Military college online learning management platform based on big data analysis

Abstract

The invention relates to the technical field of big data, and particularly discloses a military college online learning management platform based on big data analysis, which constructs a knowledge graph according to acquired military learning data and trains a personalized learning recommendation model by using the knowledge graph. And the personalized learning recommendation model is distributed to the learning equipment of each student, and the learning equipment can collect and analyze the learning behavior data of the student, so that deep analysis and personalized guidance of the learning condition of the student are realized. According to the invention, based on the personalized learning data and model training update data of the trainees, in combination with a big data analysis method, highly customized learning content and recommendation can be provided for each trainee. The recommendation strategy can be automatically adjusted according to the learning progress and performance of the student, the learning path can be optimized in real time, and the learning experience and efficiency of the student are improved.

Description

Technical Field

[0001] The present invention relates to the field of big data technology, and in particular to an online learning management platform for military academies based on big data analysis. Background Art

[0002] At present, the online learning platforms of current military academies mainly focus on the provision of courses and basic management of students, but lack in-depth analysis and intelligent management of learning data. The existing system is difficult to meet the needs of personalized teaching, and cannot effectively analyze and feedback students' learning situation in real time, thus affecting the teaching effect and students' learning efficiency. In addition, the course content of military academies is relatively complex, and teaching and training are usually carried out in parallel. The existing platform is often unable to flexibly adapt to the needs of each stage, and cannot conduct comprehensive data analysis and feedback, resulting in inaccurate monitoring of students' learning progress and quality, and insufficient support for personalized education.

[0003] Therefore, it is necessary to provide an online learning management platform for military academies based on big data analysis. Summary of the invention

[0004] The purpose of the present invention is to provide an online learning management platform for military academies based on big data analysis to solve the following technical problems: Based on the personalized learning data of students and the model training update data, combined with the big data analysis method, highly customized learning content and recommendations can be provided to each student. Not only can the recommendation strategy be automatically adjusted according to the learning progress and performance of the students, but the learning path can also be optimized in real time to improve the learning experience and efficiency of the students.

[0005] The purpose of the present invention can be achieved through the following technical solutions: A military academy online learning management platform based on big data analysis, the management platform comprising: A first data acquisition module is used to acquire military learning material data; the military learning material data includes teaching materials, question bank materials and video materials; A knowledge graph construction module, used for extracting knowledge points based on the military learning material data, and constructing a learning knowledge graph based on the knowledge points; A model training module, used to perform model training based on the learning knowledge graph to obtain a personalized learning recommendation model; A first distribution module, used to distribute the personalized learning recommendation model and the learning knowledge graph to learning devices of multiple students; A second data acquisition module is used to acquire model training update data and knowledge mining data of each student; wherein the model training update data is obtained by the learning device of each student updating and training the personalized learning recommendation model based on the local learning record, and the knowledge mining data is the knowledge data mined by the learning device of each student based on its local learning record; An aggregation module, used for aggregating the model training update data of each student to obtain aggregated update data; A verification module, used for verifying the knowledge mining data based on the aggregated update data; A knowledge graph updating module, used to update the verified knowledge mining data to the learning knowledge graph to obtain a target knowledge graph; The second distribution module is used to distribute the target knowledge graph to the learning devices of the multiple students so that the multiple students can learn based on the target knowledge graph.

[0006] As a further solution of the present invention, the management platform further includes: A model updating module, used to update the personalized learning recommendation model based on the aggregated update data; The third distribution module is used to distribute the updated personalized learning recommendation model to the learning devices of the multiple students.

[0007] As a further solution of the present invention, the aggregating the model training update data of each student to obtain aggregated update data includes: Obtain the contribution corresponding to each student's learning device; Based on the contribution, determining a first weight of the model training update data of each student; Based on the first weight, the model training update data of each student is weighted to obtain the aggregated update data.

[0008] As a further solution of the present invention, the step of obtaining the contribution corresponding to each student's learning device includes: Get the data volume of each student’s learning device; Determine a first contribution of the learning device of each student based on the data amount of the learning device of each student; Perform conflict detection based on the knowledge mining data of each student to obtain a conflict detection result; Based on the conflict detection result, determining a second contribution of the learning device of each student; The first contribution and the second contribution are merged to obtain the contribution corresponding to the learning device of each student.

[0009] As a further solution of the present invention, the fusing of the first contribution and the second contribution includes: Based on the first contribution and the second contribution, fusion is performed by the following formula; ; Among them, C i is the contribution of the i-th student’s device, T i The type of knowledge point submitted by the i-th student device, D i is the data volume of the i-th student’s device, N i is the historical update times corresponding to the i-th student device, W 1,i For the first contribution, W 2,i For the second contribution, γ i is the conflict ratio obtained according to the conflict detection result.

[0010] As a further solution of the present invention, the conflict detection is performed based on the model training update data of each student to obtain the conflict detection result, including: Perform semantic conflict detection based on the knowledge mining data of each student to obtain conflicting knowledge data; Based on the conflict knowledge data, determining the conflict level; In response to the conflict level not satisfying a preset condition, determining a conflict detection result based on the trustworthiness of the learning devices corresponding to the respective students; In response to the conflict level satisfying a preset condition, determining a conflict detection result by the following formula; ; Where ∆K f is the final mining knowledge point in the conflict detection result, m i is the quality of the mined knowledge corresponding to the learning device of the i-th student, t i is the trust level of the i-th student’s learning device, ∆K i Knowledge mining data for the learning device of the i-th student.

[0011] As a further solution of the present invention, the aggregating the model training update data of each student to obtain aggregated update data includes: Based on the meta-learning model, the second weight of the model training update data of each student is obtained by the following formula; ; Among them, α i is the second weight of the i-th student, L train,i Update the loss function of the personalized learning recommendation model locally for the i-th student, θ fare the parameters of the meta-learning model.

[0012] Based on the second weight, the model training update data of each student is weighted to obtain the aggregated update data.

[0013] A method for online learning management in military academies based on big data analysis, the method comprising: Acquire military learning material data; the military learning material data includes teaching materials, question bank materials and video materials; Extracting knowledge points based on the military learning material data, and constructing a learning knowledge graph based on the knowledge points; Performing model training based on the learning knowledge graph to obtain a personalized learning recommendation model; Distributing the personalized learning recommendation model and the learning knowledge graph to learning devices of multiple students; Obtaining model training update data and knowledge mining data of each student; wherein the model training update data is obtained by updating and training the personalized learning recommendation model based on the local learning records of the learning devices of each student, and the knowledge mining data is the knowledge data mined by the learning devices of each student based on their local learning records; Aggregating the model training update data of each student to obtain aggregated update data; verifying the knowledge mining data based on the aggregated updated data; Updating the verified knowledge mining data to the learning knowledge graph to obtain a target knowledge graph; The target knowledge graph is distributed to the learning devices of the multiple students so that the multiple students can learn based on the target knowledge graph.

[0014] The beneficial effects of the present invention are as follows: (1) By updating data based on the students' personalized learning data and model training, combined with big data analysis methods, highly customized learning content and recommendations can be provided to each student. Not only can the recommendation strategy be automatically adjusted according to the students' learning progress and performance, but the learning path can also be optimized in real time to improve the students' learning experience and efficiency. (2) The personalized learning recommendation model for students is trained and updated using a meta-learning model, so that the system can dynamically adjust the recommendation strategy and learning model based on the characteristics and performance of each student. Through training based on the meta-learning model, the platform can more efficiently adapt the model when facing the learning data of different students, thereby enhancing the platform's ability to adapt to diverse learning needs. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] The present invention will be further described below in conjunction with the accompanying drawings.

[0016] Figure 1 It is a structural diagram of the online learning management platform for military academies based on big data analysis of the present invention; Figure 2 It is a flow chart of the online learning management platform for military academies based on big data analysis of the present invention. DETAILED DESCRIPTION

[0017] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0018] See also Figure 1 ,The present invention is an online learning management platform for military academies based on big data analysis, such as Figure 1 As shown, the online learning management platform 100 may include the following modules.

[0019] The first data acquisition module 101 is used to acquire military learning material data, including teaching materials, question bank materials and video materials.

[0020] The knowledge graph construction module 102 is used to extract knowledge points based on the military learning material data and construct a learning knowledge graph based on the knowledge points.

[0021] The model training module 103 is used to perform model training based on the learning knowledge graph to obtain a personalized learning recommendation model.

[0022] The first distribution module 104 is used to distribute the personalized learning recommendation model and the learning knowledge graph to the learning devices of multiple students.

[0023] The second data acquisition module 105 is used to acquire the model training update data and knowledge mining data of each student.

[0024] The model training update data is obtained by updating and training the personalized learning recommendation model through the learning device of each student based on the local learning record, and the knowledge mining data is the knowledge data mined by the learning device of each student based on the local learning record.

[0025] The aggregation module 106 is used to aggregate the model training update data of each student to obtain aggregated update data.

[0026] The verification module 107 is used to verify the knowledge mining data based on the aggregated update data.

[0027] The knowledge graph updating module 108 is used to update the verified knowledge mining data to the learning knowledge graph to obtain the target knowledge graph.

[0028] The second distribution module 109 is used to distribute the target knowledge graph to the learning devices of the multiple students so that the multiple students can learn based on the target knowledge graph.

[0029] In some embodiments, the online learning management platform 100 may further include a model updating module for updating the personalized learning recommendation model based on the aggregated update data; and a third distribution module for distributing the updated personalized learning recommendation model to the learning devices of the multiple students.

[0030] For more detailed description of each module of the online learning management platform 100, please refer to the following Figure 2 Description of the process section.

[0031] Figure 2 It is a flow chart of the online learning management platform for military academies based on big data analysis of the present invention. In some embodiments, Figure 2 The process 200 shown may be executed by a processor or the online learning management platform 100 .

[0032] Step 201, obtaining military learning material data.

[0033] Military learning material data refers to the various information materials required by students to study military courses, including teaching materials, question bank materials and video materials.

[0034] Teaching materials include written content of military courses, such as textbooks, reference books, handouts, etc., which can be provided by military academies or educational institutions.

[0035] The question bank data includes various test questions related to the textbook, such as multiple-choice questions, fill-in-the-blank questions, short-answer questions, etc. The question bank data can be downloaded from the teaching management system, question bank platform or examination system, or accessed through the API interface.

[0036] Video materials include lecture videos, demonstration videos, and experimental videos of military courses, which can provide intuitive teaching related to the content of textbooks and question banks. Video materials can be obtained by connecting to online learning platforms and video streaming services (such as YouTube, Bilibili, etc.) to capture video content, or directly from the internal teaching platforms of military academies.

[0037] Step 202, extracting knowledge points based on the military learning material data, and constructing a learning knowledge graph based on the knowledge points.

[0038] Knowledge points refer to important concepts, theories, formulas, skills and other knowledge units in textbooks, question banks or video materials.

[0039] Learning knowledge graph is to organize the extracted knowledge points and the relationships between them (such as subordination, causality, etc.) into data in the form of graph.

[0040] In some embodiments, natural language processing (NLP) technologies such as word segmentation, named entity recognition, topic modeling, etc. can be used to automatically extract relevant knowledge points from textbooks, question banks, and video materials. For video materials, speech recognition technology can be used to convert them into text, and then the key concepts and knowledge points therein can be identified through text analysis methods.

[0041] In some embodiments, the extracted knowledge points can be connected according to their mutual relationships (for example, "Concept A is associated with Concept B", "Concept A is a subset of Concept B", etc.) to form a graph structure and obtain a learning knowledge graph.

[0042] Step 203: Perform model training based on the learning knowledge graph to obtain a personalized learning recommendation model.

[0043] The personalized learning recommendation model is based on students' learning history, preferences, and learning knowledge graph. The model generated through data analysis and machine learning algorithms can recommend suitable learning content (textbooks, videos, question banks, etc.) to students.

[0044] In some embodiments, learning data of each student can be collected, including learning progress, knowledge points learned, test scores, learning time, etc., to build a student portrait. Knowledge points in the learning knowledge graph are input to establish the correlation between each knowledge point.

[0045] Use machine learning algorithms (such as collaborative filtering, deep learning, etc.) to train personalized recommendation models and learn how to recommend content based on students’ needs, levels, and learning habits. The training process includes iteratively optimizing model parameters and using students’ historical data and knowledge graph structure information to recommend appropriate learning resources.

[0046] Step 204: distribute the personalized learning recommendation model and the learning knowledge graph to the learning devices of multiple students.

[0047] Learning devices refer to the various devices used by students for online learning, such as PCs, tablets, mobile phones and other terminals.

[0048] Distribution refers to pushing personalized learning recommendation models and learning knowledge graphs to students' learning devices.

[0049] In some embodiments, the trained personalized learning recommendation model and learning knowledge graph can be distributed to the learner's learning device through the network. The distribution can be achieved through cloud services, and the learner's device receives the model and knowledge graph data through an API or SDK.

[0050] Step 205, obtaining the model training update data and knowledge mining data of each student.

[0051] Model training update data refers to the data used by students to update and train the personalized learning recommendation model through local learning records (such as completed learning tasks, questions done, videos watched, etc.). Model training update data includes the input data for model update training and the training results obtained from model update training, such as gradients.

[0052] In some embodiments, the model training update data is obtained by updating and training the personalized learning recommendation model based on local learning records by the learning devices of the respective students.

[0053] Knowledge mining data refers to data related to knowledge points mined from local learning records by students during the learning process, such as knowledge points that students have mastered or not mastered, and extended data that is not recorded in the learning knowledge graph.

[0054] In some embodiments, the knowledge mining data is knowledge data mined by the learning devices of the respective students based on their local learning records.

[0055] In some embodiments, learning behavior data may be collected from the learner's learning device, such as learning time, learning progress, learning frequency, etc. Specifically, the data may be obtained through channels such as sensors, application logs, and background data of the learning platform.

[0056] Based on the students' learning records, the personalized learning recommendation model is updated and trained through incremental learning (such as online learning algorithms), and the model parameters are adjusted to make the recommendations more accurate.

[0057] Step 206, aggregating the model training update data of each student to obtain aggregated update data.

[0058] Aggregate update data is a unified update result obtained by summarizing the model training update data of multiple students.

[0059] In some embodiments, model training update data from multiple students may be integrated, for example, by averaging, weighted averaging, etc. to obtain the final aggregated update data.

[0060] In some embodiments, the personalized learning model update information of all students can be merged to generate a new global model to ensure that the model reflects the learning behaviors and needs of different students.

[0061] For example, the personalized learning recommendation model can be updated based on the aggregated update data by using a model update module; and the updated personalized learning recommendation model can be distributed to the learning devices of the multiple students by using a third distribution module.

[0062] In some embodiments, aggregating the model training update data of each student to obtain aggregated update data may include: based on the meta-learning model, obtaining the second weight of the model training update data of each student by the following formula (1).

[0063] (1); Among them, α i is the second weight of the i-th student, L train,i Update the loss function of the personalized learning recommendation model locally for the i-th student, θ f are the parameters of the meta-learning model.

[0064] Based on the second weight, the model training update data of each student is weighted to obtain the aggregated update data.

[0065] A meta-learning model is a high-level learning model that optimizes the learning process by learning the rules of multiple tasks or multiple data sources. The goal of a meta-learning model is to enable the learning algorithm to transfer from one task to another more effectively, and can be used to solve the problem of obtaining the best update in multiple student data.

[0066] The second weight is a coefficient calculated based on the meta-learning model, which is used to characterize the weight obtained by the i-th student during the model training process. This weight reflects the importance of the student's model training data in the entire aggregation process. The second weight is used to weight the model training update data of each student, so as to make the data aggregation process more accurate and reasonable.

[0067] In some embodiments, aggregating the model training update data of each student to obtain aggregated update data may include the following operations.

[0068] S10, obtaining the contribution corresponding to each student's learning device.

[0069] Contribution refers to the degree to which the data provided by a student's learning device contributes to the update of the learning knowledge graph. Contribution can be evaluated based on factors such as the student's device usage, learning time, learning frequency, and interaction, to measure the device's participation in the data update of the knowledge graph.

[0070] In some embodiments, activity data of the student's learning device can be collected, for example, behavioral data related to learning can be collected from the student's device, including but not limited to the time the student uses the device (e.g., learning duration, learning frequency, etc.), the number of learning tasks completed by the student (e.g., completed courses, answers, exercises, etc.), the interaction between the student and the platform (e.g., questions, discussions, feedback, etc.), and the student's behavioral trajectory on the platform (e.g., progress in watching videos, browsing learning content, etc.). The learning data of the student's device is quantified, and the contribution of each device is calculated through a predetermined algorithm (e.g., weighted average, cumulative score, etc.).

[0071] In some embodiments, obtaining the contribution corresponding to each student's learning device may include the following operations.

[0072] S20, obtaining the data volume of each student's learning device.

[0073] Data volume refers to the amount of data generated or used by students through their learning devices during the learning process.

[0074] In some embodiments, the amount of data related to the learning activities can be collected from each student's learning device. The student's learning data can be quantified and expressed in units such as bytes (B), kilobytes (KB), and megabytes (MB), or the data can be converted into the actual amount of learning time or learning content occupied, which is convenient for subsequent calculations.

[0075] S21, determining a first contribution of each student's learning device based on the amount of data of the learning device of each student.

[0076] The first contribution refers to the measure of the contribution of the student learning device to the update of the learning knowledge graph during the data processing process, which is measured by the amount of data of the device. Devices with larger data volumes are considered to have made greater contributions.

[0077] According to the amount of data on each student's learning device, the first contribution of the learning device can be calculated using a ratio or formula. For example, in the direct proportion method, the first contribution is proportional to the amount of data on the device. For example, a device with a data volume of 10MB has a contribution of 10, and a device with a data volume of 20MB has a contribution of 20. In the normalization method, in order to facilitate comparison, the data volume of all devices can be standardized to ensure that the first contribution is within a unified quantitative range (for example, between 0 and 1). In the weighted method, different weights can be assigned to different types of data volume (such as network traffic, storage capacity, etc.) to comprehensively calculate the first contribution.

[0078] S22, performing conflict detection based on the knowledge mining data of each student to obtain a conflict detection result.

[0079] Conflict detection refers to the discovery of anomalies or inconsistencies in students' learning behavior or knowledge mining data during data analysis. Conflicts can be reflected in the following aspects: inconsistencies between student behavior data (for example, the same student has completely inconsistent learning behaviors in different time periods); problems with students' learning content repetition (such as repeatedly learning the same content but without actual progress); conflicts in students' knowledge mastery (such as inconsistencies in students' knowledge mastery).

[0080] In some embodiments, a conflict detection algorithm (eg, rule-based detection, machine learning model, etc.) may be used to analyze the knowledge mining data to generate a conflict detection result.

[0081] The conflict detection result can be a Boolean value (conflict exists or does not exist) or a detailed description of the conflict (such as the type and degree of the conflict).

[0082] In some embodiments, performing conflict detection based on the model training update data of each student to obtain the conflict detection result may include the following operations.

[0083] S30, performing semantic conflict detection based on the knowledge mining data of each student to obtain conflicting knowledge data.

[0084] Semantic conflict detection refers to the analysis of students' knowledge mining data to detect whether there are semantic conflicts or inconsistencies in the understanding of knowledge points among different students. The goal of semantic conflict detection is to identify and resolve differences or errors in students' understanding of the same knowledge point and avoid conflicts in learning content.

[0085] In some embodiments, natural language processing (NLP) algorithms, knowledge graphs, or other semantic analysis technologies can be used to analyze students' knowledge mining data and identify differences in students' understanding of the same knowledge point. For example, it can be detected whether different students have conflicts in understanding of the same knowledge point (such as different interpretations of a definition).

[0086] S31, judging the conflict level based on the conflict knowledge data.

[0087] The conflict level refers to the severity level of the conflict, which is used to evaluate the scope and extent of the impact of the conflict knowledge data. The conflict level can be divided into multiple levels, such as minor conflict, medium conflict, severe conflict, etc. The conflict level can be used to determine the priority of the conflict and whether further intervention is needed, as well as the way to deal with the conflict.

[0088] In some embodiments, the impact of the conflict on the learning content and learning objectives can be analyzed based on the conflict knowledge data. For example, whether there is a misleading conflict, and whether it affects the correct understanding of basic knowledge points. Based on the evaluation results, the conflict can be divided into different levels (such as minor, moderate, and severe). For example, a minor conflict may be just a difference in the expression of a knowledge point, while a severe conflict may involve the learner's misunderstanding of basic knowledge.

[0089] S32: In response to the conflict level not satisfying a preset condition, a conflict detection result is determined based on the trust levels of the learning devices corresponding to the respective students.

[0090] In some embodiments, the preset condition may be that the conflict level reaches a severe conflict.

[0091] For learning devices with higher trust (good historical performance and reliable data), the conflict detection results can be considered more reliable; while for devices with lower trust (poor device activity and unreliable historical data), it can be considered that the conflict data needs to be rechecked.

[0092] The conflict detection result can be adjusted according to the trust level of the student's device. If the trust level is high, the conflict detection result can be considered more reliable, otherwise, further verification is required.

[0093] S33, in response to the conflict level satisfying a preset condition, determining a conflict detection result by using the following formula (2).

[0094] (2); Where ∆K f is the final mining knowledge point in the conflict detection result, m i is the quality of the mined knowledge corresponding to the learning device of the i-th student, t iis the trust level of the i-th student’s learning device, ∆K i Knowledge mining data for the learning device of the i-th student.

[0095] S23: Determine the second contribution of each student's learning device based on the conflict detection result.

[0096] The second contribution refers to the contribution assigned to the student's learning device according to the result of the conflict detection. In some embodiments, it can be determined whether the contribution of the student's learning device is affected according to the conflict detection result. For example, if an abnormality or conflict is detected in the learning data of a certain student, a lower contribution of the device may need to be assigned.

[0097] S24, integrating the first contribution and the second contribution to obtain the contribution corresponding to the learning device of each student.

[0098] Fusion refers to combining the first contribution and the second contribution to form a final contribution.

[0099] In some embodiments, a suitable fusion method can be selected according to the properties of the first contribution and the second contribution. For example, the weighted average method weights the first contribution and the second contribution according to a certain weight to obtain the final contribution. For example, the first contribution accounts for 60% and the second contribution accounts for 40%, and the final contribution is calculated according to the weights. The comprehensive scoring method combines the contributions of the two according to a certain comprehensive scoring standard to form a unified contribution. The maximum and minimum method selects a more reliable contribution as the final contribution according to the severity of the conflict detection result.

[0100] In some embodiments, the fusing the first contribution and the second contribution may include: fusing the first contribution and the second contribution according to the following formula (3): (3); Among them, C i is the contribution of the i-th student’s device, T i The type of knowledge point submitted by the i-th student device, D i is the data volume of the i-th student’s device, N i is the historical update times corresponding to the i-th student device, W 1,i For the first contribution, W 2,i For the second contribution, γ i is the conflict ratio obtained according to the conflict detection result.

[0101] S11, based on the contribution, determining the first weight of the model training update data of each student.

[0102] The first weight refers to the weight assigned to each student's learning data based on the contribution of the student's device. Students with higher contributions will receive higher weights, which means that the student's learning data will have a greater influence when the model training is updated.

[0103] In some embodiments, the first weight can be determined by various methods. For example, the linear allocation method directly calculates the weight according to the contribution. For example, if the contribution of a student's device is higher, the weight of the device is the percentage of its contribution. The normalization method standardizes the contribution of all devices (such as minimum-maximum normalization) to ensure that the weights of all students are within a uniform range. The weighted average method uses the weighted average method to weight the student's learning data according to the contribution of the device. For example, the proportion of a certain device's contribution to the total contribution of the student will give the student's data a corresponding weight.

[0104] S12: Based on the first weight, weight the model training update data of each student to obtain the aggregated update data.

[0105] In some embodiments, the learning data of each student may be weighted according to the first weight of the data of each student. After the weighted data of all students are aggregated, the aggregated update data may be obtained.

[0106] Step 207: verify the knowledge mining data based on the aggregated updated data.

[0107] Verification refers to the use of aggregated updated data to test the accuracy and effectiveness of knowledge mining results.

[0108] In some embodiments, the information in the aggregated update data can be used to verify the knowledge points mined by the students during the learning process to determine whether they are consistent with the actual learning results. For example, techniques such as cross-validation and A / B testing can be used to ensure that the mined knowledge points are accurate and reliable.

[0109] Step 208, update the verified knowledge mining data to the learning knowledge graph to obtain the target knowledge graph.

[0110] The target knowledge graph refers to the updated and optimized learning knowledge graph after the knowledge mining data has been verified.

[0111] In some embodiments, the knowledge graph can be adjusted and expanded based on the verified knowledge mining data, and new knowledge nodes and relationships can be added. A graph database (such as Neo4j) can be used to integrate new knowledge points with the existing knowledge graph to obtain a target knowledge graph.

[0112] Step 209: distribute the target knowledge graph to the learning devices of the multiple students so that the multiple students can learn based on the target knowledge graph.

[0113] Distribution refers to pushing the updated target knowledge graph to the learner's learning device. For example, the target knowledge graph can be pushed from a central server or cloud to each learner's learning device through the network or other data transmission methods, so that the learner can use the latest data for personalized learning.

[0114] The working principle of the present invention: By updating data based on the students' personalized learning data and model training, combined with big data analysis methods, highly customized learning content and recommendations can be provided to each student. Not only can the recommendation strategy be automatically adjusted according to the students' learning progress and performance, but the learning path can also be optimized in real time to improve the students' learning experience and efficiency. The personalized learning recommendation model of the students is trained and updated using a meta-learning model, so that the system can dynamically adjust the recommendation strategy and learning model according to the characteristics and performance of each student. Through training based on the meta-learning model, the platform can more efficiently adapt the model when facing the learning data of different students, thereby enhancing the platform's ability to adapt to diverse learning needs.

[0115] The above is a detailed description of an embodiment of the present invention, but the content is only a preferred embodiment of the present invention and cannot be considered to limit the scope of implementation of the present invention. All equivalent changes and improvements made within the scope of the present invention should still fall within the scope of the patent coverage of the present invention.

Claims

1. A military academy online learning management platform based on big data analysis, characterized in that: The management platform includes: The first data acquisition module is used to acquire military learning material data; the military learning material data includes teaching materials, question bank materials and video materials; A knowledge graph construction module, used to extract knowledge points based on the military learning material data, and construct a learning knowledge graph based on the knowledge points; A model training module, used to perform model training based on the learning knowledge graph to obtain a personalized learning recommendation model; A first distribution module, used to distribute the personalized learning recommendation model and the learning knowledge graph to learning devices of multiple students; A second data acquisition module is used to acquire model training update data and knowledge mining data of each student; wherein the model training update data is obtained by the learning device of each student updating and training the personalized learning recommendation model based on the local learning record, and the knowledge mining data is the knowledge data mined by the learning device of each student based on its local learning record; An aggregation module, used for aggregating the model training update data of each student to obtain aggregated update data; A verification module, used for verifying the knowledge mining data based on the aggregated update data; A knowledge graph updating module, used to update the verified knowledge mining data to the learning knowledge graph to obtain a target knowledge graph; The second distribution module is used to distribute the target knowledge graph to the learning devices of the multiple students so that the multiple students can learn based on the target knowledge graph.

2. According to the military academy online learning management platform based on big data analysis according to claim 1, it is characterized in that: The management platform also includes: A model updating module, used to update the personalized learning recommendation model based on the aggregated update data; The third distribution module is used to distribute the updated personalized learning recommendation model to the learning devices of the multiple students.

3. According to the military academy online learning management platform based on big data analysis as described in claim 1, it is characterized in that: The aggregating the model training update data of each student to obtain aggregated update data includes: Obtain the contribution corresponding to each student's learning device; Based on the contribution, determining a first weight of the model training update data of each student; Based on the first weight, the model training update data of each student is weighted to obtain the aggregated update data.

4. According to claim 3, a military academy online learning management platform based on big data analysis is characterized in that: The obtaining of the contribution corresponding to each student's learning device includes: Get the data volume of each student’s learning device; Determine a first contribution of the learning device of each student based on the data amount of the learning device of each student; Perform conflict detection based on the knowledge mining data of each student to obtain a conflict detection result; Based on the conflict detection result, determining a second contribution of the learning device of each student; The first contribution and the second contribution are merged to obtain the contribution corresponding to the learning device of each student.

5. The online learning management platform for military academies based on big data analysis according to claim 4 is characterized in that: The fusing the first contribution and the second contribution comprises: Based on the first contribution and the second contribution, fusion is performed by the following formula; ； Among them, C i is the contribution of the i-th student’s device, T i The type of knowledge point submitted by the i-th student device, D i is the data volume of the i-th student’s device, N i is the historical update times corresponding to the i-th student device, W 1,i For the first contribution, W 2,i For the second contribution, γ i is the conflict ratio obtained according to the conflict detection result.

6. The online learning management platform for military academies based on big data analysis according to claim 4 is characterized in that: The conflict detection is performed based on the model training update data of each student to obtain the conflict detection result, including: Perform semantic conflict detection based on the knowledge mining data of each student to obtain conflicting knowledge data; Based on the conflict knowledge data, determining the conflict level; In response to the conflict level not satisfying a preset condition, determining a conflict detection result based on the trustworthiness of the learning devices corresponding to the respective students; In response to the conflict level satisfying a preset condition, determining a conflict detection result by the following formula; ； Where ∆K f is the final mining knowledge point in the conflict detection result, m i is the quality of the mined knowledge corresponding to the learning device of the i-th student, t i is the trust level of the i-th student’s learning device, ∆K i Knowledge mining data for the learning device of the i-th student.

7. The online learning management platform for military academies based on big data analysis according to claim 1 is characterized in that: The aggregating the model training update data of each student to obtain aggregated update data includes: Based on the meta-learning model, the second weight of the model training update data of each student is obtained by the following formula; ； Among them, α i is the second weight of the i-th student, L train,i Update the loss function of the personalized learning recommendation model locally for the i-th student, θ f are the parameters of the meta-learning model; Based on the second weight, the model training update data of each student is weighted to obtain the aggregated update data.

8. A method for online learning management in military academies based on big data analysis, characterized in that: The method comprises: Acquire military learning material data; the military learning material data includes teaching materials, question bank materials and video materials; Extracting knowledge points based on the military learning material data, and constructing a learning knowledge graph based on the knowledge points; Performing model training based on the learning knowledge graph to obtain a personalized learning recommendation model; Distributing the personalized learning recommendation model and the learning knowledge graph to learning devices of multiple students; Obtaining model training update data and knowledge mining data of each student; wherein the model training update data is obtained by updating and training the personalized learning recommendation model based on the local learning records of the learning devices of each student, and the knowledge mining data is the knowledge data mined by the learning devices of each student based on their local learning records; Aggregating the model training update data of each student to obtain aggregated update data; verifying the knowledge mining data based on the aggregated updated data; Updating the verified knowledge mining data to the learning knowledge graph to obtain a target knowledge graph; The target knowledge graph is distributed to the learning devices of the multiple students so that the multiple students can learn based on the target knowledge graph.

Images