To address the limitations currently experienced regarding the comprehensiveness and effectiveness of online learning evaluation in the smart education context, this paper proposes a novel framework for assessing online learning behavior based on the fuzzy analytic hierarchy process(FAHP) and the fuzzy synthetic evaluation method(FSEM). Drawing upon the CIPP(context, input, process, product) educational evaluation model and integrating the educational evaluation tag taxonomy system, the framework identifies five key dimensions: learning exploration, programming practice, knowledge acquisition, collaborative innovation, and communication interaction. These dimensions are further delineated into secondary and tertiary indicators to ensure comprehensive evaluation coverage. The framework utilizes FAHP-FSEM to determine the weights of each indicator level and employs consistency testing to validate the scientific and rational nature of the evaluation process. Implemented on the Shuishan Online platform, the framework leverages extensive multi-source process learning data to facilitate comprehensive evaluation from multiple perspectives and across various dimensions. Student profiles and learning behavior patterns are presented via a visual dashboard. This framework provides robust data support for enhancing personalized learning outcomes and advancing educational reform, demonstrating its broad applicability and potential.

This study presents an OpenRank-based method for evaluating open-source contributions, designed to address the challenge of quantifying student contributions in open-source projects. Taking the “Open-Source Software Design and Development” course as a case study, we developed a method to assess student contributions in open-source practice. The OpenRank algorithm, which is based on developer collaboration networks, evaluates student contributions in discussions, problem-solving, and coding. Experimental results indicate that OpenRank not only aligns with traditional grading methods but also provides a more comprehensive view of student contributions. Combining OpenRank with traditional grading offers a more scientific and thorough evaluation of student contributions and skills in open-source projects.

This study proposes a multi-graph knowledge tracing method integrated with a self-attention mechanism (SA-MGKT), The aim is to model students’ knowledge mastery based on their historical performance on problem-solving exercises and evaluate their future learning performance. Firstly, a heterogeneous graph of student-exercise is constructed to represent the high-order relationships between these two factors. Graph contrastive learning techniques are employed to capture students’ answer preferences, and a three-layer LightGCN is utilized for graph representation learning. Secondly, we introduce information from concept association hypergraphs and directed transition graphs, and obtain node embeddings through hypergraph convolutional networks and directed graph convolutional networks. Finally, by incorporating the self-attention mechanism, we successfully fuse the internal information within the exercise sequence and the latent knowledge embedded in the representations learned from multiple graphs, leading to a substantial enhancement in the accuracy of the knowledge tracing model. Experimental outcomes on three benchmark datasets demonstrate promising results, showcasing remarkable improvements of 3.51%, 17.91%, and 1.47% respectively in the evaluation metrics, compared to the baseline models. These findings robustly validate the effectiveness of integrating multi-graph information and the self-attention mechanism in enhancing the performance of knowledge tracing models.

In recent years, massive open online courses (MOOCs) have become a significant pathway for acquiring knowledge and skills. However, the increasing number of courses has led to severe information overload. Knowledge concept recommendation aims to identify and recommend specific knowledge points that students need to master. Existing research addresses the challenge of data sparsity by constructing heterogeneous information networks; however, there are limitations in fully leveraging these networks and considering the diverse interactions between learners and knowledge concepts. To address these issues, this study proposes a novel method, heterogeneous learning behavior-aware knowledge concept recommendation (HLB-KCR). First, it uses metapath-based random walks and skip-gram algorithms to generate semantically rich metapath embeddings and optimizes these embeddings through a two-stage enhancement module. Second, a multi-type interaction graph incorporating temporal contextual information is constructed, and a graph neural network (GNN) is employed for message passing to update the nodes, obtaining deep embedded representations that include time and interaction type information. Third, a semantic attention module is introduced to integrate meta-path embeddings with multi-type interaction embeddings. Finally, an extended matrix factorization rating prediction module is used to optimize the recommendation algorithm. Extensive experiments on the large-scale public MOOCCubeX dataset demonstrate the effectiveness and rationality of the HLB-KCR method.

In massive open online courses (MOOCs), knowledge concept recommendation aims to analyze and extract learning records from a platform to recommend personalized knowledge concepts to users, thereby avoiding the inefficiencies caused by the blind selection of learning content. However, existing methods often lack comprehensive utilization of the multidimensional aspects of user behavior data, such as sequential information and complex interactions. To address this issue, we propose STRec, a sequence-aware and multi-type behavioral data driven knowledge concept recommendation method for MOOCs. STRec extracts the sequential information of knowledge concepts and combines it with the features produced by graph convolutional networks using an attention mechanism. This facilitates the prediction of a user's next knowledge concept of interest. Moreover, by employing multi-type contrastive learning, our method integrates user-interest preferences with various interaction relationships to accurately capture personalized features from complex interactions. The experimental results on the MOOCCube dataset demonstrate that the proposed method outperforms existing baseline models across multiple metrics, validating its effectiveness and practicality in knowledge concept recommendation.

Automated content reviews on digital educational resources are urgently in demand in the educational informatization era. Especially in the applicability review of whether educational resources exceed the standard, there are problems with knowledge which are easy to exceed national curriculum standards and difficult to locate. In response to this demand, this study proposed a review method for educational resources based on the collaboration of an educational knowledge graph and a large language model . Specifically, this study initially utilized the ontology concept to design and construct a knowledge graph for curriculum education in primary and secondary schools. A knowledge localization method was subsequently designed based on teaching content generation, sorting, and pruning, by utilizing the advantages of large language models for text generation and sorting tasks. Finally, by detecting conflicts between the core knowledge sub-graph of teaching content and the knowledge graph teaching path, the goal of recognizing teaching content that exceeded the national standard was achieved. Experimental results demonstrate that the proposed method effectively addresses the task of reviewing exceptional standard knowledge in educational resource content. This opens up a new technological direction for educational application based on the knowledge graph and large language model collaboration.

Drawing on constructivism and competency-based theory, this paper proposes an online learning system design method based on a knowledge graph, which breaks the traditional knowledge structure and builds a multi-dimensional competence framework of knowledge and skills with the goal of improving competence. A learning system with a knowledge graph as the underlying logic and linked digital learning resources was built. Teaching practice and empirical research were then carried out. First, the learning system was verified with a questionnaire. Second, taking the ability to “read English academic papers” as the learning task, experimental and control groups were created to evaluate the understanding of knowledge and skills, memory level, and comprehensive application ability of the participants. The results showed that the effectiveness and usability of the learning system were higher in the experimental group than in the control group in terms of total, knowledge, skill, and ability scores. Among these, total and ability scores showed significant differences, indicating that the system played a role in promoting the effect of online learning.

With the rapid development of artificial intelligence technology, large language models (LLMs) have demonstrated strong abilities in natural language processing and various knowledge applications. This study examined the application of Chinese large language models in the automatic labelling of knowledge graphs for primary and secondary school subjects in particular compulsory education stage morality and law and high school mathematics. In education, the construction of knowledge graphs is crucial for organizing systemic knowledge . However, traditional knowledge graph methods have problems such as low efficiency and labor-cost consumption in data labelling. This study aimed to solve these problems using LLMs, thereby improving the level of automation and intelligence in the construction of knowledge graphs. Based on the status quo of domestic LLMs, this paper discusses their application in the automatic labelling of subject knowledge graphs. Taking morality and rule of law and mathematics as examples, the relevant methods and experimental results are explained. First, the research background and significance are discussed. Second, the development status of the domestic large language model and automatic labelling technology of the subject knowledge graph are then presented. In the methods and model section, an automatic labelling method based on LLMs is explored to improve its application in a subject knowledge graph. This study also explored the subject knowledge graph model to compare and evaluate the actual effect of the automatic labelling method. In the experiment and analysis section, through the automatic labelling experiments and results analysis of the subjects of morality and law and mathematics, the knowledge graphs of the two disciplines are automatically labeled to achieve high accuracy and efficiency. A series of valuable conclusions are obtained, and the effectiveness and accuracy of the proposed methods are verified. Finally, future research directions are discussed. In general, this study provides a new concept and method for the automatic labelling of subject knowledge graphs, which is expected to promote further developments in related fields.

Advancements in machine-learning technology has enabled automated program-repair techniques that learn human patterns of erroneous-code fixing, thereby assisting students in debugging and enhancing their self-directed learning efficiency. Automatic program-repair models are typically based on either manually designed symbolic rules or data-driven methods. Owing the availability of large language models that possess excellent natural-language understanding and code-generation capabilities, researchers have attempted to use prompt engineering for automatic program repair. However, existing studies primarily evaluate commercial models such as Codex and GPT-4, which may incur high costs for large-scale adoption and cause data-privacy issues in educational scenarios. Furthermore, these studies typically employ simple prompt forms to assess the program-repair capabilities of large language models, whereas the results are not analyzed comprehensively. Hence, we evaluate two representative open-source code large language models with excellent code-generation capability using prompt engineering. We evaluate different prompting methods, such as chain-of-thought and few-shot learning, and analyze the results comprehensively. Finally, we provide suggestions for integrating large language models into programming educational scenarios.

Against the backdrop of the national new engineering education initiative, early C++ teaching has failed to meet the requirements of high-level sophistication, innovation, and challenges. Furthermore, issues such as fragmented knowledge points, difficulty in integrating theory with practice, and single-perspective bias are prevalent in this field. To address these problems, we propose an innovative teaching model that effectively integrates QT(Qt Toolkit) and C++ by merging the two courses. This model facilitates the teaching process via a course knowledge graph deployed on the Zhihuishu platform. The breadth of teaching is expanded by effectively linking course knowledge points, integrating and sharing multimodal teaching resources, enhancing multiperspective learning, showcasing the course’s innovative nature, and avoiding single-perspective bias. Simultaneously, the depth of teaching is increased through the construction of a knowledge graph that integrates QT and object-oriented programming (C++), organically combining the knowledge points of both courses. This approach bridges the gap between theory and practice by enhancing the course’s sophistication and level of challenge. Consequently, this study pioneers the reform of C++ teaching by providing valuable references and insights for programming courses under the new engineering education framework.

In the digital education application domain, developers of platforms such as online classrooms face the challenges of privacy issues and existing datasets’ insufficient size in their pursuit of data-driven optimization. To address this, a set of heterogeneous data models adapted to the characteristics of education were constructed, and corresponding data generation tools (E-Tools) that can be used to simulate data interactions in complex educational scenarios were implemented. Experimental results have shown that the tool can maintain an efficient data generation speed (64–74 $ {\rm{MB}}\cdot {{\rm{s}}^{-1}} $) under a variety of data sizes, demonstrating good linear scaling ability, which validates the model’s effectiveness and the tool’s ability to generate larger data volumes. A heterogeneous data query load reflecting students’ learning behaviors was also designed to provide strong support for performance evaluation and the education platform’s optimization.

Conventional education big data management is faced with security risks such as privacy data leakage, questionable data credibility, and unauthorized access. To avoid the above risks, a novel type of education big data security management and privacy protection method, Algorithm for security management and privacy protection of education big data based on smart contracts (ASPES), is proposed. It integrates an improved key splitting and sharing algorithm based on the secret sharing of Shamir, a hybrid encryption algorithm based on SM2-SHA256-AES, and a smart contract management algorithm based on hierarchical data access control. Experiments are conduced on the real dataset of MOOCCube and the results indicate that the execution efficiency and security of ASPES are significantly improved when compared with the state-of-the-art methods, which can effectively store and manage education big data and realize the reasonable distribution of educational resources. By embedding smart contracts into the blockchain and inputting operations like data reading and writing into the blockchain, ASPES can optimize the management path, improve management efficiency, ensure the fairness of education, and considerably improve the quality of education.

Query optimization can significantly enhance the analysis efficiency of online analytical processing (OLAP) database systems for massive educational data, providing fast and accurate data support for intelligent educational systems. The optimizer mainly consists of three modules: cardinality estimation, space enumeration, and cost models. Specifically, cardinality estimation determines the results of the cost model and guides the selection of query plans. Therefore, the evaluation of the cardinality estimation module of the optimizer plays a crucial role in the optimization of OLAP database systems. This study designs and implements an effective workload generation tool based on primary key-driven diversified data distribution and data relationship construction. The tool includes data generation technology with custom relationships, workload template generation technology based on finite state machines, and parameter instantiation technology driven by target cardinality. Experiments were conducted on three databases: OceanBase, TiDB, and PostgreSQL, analyzing the issues of their optimizers and providing suggestions.

In the modern educational environment, efficient and reliable data management systems are essential for the operation of online education platforms and student information management systems. With the continuous growth of educational data and the increase in the frequency of multi-user access, database systems face the challenge of high throughput requirements owing to concurrent conflict operations. Among the many concurrency control strategies, the lock-based control strategy is commonly used in database systems. However, the blocking caused by locks affects the performance of concurrent execution of transactions in the database. Existing work mainly reduces lock contention by scheduling the execution order between transactions or optimizing stored procedures. To improve transaction throughput further, this study conducts blocking analysis and cost modeling within transactions based on lock avoidance, and proposes an intra-transaction scheduling strategy. The scheduling cost is estimated by analyzing the blocking of the workload, and then the operation order is exchanged to a limited extent within the transaction according to certain rules to reduce the delay caused by lock blocking, thereby improving performance. Finally, comparing the conventional and proposed scheduling strategies, the latter is verified to improve throughput and reduce the average transaction delay.

This study introduces and implements a local, lightweight, intelligent teaching-assistant system. Using the IPEX-LLM (Intel PyTorch extention for large language model) acceleration library, the system can efficiently deploy and execute large language models that are fine-tuned using the QLoRA (quantum-logic optimized resource allocation) framework on devices with limited computational resources. Combining this with enhanced retrieval techniques, the system provides flexible course customization through four major functional modules: intelligent Q&A, automated question generation, syllabus creation, and course PPT generation. This system is intended to assist educators in improving the quality and efficiency of lesson preparation and delivery, safeguarding data privacy, supporting personalized student learning, and offering real-time feedback. Performance tests exemplified by the optimized Chatglm3-6B model show the rapid inference capability of the system via the processing of a 64-token output task within 4.08 s in a resource-constrained environment. A practical case study comparing the functionality of the system with native Chatglm-6B and ChatGPT 4.0 further validates its superior accuracy and practicality.

Building an intelligent education platform is an important process in the promotion of the intelligence of education. However, the artificial intelligence model on which intelligent education platforms rely consumes a large amount of electricity and energy during its training process, therefore, it is of great significance to carry out a short-term power load prediction for building an intelligent education platform. However, the major issue is the weak correlations between some attributes and power load data when considering multiple attributes during short-term power load forecasting, and the Transformer cannot capture the temporal correlation of power load data, which leads to a lack of accuracy in power load forecasting. Therefore, a short-term power load forecasting model, SF-Transformer is proposed that is based on the SR (Székely and Rizzo) distance correlation coefficient, fusion temporal localization coding and Transformer. The SF-Transformer filters the attributes that affect the power load data by using the SR distance correlation coefficient and selects the attributes that have higher SR distance correlation coefficients with the power load data. The SF-Transformer adopts fusion time localization coding, thereby combining global time coding and local position coding, which helps the model to comprehensively obtain time and localization information regarding power load data. Experiments conducted on the dataset show that SF-Transformer has a lower RMSE (root mean square error) and MAE (mean absolute error), compared with those of other power load forecasting models over two-time durations.

Recently, with the development of artificial intelligence, visual recognition, and edge intelligent computing, intelligent patrol or online monitoring technology based on visual recognition has found important applications in conventional campus security, laboratory safety monitoring, and industrial production operation and maintenance monitoring. Campus security and laboratory safety monitoring aim to protect the personal safety of students and teachers and avoid incidents such as campus bullying or laboratory safety accidents. Industrial production operation and maintenance monitoring is the identification and early warning of hidden dangers and defects in equipment or operational behaviors in industrial scenarios to avoid huge losses caused by faults and hazards. In security and production operation monitoring tasks, using manual methods for real-time detection can be labor-intensive and inefficient, and human negligence that leads to undetected dangers could occur. Therefore, based on the needs of campus security and safety or industrial production operation and maintenance monitoring, this study designs and implements an intelligent patrol system based on microservices and the operation and maintenance monitoring of industrial substations. The system does not require excessive manual participation and can automatically conduct patrols, identify dangers, and provide early warnings. Subsequently, the system adopts an advanced scheduling system that takes only 3–5 min to perform one patrol, considerably improving the efficiency of hazard detection during patrols. The system can be applied to intelligent patrols of campus and industrial substations, security, and safety.