, extract the current path set P″′={P″′1, P″′2, …, P″′ q }, analyze the core knowledge unit k i Relevance to the new learning goal set G′: R′(k i ,g′ j )=(α·A(k i ,g′ j )+β·C(k i ,g′ j ))+γ·W′(k i ); Among them, R′(k i ,g′ j ) represents node k i With the new target g′ j The priority association degree, A(k i ,g′ j ) represents the semantic association strength, C(k i ,g′ j ) represents context adaptability, W′(k i ) is the importance weight of the node, α, β, γ are adjustment factors; S62, according to the priority association R′(k i ,g′ j ), sort the core knowledge unit set K, give priority to the nodes with high correlation, and migrate them to the new path framework to form the core knowledge unit set K′ that adapts to the new goal; S63. Based on the core knowledge unit set K′, the extended knowledge unit and the context knowledge unit are combined to construct a multi-scenario adaptation path set P scene ; S64: Adapting the path set P for multiple scenarios scene The content layout and logical structure are optimized, the order and jump relationship of the path nodes are adjusted according to the dynamic priority matrix, and the target adaptation path set is generated.

8. A personalized online learning path planning and optimization method according to claim 1, characterized in that: The S7 specifically includes: S71. In the generated target adaptation path set Based on the above, the path structure is displayed through a visual interface, including the priority score of the path node, the logical jump relationship, the learning progress and the adjustment results; S72, collect user learning behavior data, and construct a behavior data set D = {d1, d2, ..., d v }, where d i Include path node p i The residence time t(p i ), completion rate c(p i ), feedback result f(p i ) and node association behavior B(p i ); S73, input the behavior data set D and the path execution data into the dynamic priority matrix update model, and recalculate the priority score P of the path node ′ rank (p i ), update the path planning rules; S74. According to the updated priority score P ′ rank (p i ), adjust the content layout and logical order of the path nodes, rearrange the path nodes according to the priority weights, and generate an optimized path set S75, the generated target adaptation path set P adapt and the optimized path set P final The integration is fed back into the dynamic priority matrix update model to optimize the target sorting and path planning rules, forming a closed-loop path optimization mechanism.

Images