Self-adaptive education gamification learning system design method, device, equipment and medium

Abstract

The invention relates to an adaptive education gamification learning system design method and device, equipment and a medium. According to the method, a knowledge graph containing knowledge point nodes and associated edges is constructed, memory parameters are initialized, and based on the graph, a memory intensity dynamic model is utilized to simulate the attenuation and mutual influence process of knowledge point memory intensity, so that dynamic memory intensity distribution is obtained; when a review event is detected, updating the memory intensity of the reviewed node and the associated node through the review intervention model to obtain an updated memory state; on this basis, an optimal review path which maximizes the overall memory stability with the minimum review cost is generated by using a path planning algorithm, and the path is converted into a gamification task, and finally, the learning path is adaptively planned through structured knowledge modeling and dynamic memory optimization, so that the learning efficiency is effectively improved, the cognitive load is reduced, and the learning efficiency is improved. And the learning motivation and the participation degree are enhanced through gamification elements.

Description

Technical Field

[0001] This invention belongs to the field of educational technology, and in particular relates to a design method, device, equipment and medium for an adaptive gamified learning system. Background Technology

[0002] In current adaptive learning systems, spaced repetition-based algorithms have become the mainstream technique for improving memory retention. These systems typically rely on the Ebbinghaus forgetting curve theory or its improved models, treating each knowledge point as an independent memory unit and calculating the optimal review time for it. However, this atomistic approach has significant limitations, ignoring the inherent network structure of knowledge itself. In real-world learning scenarios, knowledge points exist within complex preconditions and logical connections, forming an organic knowledge system. When learners forget basic concepts, this memory decay propagates along the knowledge network, affecting the mastery of related higher-level concepts as well. This phenomenon of "associative forgetting" is completely unconsidered in existing independent memory item models.

[0003] Another prominent problem with existing systems is their local optimization strategy. Whether it's an adaptive algorithm based on half-life theory or a traditional interval repetition scheme, both only plan the review of individual knowledge points. This approach can lead to low learning efficiency; learners may repeatedly review isolated knowledge points while neglecting the coherence and stability of the entire knowledge system. More importantly, when reviewing a core foundational concept, its positive effects can radiate to multiple related knowledge points; this network gain cannot be quantified and utilized in existing models.

[0004] From a technical implementation perspective, most current systems focus on fitting and predicting a single forgetting curve, lacking the ability to model the dynamic interactions of knowledge structures. Review decisions are typically based on simple memory strength thresholds or time intervals, failing to assess the impact of different review strategies on overall knowledge mastery from a holistic perspective. This limitation makes it difficult to truly personalize learning path planning, hindering intelligent adjustments based on the learner's actual knowledge structure.

[0005] Therefore, there is an urgent need for a new learning system that can model the dynamic interactions within a knowledge network and perform global optimization planning based on this. Summary of the Invention

[0006] Therefore, it is necessary to provide a design method, device, equipment, and medium for an adaptive gamified learning system to address the aforementioned technical problems.

[0007] Firstly, this application provides a design method for an adaptive gamified learning system, including:

[0008] S1. Based on the attribute information and relationships of knowledge points, construct a basic knowledge graph containing nodes and edges, and initialize the memory strength parameter and forgetting rate parameter for each node in the basic knowledge graph, and initialize the edge weight parameter for each edge to obtain the initial knowledge graph;

[0009] S2. Based on the initial knowledge graph, the memory intensity of each node decays over time and interacts with each other through the edge by simulating the memory intensity dynamics model, thus obtaining the memory intensity distribution.

[0010] S3. When a review event is detected, based on the memory intensity distribution and the initialized knowledge graph, the memory intensity of the reviewed node and its neighboring nodes is updated through the review intervention model to obtain the updated memory intensity distribution.

[0011] S4. Based on the updated memory strength distribution, the optimal review path is generated through a path planning algorithm, and the review sequence is obtained according to the optimal review path;

[0012] S5. Based on the optimal review path and basic knowledge graph, the review sequence is transformed into gamified tasks.

[0013] Secondly, this application also provides an adaptive educational gamified learning system design apparatus for implementing the method described in the first aspect, the apparatus comprising:

[0014] The knowledge graph initialization module is used to construct a basic knowledge graph containing nodes and edges based on the attribute information and relationships of knowledge points. It initializes the memory strength parameter and forgetting rate parameter for each node in the basic knowledge graph and initializes the edge weight parameter for each edge to obtain the initial knowledge graph.

[0015] The memory dynamic simulation module is used to simulate the decay of the memory strength of each node over time and the process of mutual influence through edges based on the initial knowledge graph, and to obtain the memory strength distribution.

[0016] The review intervention module is used to update the memory intensity of the reviewed node and its neighboring nodes through the review intervention model when a review event is detected, based on the memory intensity distribution and the initialized knowledge graph, to obtain the updated memory intensity distribution.

[0017] The path planning module is used to generate the optimal review path based on the updated memory strength distribution using a path planning algorithm, and to obtain the review sequence based on the optimal review path;

[0018] The gamification conversion module is used to transform review sequences into gamified tasks based on the optimal review path and basic knowledge graph.

[0019] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement an adaptive educational gamified learning system design method as described in the first aspect.

[0020] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements an adaptive educational gamified learning system design method as described in the first aspect.

[0021] The aforementioned adaptive gamified learning system design method, device, equipment, and medium construct a knowledge graph containing knowledge point nodes and associated edges, initialize memory parameters, and use a memory intensity dynamics model based on this graph to simulate the decay and mutual influence process of knowledge point memory intensity, thereby obtaining a dynamic memory intensity distribution. When a review event is detected, the memory intensity of the reviewed node and associated nodes is updated through a review intervention model to obtain the updated memory state. On this basis, a path planning algorithm is used to generate an optimal review path that maximizes overall memory stability with minimal review cost, and this path is transformed into a gamified task. Ultimately, adaptive planning of learning paths through structured knowledge modeling and dynamic memory optimization is achieved, effectively improving learning efficiency, reducing cognitive load, and enhancing learning motivation and participation through gamified elements. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 A flowchart illustrating an adaptive gamified learning system design method provided by this invention;

[0024] Figure 2 This is a schematic diagram of the process for obtaining memory intensity distribution in an optional embodiment of the present invention;

[0025] Figure 3 This is a schematic diagram of the structure of an adaptive educational gamified learning system design device provided by the present invention. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0027] refer to Figure 1 The document presents a flowchart illustrating an adaptive gamified learning system design method provided in this application, which includes the following steps:

[0028] S1. Based on the attribute information and relationships of knowledge points, construct a basic knowledge graph containing nodes and edges, and initialize the memory strength parameter and forgetting rate parameter for each node in the basic knowledge graph, and initialize the edge weight parameter for each edge to obtain the initialized knowledge graph.

[0029] Specifically, the attribute information of knowledge points covers core dimensions such as difficulty level, knowledge module affiliation, cognitive level requirements (e.g., memorization, understanding, application, analysis), and estimated mastery time. This information is obtained through a combination of domain expert annotation and historical learning data statistics. For example, knowledge modules are divided with reference to curriculum standards in the education field, and the average mastery time for a particular knowledge point is determined based on a large amount of learner learning time data. The relationships between knowledge points include prerequisite dependencies, logical deduction relationships, and content relationships. These relationships are structured and stored using knowledge graph construction tools (such as Neo4j) to form a basic knowledge graph containing nodes (representing knowledge points) and edges (representing relationships).

[0030] Based on the fundamental knowledge graph, memory strength and forgetting rate parameters are initialized for each node. The memory strength parameter quantifies the learner's mastery of the corresponding knowledge point; its initial value is defined by considering both the initial cognitive difficulty of the knowledge point and the average learner's initial mastery level. ,in Representing the The node (i.e., the node) (One knowledge point) The value range can be set within the interval [0,1], where 0 represents complete lack of understanding and 1 represents complete understanding. For example, for basic concept knowledge points, the range can be set to [0,1]. Initialize to 0.8, but for complex application-related knowledge points, it can be initialized to 0.3. The forgetting rate parameter describes the rate at which the memory strength of knowledge points naturally decays over time, and is defined as follows: Similarly, it is related to the attributes of the knowledge points; knowledge points that are more difficult and abstract are usually forgotten at a higher rate. The value of is within the interval (0,1).

[0031] For each edge (i.e., the relationship between knowledge points), initialize the edge weight parameter, defined as follows: ,in and These represent the two nodes (i.e., two knowledge points) connected by the edge. The edge weight is used to quantify the strength of the relationship between two knowledge points, with a value ranging from [0,1]. The stronger the relationship, the larger the edge weight. The initialization of the edge weight parameter requires comprehensive calculation combining domain experts' assessment of the relationship strength with data such as the co-learning frequency and erroneous association rate of the two knowledge points in historical learning data, ensuring that the weight value accurately reflects the actual impact of the relationship. Through the initialization of the node and edge parameters, an initialized knowledge graph that can be used for subsequent calculations is finally obtained.

[0032] S2. Based on the initial knowledge graph, the memory intensity of each node decays over time and interacts with each other through the edge by simulating the memory intensity dynamics model, thus obtaining the memory intensity distribution.

[0033] Specifically, the core of the memory strength dynamics model is to simulate the dynamic change process of knowledge point memory strength. It considers both the natural decay of individual knowledge point memory strength over time and the mutual influence between different knowledge points through their relationships. The model's construction must be based on dynamic systems theory, treating the memory strength of each node as a dynamic variable that changes over time. The core formula of the model is defined as:

[0034]

[0035] in, Indicates the first Each node in time The rate of change of memory strength at any given time, that is, the increase or decrease in memory strength per unit of time; Indicates the first Each node in time The memory strength at any given moment changes dynamically over time, with the initial value set in S1. ; For the first The forgetting rate parameter for each node is defined in S1 and is used to quantify the decay effect of natural forgetting on memory strength. The negative sign indicates that memory strength decays over time under natural conditions. Indicates the first The set of neighboring nodes of the nth node, that is, the set of neighboring nodes of the nth node connected by an edge. The determination of neighboring nodes requires traversing and initializing the knowledge graph to find all nodes directly connected to the first node. Get all associated edges of each node; For the first The node and the first The edge weight parameters between the nth nodes are consistent with those defined in S1 and are used to quantize the nth node. The node is the first The strength of the correlation influence of each node; Let be the neighbor node influence function, used to describe the th The memory strength of the nth node affects the nth node. The specific way in which the memory strength of each node is affected depends on the form of the function, which needs to be determined based on the type of knowledge association. For example, for prerequisite dependencies, a linear function can be used. (in The influence coefficient (value in the range (0,1), adjusted according to the strength of dependency), indicates that the higher the mastery of prerequisite knowledge points, the greater the positive impact on the memory strength of subsequent knowledge points; for logical derivation relationships, a nonlinear function can be used. This indicates that the degree of mastery of the source knowledge points in the derivation has a secondary increasing trend in its influence on the knowledge points in the derivation result.

[0036] In the specific simulation process of the model, numerical calculation methods are used to solve the above differential equations. Commonly used methods include the Euler method and the Runge-Kutta method. Taking the Euler method as an example, the time step is set... (The choice of time step needs to balance computational accuracy and efficiency, and can be set according to the learning cycle, such as in days or hours.) Divide the time axis into a series of discrete time points. (in At the initial moment, Within each time step, first calculate the natural decay of each node. Then calculate the total influence of all neighboring nodes on this node. The sum of the two yields the change in node memory strength within that time step, which is then used to update the node's memory strength. Intensity of memory at any moment By iteratively calculating all time steps, the memory intensity values ​​of all nodes at different time points can be obtained. By classifying and organizing these values ​​by node, the memory intensity distribution of the entire knowledge graph can be formed. This distribution can intuitively reflect the changes in learners' memory status of each knowledge point over time without review intervention.

[0037] S3. When a review event is detected, based on the memory intensity distribution and the initialized knowledge graph, the memory intensity of the reviewed node and its neighboring nodes is updated through the review intervention model to obtain the updated memory intensity distribution.

[0038] Specifically, the detection of review events is achieved through the system's built-in event monitoring module. Conditions triggering review events include learners actively initiating review actions (such as clicking a review button or selecting a specific knowledge point for review), automatic triggering based on memory intensity distribution (such as when the memory intensity of a certain node falls below a preset threshold), and triggering at learning cycle nodes (such as at fixed daily review times or after the end of a learning unit). Once a review event is detected, the set of nodes to be reviewed is determined, i.e., the nodes corresponding to the knowledge points targeted in this review, denoted as . (in (This represents the number of nodes to be reviewed). Simultaneously, by traversing and initializing all the associated edges of the reviewed nodes in the knowledge graph, the set of neighboring nodes for each reviewed node is determined. ( These neighboring nodes will be affected by the review intervention due to their association relationships.

[0039] The core of the review intervention model is to quantify the effect of review operations on improving the memory strength of the reviewed node and its neighboring nodes. The model's update formula is defined as:

[0040]

[0041] in, Indicates the first day after the review intervention Each node in time The strength of memory at any given moment, i.e., the strength of memory after the update; Indicates the review before intervention. Each node in time The memory intensity at any given time is taken from the memory intensity distribution obtained in S2; The intervention coefficient for the reviewed node is used to quantify the improvement in the memory strength of the node by direct review. Its value needs to be determined in combination with the review method (such as passive review, active practice, test, etc.) and the type of knowledge point. The value range is within the (0,1) interval to avoid exceeding the upper limit of 1 after the memory strength is updated; To review indicator variables, when the first... Each node belongs to the set of nodes to be reviewed. hour, This indicates that the node is subject to direct review intervention, when the first... The node does not belong to hour, This indicates that the node was not directly affected by the review process. This indicates that it simultaneously belongs to the set of nodes being reviewed. and the The set of neighboring nodes of a node The node, that is, the node with the first The nodes that are related to each other and are directly reviewed this time; For the first The node and the first The edge weight parameters between the nth nodes are defined in S1 and are used to quantify the edge weights of the reviewed neighbor nodes on the nth node. The strength of the correlation influence of each node; The correlation intervention coefficient is used to quantify the indirect increase in the memory strength of the current node by the reviewed neighbor nodes. Its value is adjusted according to the correlation relationship type, and the value range is also within the (0,1) interval. To review the pre-intervention period Each node (the reviewed neighbor node) in time The memory intensity at any given moment is taken from the memory intensity distribution in S2.

[0042] The specific update process is executed as follows: First, traverse all nodes in the initialized knowledge graph, and for each node... Determine whether it belongs to the set of nodes to be reviewed. ,Sure The value of; secondly, for For the nodes (the nodes being reviewed), calculate the increase in memory strength resulting from direct review. Simultaneously, find all neighboring nodes of the given node and calculate the indirect boost for each neighboring node resulting from the review of this node (this step is completed during the update of neighboring nodes); furthermore, for For a node (not the node being reviewed), find its neighboring nodes that belong to... The node, i.e. Calculate the total indirect boost of these reviewed neighbor nodes to that node. Finally, add the direct improvement (if any) and the indirect improvement (if any) to the memory strength before review. The updated memory strength is obtained. By performing the above update operation on all nodes, the updated memory intensity distribution of the entire knowledge graph can be obtained. This distribution can accurately reflect the changes in the memory status of each knowledge point after the review intervention, providing data support for subsequent optimal review path planning.

[0043] S4. Based on the updated memory strength distribution, the optimal review path is generated through a path planning algorithm, and the review sequence is obtained according to the optimal review path.

[0044] Specifically, the core objective of the path planning algorithm is to select the knowledge points that need to be reviewed from a global perspective based on the updated memory strength distribution, determine the optimal review order for these knowledge points, form the optimal review path, and thus obtain the review sequence. Its design should aim to maximize the overall memory stability of the knowledge system and minimize the risk of memory decay. First, the objective function of path planning needs to be determined to quantify the quality of the review path. The objective function is defined as:

[0045]

[0046] in, The index represents the overall evaluation of the review path. The higher the index value, the better the optimization effect of the path on the overall memory state of the knowledge system. , is a weighting coefficient used to balance the memory stability of individual nodes and the overall association stability of the knowledge system, and its value ranges from [0,1]. This represents the set of nodes to be reviewed. The selection of this set needs to be determined based on the updated memory strength distribution. Nodes below a preset memory threshold are included, i.e. ; For the updated version Each node in time The memory intensity at each moment is taken from the updated memory intensity distribution obtained in S3; For the first The forgetting rate parameter for each node is consistent with the definition in S1. The ability to quantify the memory decay resistance of a node is used; the product of the two can reflect the long-term stability of the node under the current memory strength. For the first The set of neighbor nodes of each node is consistent with the definition in S2; For the first The node and the first The edge weight parameters between nodes are consistent with those defined in S1; For the updated version Each node in time The memory strength at any given moment is summed to quantify the stability of the association memory between the node to be reviewed and its neighboring nodes. The larger the edge weight, the higher the memory strength between the two nodes. The larger this value, the stronger the association and coherence of the knowledge system.

[0047] After determining the objective function, a suitable path planning algorithm is selected to solve for the optimal path. Here, the improved Dijkstra algorithm is used as an example for detailed explanation. The algorithm implementation steps are as follows: First, initialize the set of nodes to be reviewed. Each node is assigned an initial path weight, which is set to the local contribution value of that node to the objective function. ,in The first step is to identify neighboring nodes that also belong to the set to be reviewed; the second step is to construct a path graph, where the nodes of the path graph represent the set of nodes to be reviewed. In the path graph, the edge weight is defined as the knowledge association cost between two nodes, i.e. The smaller the edge weight, the stronger the knowledge connection between the two nodes, and the more significant the improvement in overall memory stability when the review order is adjacent; the third step is to execute the improved Dijkstra algorithm, using the node with the smallest initial path weight as the starting node (the selection of the starting node should prioritize the node with the lowest memory strength and the highest forgetting rate, i.e., For each corresponding node, prioritize reinforcing the knowledge points with the highest memory risk. Each time, select the node with the lowest path weight from the nodes to be visited and add it to the visited path, while simultaneously updating the path weights of its neighboring nodes (update rule: ...). The fourth step involves arranging the nodes in the visited path in the order of visit to form the optimal review path, which satisfies the objective function. The goal is to maximize the effectiveness of the review process along this path, prioritizing the consolidation of weak knowledge points while simultaneously strengthening the coherence of the knowledge system through the review of closely related knowledge points. The fifth step involves extracting the corresponding sequence of knowledge points based on the node order of the optimal review path, thus obtaining a review sequence. This review sequence must clearly define the review priority of each knowledge point; for example, if the path is... The review sequence is as follows: .

[0048] During algorithm implementation, pay attention to the constraints of path optimization, such as review time constraints (setting an upper limit on the total duration of a single review session). Review time for each knowledge point Determined based on its memory strength and cognitive level, such as ,in This is the duration coefficient. To minimize the value, avoid a denominator of 0 and ensure the total duration. Cognitive load constraints (the difference in cognitive level between adjacent review points must be controlled within a preset range to avoid cognitive overload) can be introduced to make the generated review sequence more consistent with actual learning scenarios, thereby improving learning efficiency and user experience.

[0049] S5. Based on the optimal review path and basic knowledge graph, the review sequence is transformed into gamified tasks.

[0050] Specifically, the core of transforming review sequences into gamified tasks is establishing a mapping relationship between the review sequences and the gamified tasks. This mapping simultaneously satisfies the knowledge review objectives and the needs of the gamified experience. Furthermore, it is designed based on the priority order of the optimal review path and the correlation of the basic knowledge graph, ensuring that the gamified tasks accurately cover the knowledge points in the review sequence while strengthening the coherence of the knowledge system through the connections between tasks. First, the design framework for the gamified tasks is determined. This framework includes a task type matching module, a task difficulty adjustment module, a task association construction module, and a task reward design module. These modules work together to achieve the transformation of review sequences into gamified tasks.

[0051] The core function of the task type matching module is to match a corresponding gamified task type for each knowledge point in the review sequence based on the cognitive level attributes of the knowledge points (derived from the attribute information of the knowledge points in S1), ensuring that the task type is consistent with the learning objectives of the knowledge points. Task type matching is achieved through a pre-defined mapping rule base, which is jointly developed by domain education experts and game design experts. The rule base clearly defines the task types, task formats, and task objectives corresponding to different cognitive levels and different knowledge modules, ensuring the educational value and rationality of the task design.

[0052] The task difficulty adjustment module adjusts the difficulty of each gamified task based on the updated memory strength distribution (taken from S3), ensuring that the task difficulty matches the learner's current mastery of the knowledge points. This avoids frustration due to excessive difficulty or inefficiency due to insufficient difficulty. The difficulty adjustment employs a dynamic adjustment mechanism, with the core adjustment formula defined as:

[0053]

[0054] in, For the first in the review sequence The difficulty value of the gamified task corresponding to each knowledge point can be set in the range of [1,5], where 1 represents the lowest difficulty and 5 represents the highest difficulty. For the first The basic difficulty value of each knowledge point is determined by the difficulty level of the knowledge point itself (taken from the attribute information of the knowledge point in S1); The memory strength influence coefficient is used to quantify the negative impact of memory strength on task difficulty; that is, the higher the memory strength, the lower the task difficulty. The forgetting rate influence coefficient is used to quantify the positive impact of the forgetting rate on task difficulty. That is, the higher the forgetting rate, the more the task difficulty needs to be reduced to help learners rebuild their memories. For the updated version The memory strength of each node, For the first The forgetting rate parameters for each node are consistent with the definitions above. In addition to numerical adjustment, task difficulty can also be achieved by adjusting task parameters. For example, the difficulty of the "memory challenge" task can be adjusted by adjusting the time limit (the higher the difficulty, the shorter the time) and the number of errors tolerated (the higher the difficulty, the fewer the number of tolerances); the difficulty of the "problem solving" task can be adjusted by adjusting the complexity of the problem conditions (the higher the difficulty, the more hidden the conditions and the more interference items).

[0055] The task association construction module builds associations between gamified tasks based on the edge weight parameters (taken from S1) of the basic knowledge graph, ensuring that the task order aligns with the optimal review path, and that the associations between tasks strengthen the logical connections between knowledge points. The specific implementation is as follows: First, based on the node order of the optimal review path, the basic order of the gamified tasks is determined, i.e., the review sequence is... At that time, the basic order of tasks is as follows: ( For knowledge points (Corresponding gamified tasks); secondly, calculate the edge weights between knowledge points corresponding to adjacent tasks. ,like ( (For the preset association weight threshold), then in and The design incorporates "connecting tasks" that visually demonstrate the prerequisite dependencies between two knowledge points, strengthening the connection between them; if... Then in and Transitional tasks can be designed to guide the flow of tasks and prevent abrupt transitions. Additionally, hidden tasks can be designed based on non-adjacent relationships within the knowledge graph, such as knowledge points. and There is an indirect connection (through) , (Connection), then upon completion This will trigger the hidden quest "Comprehensive Application". , , , "Using knowledge points to solve cross-module problems" further strengthens the integrity of the knowledge system.

[0056] The task reward design module combines the strength of knowledge retention, task difficulty, and knowledge correlation to design a reward mechanism to enhance learner participation and review motivation. Reward types include points rewards, badge rewards, and task unlock rewards. The formula for calculating points rewards is defined as follows:

[0057]

[0058] in, To complete the knowledge points corresponding gamified tasks Points earned; The basic integral value (e.g., 100 points); This represents the task difficulty value, consistent with the previous definition; the higher the difficulty, the more points are awarded. For knowledge points The average side weight of the review sequence and other reviewed knowledge points, i.e. ( This represents the number of knowledge points that have been reviewed so far. For the first (For each reviewed knowledge point), the higher the average edge weight, the stronger the connection between that knowledge point and the reviewed knowledge, and the better the reinforcement effect on the knowledge system after completing the task. Therefore, extra points are awarded. The badge reward adopts a phased reward mechanism; when consecutive review sequences are completed... One task ( When a preset number of questions (e.g., 5) are completed and the average accuracy reaches a preset threshold (e.g., 90%), a "Knowledge Coherence Badge" is awarded. When a high-difficulty task at the cognitive level of "analysis" or "application" is completed, a "Skill Master Badge" is awarded. Task unlock rewards are linked to the relationships within the knowledge graph. After completing a task on a certain knowledge point, if its neighboring nodes are not included in the current review sequence but their memory strength is close to the threshold, a "preview task" for that neighboring node is unlocked, laying the foundation for subsequent review. Through multi-dimensional reward design, the attractiveness of gamified tasks can be effectively enhanced, stimulating learners' proactive review willingness, ultimately achieving an organic combination of education and playfulness.

[0059] The aforementioned adaptive gamified learning system design method constructs a knowledge graph containing knowledge point nodes and associated edges, initializes memory parameters, and uses a memory intensity dynamics model to simulate the decay and mutual influence of knowledge point memory intensity, thereby obtaining a dynamic memory intensity distribution. When a review event is detected, the memory intensity of the reviewed node and associated nodes is updated through a review intervention model, resulting in an updated memory state. Based on this, a path planning algorithm is used to generate an optimal review path that maximizes overall memory stability with minimal review cost, and this path is transformed into a gamified task. Ultimately, this achieves adaptive learning path planning through structured knowledge modeling and dynamic memory optimization, effectively improving learning efficiency, reducing cognitive load, and enhancing learning motivation and participation through gamified elements.

[0060] refer to Figure 2 In one optional embodiment, based on the initialized knowledge graph, the memory intensity distribution is obtained by simulating the decay of the memory intensity of each node over time and the process of mutual influence through edges using a memory intensity dynamics model, including the following steps:

[0061] S11. Based on the forgetting rate parameter and edge weight parameter of each node in the initialized knowledge graph, a partial differential equation for the change in memory strength is established, resulting in a partial differential equation model; the expression of the partial differential equation model is:

[0062]

[0063] in, and This indicates the initialization of node indices in the knowledge graph. Represents a node In time The memory strength value; Represents a node The forgetting rate parameter is used to control the node. The rate of memory decay; This represents the diffusion coefficient, used to control the rate at which memory strength propagates through edges in the initial knowledge graph; For nodes The set of neighboring nodes, representing the node The set of directly connected nodes; Indicates from node To the node The edge weight parameter is used to adjust the memory strength from the node. To the node The extent of the impact.

[0064] Specifically, in simulating the memory intensity changes of each node and obtaining the memory intensity distribution based on the initialized knowledge graph, it is first necessary to establish a partial differential equation model to accurately characterize the dynamic changes in memory intensity. This model must simultaneously reflect the natural decay of memory over time and the memory diffusion effect generated by the associations between nodes. Specifically, based on the forgetting rate parameter and edge weight parameter of each node in the initialized knowledge graph, a partial differential equation model of memory intensity changes needs to be constructed. In its expression, i and j represent the node indices in the initialized knowledge graph. This is used to describe the memory strength value of node i at time t, which directly reflects the learner's mastery of the knowledge point corresponding to that node. The forgetting rate parameter for node i determines how quickly the memory strength of node i decays over time; the larger the value, the faster the memory decays. As a diffusion coefficient, its main function is to control the rate at which memory intensity propagates through the edges between nodes in the knowledge graph. The higher the value, the more significant the mutual influence of memory strength between adjacent nodes; The set of neighboring nodes of node i is the collective term for all nodes that are directly connected to node i through an edge. These are the edge weight parameters from node j to node i, used to adjust the degree to which the memory strength of node j affects the memory strength of node i. The larger the value, the stronger the influence of node j on the memory strength of node i.

[0065] S12. Based on the partial differential equation model, after setting the discrete time step, numerical integration is performed using the Euler method to obtain the memory strength value of each node in the next time step; the calculation formula for numerical integration is:

[0066]

[0067] in, This represents the discrete time step, the time interval used in numerical simulations.

[0068] Specifically, after constructing the partial differential equation model, the model is solved numerically to obtain the memory strength values ​​of each node at different time points. Here, the Euler method is used for numerical integration, and the discrete time step is set before the calculation. The time step represents the time interval between two adjacent calculations in the numerical simulation process, and is determined according to the time granularity requirements of the actual learning scenario.

[0069] This formula allows us to determine the memory strength of node i at the current time t. The rate of change of memory intensity calculated from the previously constructed partial differential equation model is then multiplied by the discrete time step. To obtain the time of node i Memory strength value at time This enables discrete updates of memory strength from the current moment to the next moment.

[0070] S13. Based on the initial memory intensity distribution and numerical integration calculation, iteratively execute the simulation process to obtain time series memory intensity change data; based on the memory intensity change data, output the memory intensity distribution.

[0071] Specifically, after obtaining the numerical integration calculation method, an iterative simulation process is initiated based on the initial memory intensity distribution. The initial memory intensity distribution originates from the initial memory intensity parameters of each node in the knowledge graph. During the iterative simulation, starting from the initial time step, the simulation proceeds according to a set discrete time step. Repeat the above numerical integration calculation steps to obtain the value of each node in turn. , , The memory intensity values ​​at various time points are used to form a complete time-series data on memory intensity changes. The iteration stops when the simulation reaches the preset total duration, or when the magnitude of memory intensity changes at each node is less than a preset threshold (i.e., the memory state tends to stabilize). Finally, based on the generated time-series memory intensity change data, a memory intensity distribution containing information on the memory intensity of each node at different time points is obtained. This distribution can intuitively reflect the evolution of memory intensity for each knowledge point over time, providing data support for subsequent review interventions.

[0072] In one optional embodiment, based on the memory strength distribution and the initialized knowledge graph, the memory strength of the reviewed node and its neighboring nodes is updated through a review intervention model to obtain the updated memory strength distribution, including the following steps:

[0073] S21, Nodes triggered by review events And directly review the gain parameters Update nodes The memory strength value is used to obtain the node. The updated memory strength value; node The formula for updating memory strength values ​​is:

[0074]

[0075] in, Indicates the index of the node being reviewed. Indicates the node before review The memory strength value, Indicates the node after review The memory strength value; This indicates the direct review gain parameter, used to quantize review nodes. The direct increase in its own memory strength value.

[0076] Specifically, in the process of updating the memory intensity of the reviewed node and its neighboring nodes to obtain the updated memory intensity distribution based on the memory intensity distribution and the initialized knowledge graph through the review intervention model, the memory intensity of the reviewed node itself is updated first. Specifically, when a review event is detected, the reviewed node that triggered the review event (denoted as node) is identified. ), and combine this with a direct review of gain parameters For nodes The memory strength value is updated, with the core being the quantification of the positive effect of direct review on the memory strength of node k. In the formula for this update process, The index represents the node being reviewed, used to uniquely identify the node corresponding to the knowledge point being reviewed this time; This indicates that before the review operation is executed, the node... At the current time The memory strength value is directly taken from the previously obtained memory strength distribution and reflects the learner's mastery of the knowledge points corresponding to the node before review. This indicates that the node has completed the review process. The updated memory strength value is used to reflect the effect of review on improving the memory strength of key points; Its core function, in order to directly review the gain parameters, is to quantify the effect of a single review operation on the nodes. The direct increase in one's own memory strength. The value is determined by combining the review method (such as active testing, passive review, etc.) and the type of knowledge point. For example, for active testing-based review... The value should be higher than that of passive review classes to ensure that the increase matches the review effect, and to avoid the value exceeding a reasonable range after the initial update.

[0077] S22, Based on indirect review gain parameters Update nodes based on edge weight parameters. The memory strength values ​​of the neighboring nodes are used to obtain the updated memory strength values ​​of the neighboring nodes; where, for each neighboring node... Neighbor nodes The formula for updating memory strength values ​​is:

[0078]

[0079] in, Represents a node The neighbor node index, This represents the indirect review gain parameter, used to quantify review nodes. The indirect increase in the memory strength of neighboring nodes; Indicates from node To the node The edge weight parameter is used to adjust the indirect gain from the node. To the node The intensity of transmission; Represents a node The set of neighboring nodes.

[0080] Specifically, after directly updating the memory strength of the reviewed node, the indirect influence of the reviewed node on its neighboring nodes is further considered. That is, based on the indirect review gain parameter and the edge weight parameter in the initial knowledge graph, the node is updated. The memory strength values ​​of all neighboring nodes are considered. Since there are relationships between nodes in the knowledge graph, the increased memory strength of a reviewed node will radiate to its neighboring nodes through the edge connections. Therefore, it is necessary to analyze the memory strength of each node... The set of neighboring nodes (denoted as) The neighboring nodes of ) (denoted as node) ), calculate their updated memory strength values ​​respectively, and in the corresponding update formula, Representative node The index of a certain neighboring node. Represents all nodes The set of nodes directly connected by edges, i.e., nodes The entirety of the neighboring nodes; To indirectly review the gain parameters, its function is to quantify the factors at the review nodes. The indirect increase in the memory strength of its neighboring nodes. The value is set based on the overall tightness of the knowledge connections; the tighter the connections within a knowledge system, the better. The value can be appropriately increased to reflect a more significant indirect radiation effect; Indicates from node To the node The edge weight parameter, which is directly taken from the initialized knowledge graph, has the core function of adjusting the indirect review gain from the nodes. To the node The intensity of transmission, The larger the value, the stronger the node. With nodes The closer the relationship between nodes, the more important it is to maintain the connection between them. From node The greater the increase in indirect memory strength gained from reviewing, the better; To review previous neighbor nodes The memory strength value (taken from the memory strength distribution). This represents the updated memory strength value of neighbor node i, thus quantifying the indirect review effect.

[0081] S23, By analyzing the nodes The updated memory strength values ​​of the node and its neighboring nodes are pruned to ensure that the memory strength values ​​do not exceed a preset maximum value, resulting in pruned memory strength data. Based on the pruned memory strength data, an updated memory strength distribution is generated.

[0082] Specifically, upon completion of the review milestones After the initial update of the memory strength of a node and all its neighboring nodes, the updated memory strength values ​​are pruned. This is because memory strength values ​​have a reasonable range in a practical physical sense (which can be set to [0,1]), and values ​​exceeding this range would violate the core logic that "memory strength reflects the degree of mastery of knowledge points" (e.g., values ​​greater than 1 have no actual cognitive meaning). The core rule of the pruning process is: if a node (including node...)... The updated memory strength value (and its neighboring nodes) If the memory strength value of a node exceeds a preset maximum value (which can be set to 1, representing complete mastery of the knowledge point), the memory strength value of that node is forcibly adjusted to the preset maximum value. If the memory strength value does not exceed the preset maximum value, the initially updated value is retained, thus obtaining the pruned memory strength data. Finally, based on the pruned memory strength data of all nodes (including those that did not participate in this review update, whose memory strength values ​​remain unchanged and are directly included in the pruned data), the data is structured according to node index and time dimension to generate an updated memory strength distribution covering the entire knowledge graph. This distribution can accurately reflect the memory status of each knowledge point after the review intervention, providing the latest data support for the planning of the optimal review path.

[0083] In one optional embodiment, based on the updated memory strength distribution, an optimal review path is generated using a path planning algorithm, and a review sequence is obtained based on the optimal review path, including the following steps:

[0084] S31. Construct a reinforcement learning environment based on the updated memory strength distribution and current time information; wherein, the state of the reinforcement learning environment... for The memory strength vector of all nodes at any given time, and the action. for Choose when to review and reward yourself. for The degree to which overall memory stability is improved at all times.

[0085] Specifically, in the process of generating the optimal review path and review sequence based on the updated memory strength distribution, a reinforcement learning environment adapted to the review decision-making scenario is first constructed to provide a basic framework for subsequent policy learning. Specifically, when constructing the reinforcement learning environment, the three core elements of the environment are clearly defined: state, action, and reward. Among them, state... Defined as the memory strength vector of all nodes at time t, the dimension of this vector is consistent with the total number of nodes in the initial knowledge graph. Each element in the vector corresponds to the memory strength value of a node after the update at time t, i.e. (where n is the total number of nodes), this vector can fully characterize the learner's memory state of the entire knowledge system at time t; action Defined as the node selected for review at time t, its action space encompasses the entire set of nodes in the knowledge graph; that is, each action corresponds to selecting a specific node for review, ensuring that the decision covers all potentially reviewable knowledge points; reward. Defined as the degree of improvement in overall memory stability at time t, its quantification combines the changes in node memory intensity with the characteristics of knowledge network association. For example, it can be determined by calculating the difference between the weighted sum of the memory intensity of all nodes after review and before review. The weights refer to the sum of the edge weights of the nodes to reflect the greater influence of core nodes on overall memory stability, ensuring that the reward can truly reflect the optimization effect of the review action on the memory state of the knowledge system.

[0086] S32. Based on the reinforcement learning environment, the DQN algorithm is used to learn the optimal policy, resulting in a trained DQN model; where the input of the DQN model is the state. The output is the Q-value for each action.

[0087] Specifically, after constructing the reinforcement learning environment, the optimal review strategy is learned using the DQN (Deep Q-Network) algorithm. The core logic of the DQN algorithm is to use a deep neural network to approximate the action-value function (Q-function) to address the problem of large state spaces in reinforcement learning (e.g., high dimensionality of state vectors when there are many nodes). In this step, the input to the DQN model is the state of the reinforcement learning environment. The input vector is the memory strength vector at time t. After processing by the fully connected layers and activation layers (such as the ReLU activation function) of the neural network, the output is the Q-value corresponding to each possible action (i.e., each node). The Q-value represents the expected cumulative reward that can be obtained in the future after choosing the action in the current state. The higher the Q-value, the more beneficial the action is to improving the stability of long-term memory. During the model training process, the key mechanisms of DQN are introduced to ensure training stability: one is the experience playback mechanism, which replays the results of each interaction (…). The training process employs several mechanisms: First, empirical data is stored in an experience pool. During training, batches of data are randomly sampled to update the network, preventing correlation between samples from affecting training performance. Second, a target network mechanism is used to construct a target network with the same structure as the main network to calculate the target Q-value. The parameters of the target network are periodically copied and updated from the main network, reducing fluctuations in the target value during training. Through iterative training, the error between the Q-value output by the main network and the target Q-value is continuously minimized (e.g., using a mean squared error loss function), ultimately resulting in a well-trained DQN model that stably outputs action Q-values.

[0088] S33. Based on the trained DQN model and the current state, a review sequence is generated by selecting the action with the highest Q value as the review node for the next time step.

[0089] Specifically, a greedy decision-making process is used with the trained model, that is, the optimal action is selected based on the current state, gradually generating an ordered sequence of review nodes. In practice, the current state is first obtained. This state is the updated memory strength vector of all nodes at the current time point, which is related to the state defined in S31. Consistent structure; Input the trained DQN model, and the model outputs the Q-value corresponding to each node (action). Select the action with the highest Q-value from all actions; the node corresponding to this action is the next node to be reviewed and is added to the review sequence. Subsequently, based on the review effect of this node (as shown in the memory strength update logic in steps S21-S23), update the current state to obtain a new state. Repeat the above process—input a new state, obtain the Q value, select the optimal action, and update the state—until the number of generated review nodes reaches the preset review task requirement, or the overall memory strength of the current knowledge system reaches the preset stability threshold. The resulting ordered list of nodes is the required review sequence. This sequence ensures that each review action is the optimal choice for improving the overall memory stability in the current state, thus achieving global optimization of the review path.

[0090] In one alternative embodiment, the reward The expression is:

[0091]

[0092] in, Indicates the node index. This represents the reward value at time t. Represents a node Importance weights; Indicates the node after review The memory strength value, Indicates the node before review Memory strength value; importance weight The calculation steps are as follows: Based on the topological structure of the basic knowledge graph, the centrality score of each node is calculated using the PageRank algorithm; based on the centrality score, combined with the basic importance weights of the knowledge points specified by experts, the importance weight of each node is calculated using a linear combination formula; the linear combination formula is... ,in For balance coefficient, This represents the centrality score of node i. This represents the basic importance weight of the knowledge points at node i.

[0093] Specifically, in the process of generating the optimal review path based on the reinforcement learning environment, rewards... As a core indicator for measuring the effect of each review action on improving overall memory stability, it is used to quantify the relationship between changes in memory intensity and the importance of each node before and after review. (Reward) In the expression, This represents the node index in the knowledge graph, corresponding to a specific knowledge point; Indicates time The higher the reward value obtained after performing a review action, the more significant the effect of the review action on the overall memory and stability of the knowledge system. For nodes Importance weights are used to distinguish the core importance of different knowledge points within the knowledge system; nodes with higher importance... The larger the value, the greater the impact of changes in memory strength on reward. The greater the contribution, the greater the contribution; It is a review intervention node The memory strength values ​​were taken from the updated memory strength data of the review intervention model; It is a review of the pre-intervention node The memory strength value is taken from the memory strength distribution before the update. This expression achieves a global evaluation of the review action's effectiveness by summing the "importance weight × change in memory strength" of all nodes. This avoids the limitations of focusing only on the memory change of a single node, and also... It highlights the key role of core knowledge points, ensuring that reward calculations are consistent with the overall stability requirements of the knowledge system.

[0094] node Importance weight Combining the topology of the knowledge graph with expert experience, the calculation process consists of two steps: The first step is to calculate the node centrality score based on the topology of the basic knowledge graph. The PageRank algorithm is used here. Originally used to evaluate the importance of web pages on the internet, its application to knowledge graphs allows for the quantification of a node's centrality within the knowledge network through the connections between nodes. Specifically, an adjacency matrix of the knowledge graph is first constructed, with matrix elements corresponding to edge weight parameters between nodes. Then, the iterative formula of the PageRank algorithm is used... (in For nodes The degree of exit, The damping coefficient is... (The total number of nodes) is iterated repeatedly until the total number of nodes is reached. (i.e., node) The centrality score converges and becomes stable. The higher the value, the more central the node is in the knowledge graph, the closer its connection with other knowledge points, and the greater its impact on the overall coherence of the knowledge system.

[0095] The second step is to combine the centrality score with the basic importance weights specified by the expert using a linear combination formula to obtain the final importance weights. The linear combination formula is defined as follows:

[0096]

[0097] in, This is a balancing coefficient, with values ​​ranging from [0,1], used to adjust the centrality score. With basic importance weight exist The proportion of contribution in; The nodes obtained in the first step The centrality score has been normalized and its value ranges from [0,1]. For nodes The importance weights of the knowledge points are manually assigned by domain education experts based on factors such as curriculum standards, cognitive level difficulty, and subsequent learning dependence, and are also normalized to the [0,1] interval. Through this linear combination, It takes into account both the topological structure of the knowledge graph itself (reflecting the objective importance of the connections between knowledge points) and the educational experience of experts (reflecting the subjective importance of teaching knowledge points), ensuring that the calculation of importance weights is both objective and practical, thus providing a basis for rewards. It provides reliable support for accurate calculations.

[0098] The aforementioned adaptive gamified learning system design method constructs a knowledge graph containing knowledge point nodes and associated edges, initializes memory parameters, and uses a memory intensity dynamics model to simulate the decay and mutual influence of knowledge point memory intensity, thereby obtaining a dynamic memory intensity distribution. When a review event is detected, the memory intensity of the reviewed node and associated nodes is updated through a review intervention model, resulting in an updated memory state. Based on this, a path planning algorithm is used to generate an optimal review path that maximizes overall memory stability with minimal review cost, and this path is transformed into a gamified task. Ultimately, this achieves adaptive learning path planning through structured knowledge modeling and dynamic memory optimization, effectively improving learning efficiency, reducing cognitive load, and enhancing learning motivation and participation through gamified elements.

[0099] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0100] Based on the same inventive concept, this application also provides an apparatus for implementing the adaptive gamified learning system design method described above. The solution provided by this apparatus is similar to the implementation scheme described in the above method; therefore, the specific limitations in one or more embodiments of the adaptive gamified learning system design apparatus provided below can be found in the limitations of the adaptive gamified learning system design method described above, and will not be repeated here.

[0101] In one exemplary embodiment, such as Figure 3 As shown, an adaptive gamified learning system design device 30 is provided to implement the methods in the above-described method embodiments. The device includes:

[0102] The knowledge graph initialization module 31 is used to construct a basic knowledge graph containing nodes and edges based on the attribute information and associations of knowledge points, and to initialize the memory strength parameter and forgetting rate parameter for each node in the basic knowledge graph, and to initialize the edge weight parameter for each edge, thus obtaining the initialized knowledge graph.

[0103] The memory dynamic simulation module 32 is used to simulate the decay of the memory intensity of each node over time and the process of mutual influence through edges based on the initialized knowledge graph and through the memory intensity dynamic model, so as to obtain the memory intensity distribution.

[0104] The review intervention module 33 is used to update the memory intensity of the reviewed node and its neighboring nodes through the review intervention model when a review event is detected, based on the memory intensity distribution and the initialized knowledge graph, to obtain the updated memory intensity distribution.

[0105] The path planning module 34 is used to generate the optimal review path based on the updated memory strength distribution using a path planning algorithm, and to obtain the review sequence based on the optimal review path.

[0106] Gamification Transformation Module 35 is used to transform review sequences into gamified tasks based on the optimal review path and basic knowledge graph.

[0107] Embodiments of this application also provide a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the aforementioned method embodiments.

[0108] Embodiments of this application also provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in the above-described method embodiments.

[0109] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The components described as separate parts may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this disclosure according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0110] The above-described embodiments are merely illustrative of several implementation methods of the embodiments of this application, and their descriptions are relatively specific and detailed. However, they should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the embodiments of this application, and these modifications and improvements all fall within the protection scope of the embodiments of this application.

Claims

1. A design method for an adaptive gamified learning system, characterized in that, The method includes: S1. Based on the attribute information and association of knowledge points, construct a basic knowledge graph containing nodes and edges, and initialize memory strength parameters and forgetting rate parameters for each node in the basic knowledge graph, and initialize edge weight parameters for each edge to obtain the initialized knowledge graph; S2. Based on the initial knowledge graph, the memory intensity of each node decays over time and interacts with each other through the edge by simulating the memory intensity dynamics model to obtain the memory intensity distribution. S3. When a review event is detected, based on the memory intensity distribution and the initial knowledge graph, the memory intensity of the reviewed node and its neighboring nodes is updated through the review intervention model to obtain the updated memory intensity distribution. S4. Based on the updated memory strength distribution, generate the optimal review path using a path planning algorithm, and obtain the review sequence according to the optimal review path; S5. Based on the optimal review path and the basic knowledge graph, the review sequence is transformed into a gamified task.

2. The method according to claim 1, characterized in that, Based on the initialized knowledge graph, the memory intensity distribution is obtained by simulating the decay of memory intensity of each node over time and the process of mutual influence through edges using a memory intensity dynamics model, including: S11. Based on the forgetting rate parameter and the edge weight parameter of each node in the initialized knowledge graph, establish a partial differential equation for the change in memory strength, and obtain a partial differential equation model; the expression of the partial differential equation model is: in, and This represents the node index in the initialized knowledge graph. Represents a node In time The memory strength value; Represents a node The forgetting rate parameter is used to control the node. The rate of memory decay; This represents the diffusion coefficient, used to control the rate at which memory strength propagates through edges in the initialized knowledge graph; For nodes The set of neighboring nodes, representing the node The set of directly connected nodes; Indicates from node To the node The edge weight parameter is used to adjust the memory strength from the node. To the node The degree of impact; S12. Based on the partial differential equation model, after setting the discrete time step, numerical integration is performed using the Euler method to obtain the memory strength value of each node in the next time step; the calculation formula for the numerical integration is: in, This represents the discrete time step, the time interval used in numerical simulations. S13. Based on the initial memory intensity distribution and the numerical integration calculation, the simulation process is iteratively executed to obtain time series memory intensity change data; based on the memory intensity change data, the memory intensity distribution is output.

3. The method according to claim 2, characterized in that, The process of updating the memory intensity of the reviewed node and its neighboring nodes through a review intervention model based on the memory intensity distribution and the initialized knowledge graph, to obtain the updated memory intensity distribution, includes: S21, Nodes triggered by the aforementioned review event And directly review the gain parameters Update nodes The memory strength value is used to obtain the node. The updated memory strength value; the node The formula for updating memory strength values ​​is: in, Indicates the index of the node being reviewed. Indicates the node before review The memory strength value, Indicates the node after review The memory strength value; This indicates the direct review gain parameter, used to quantize review nodes. The direct increase in its own memory strength value; S22, Based on indirect review gain parameters Update the node with the edge weight parameters. The memory strength values ​​of the neighboring nodes are used to obtain the updated memory strength values ​​of the neighboring nodes; where, for each neighboring node... The neighbor node The formula for updating memory strength values ​​is: in, Represents a node The neighbor node index, The indirect review gain parameter is used to quantify review nodes. The indirect increase in the memory strength of neighboring nodes; Indicates from node To the node The edge weight parameter is used to adjust the indirect gain from the node. To the node The intensity of transmission; Represents a node The set of neighboring nodes; S23, By analyzing the nodes The updated memory strength values ​​of the node and its neighboring nodes are pruned to ensure that the memory strength values ​​do not exceed a preset maximum value, resulting in pruned memory strength data; based on the pruned memory strength data, the updated memory strength distribution is generated.

4. The method according to any one of claims 2 to 3, characterized in that, The step of generating an optimal review path based on the updated memory strength distribution using a path planning algorithm, and obtaining a review sequence based on the optimal review path, includes: S31. Based on the updated memory strength distribution and current time information, construct a reinforcement learning environment; wherein, the state of the reinforcement learning environment... for The memory strength vector of all nodes at any given time, and the action. for Choose when to review and reward yourself. for The degree of improvement in the overall stability of memory at all times; S32. Based on the reinforcement learning environment, the DQN algorithm is used to learn the optimal policy, resulting in a trained DQN model; wherein, the input of the DQN model is the state. The output is the Q-value for each action; S33. Based on the trained DQN model and the current state, the review sequence is generated by selecting the action with the highest Q value as the review node for the next time step.

5. The method according to claim 4, characterized in that, The reward The expression is: in, Indicates the node index. This represents the reward value at time t. Represents a node Importance weights; Indicates the node after review The memory strength value, Indicates the node before review The memory strength value; The importance weight The calculation steps are as follows: Based on the topological structure of the basic knowledge graph, the centrality score of each node is calculated using the PageRank algorithm; based on the centrality score, combined with the basic importance weights of knowledge points specified by experts, the importance weights of each node are calculated using a linear combination formula; the linear combination formula is... ,in For balance coefficient, This represents the centrality score of node i. This represents the basic importance weight of the knowledge points at node i.

6. An adaptive gamified learning system design apparatus for implementing the method according to any one of claims 1 to 5, characterized in that, The device includes: The knowledge graph initialization module is used to construct a basic knowledge graph containing nodes and edges based on the attribute information and associations of knowledge points, and to initialize memory strength parameters and forgetting rate parameters for each node in the basic knowledge graph, and to initialize edge weight parameters for each edge, thereby obtaining the initialized knowledge graph. The memory dynamic simulation module is used to simulate the decay of the memory intensity of each node over time and the process of mutual influence through edges based on the initialized knowledge graph, thereby obtaining the memory intensity distribution. The review intervention module is used to update the memory intensity of the reviewed node and its neighboring nodes through the review intervention model when a review event is detected, based on the memory intensity distribution and the initial knowledge graph, to obtain the updated memory intensity distribution; The path planning module is used to generate an optimal review path based on the updated memory strength distribution using a path planning algorithm, and to obtain a review sequence based on the optimal review path. The gamification conversion module is used to convert the review sequence into gamified tasks based on the optimal review path and the basic knowledge graph.

7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the method of any one of claims 1 to 5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 5.

Images