International Journal of Engineering Intelligent Systems

Computer Software Fault Location Method Based on Machine Vision and Image Processing

2025-12-12T22:24:43-08:00

With the rapid development of science and technology, software applications have penetrated various industries, and the requirements for software systems have become increasingly higher. A variety of software systems have gradually changed people’s lives and their economies. When people demand more and more software, software failures tend to occur. In order to reduce the harm caused by software errors, this paper proposes a computer software fault location method based on machine vision and image processing. First, an analysis was conducted of the graph mining technology, and then an experiment was conducted to compare the two different graph reduction methods in series and parallel to determine their fault location efficiency. According to the experimental results, when the number of nodes was small, the time consumed by serial reduction in fault
location was shorter. However, when the number of nodes increased, the advantages of parallel reduction became increasingly obvious. When the number of nodes was 5646, the efficiency reached 55.22%, which significantly improved the efficiency of fault location. This can help programmers to locate and repair the application faults, improve the performance of software security, and reduce the loss caused by software failure.

Keywords: software fault location; machine vision; image processing; graph mining technology

Cite As

W. Yan, C. Yang, "Computer Software Fault Location Method Based on Machine Vision and Image
Processing", Engineering Intelligent Systems, vol. 33 no. 6, pp. 611-619, 2025.

Stability Analysis and Optimization of Power System in New Energy Grid Connection Control

2025-12-12T22:33:21-08:00

In response to the challenges that the volatility and uncertainty of new energy pose to the stability of the power system, this study applied the Long Short-Term Memory (LSTM) algorithm to establish a prediction model for new energy generation. By combining historical data and meteorological information, the stability of the system was analyzed in depth. The LSTM model adopted a multi-layer stacked structure, with the input layer receiving multidimensional features from the SDWPF (Spatial Dynamic Wind Power Forecasting) dataset and using 128 hidden units to capture dynamic features in the time series. This study combined Adam optimizer to improve training efficiency and introduced the Dropout mechanism to reduce the risk of overfitting. Based on the prediction results of LSTM, this study designed dynamic scheduling and energy storage system optimization strategies to
improve the stability of the power system. The dynamic dispatch strategy adjusted the power generation of the generator sets in real-time to match the load demand, and the energy storage system smooths wind power fluctuations by means of charge and discharge management, thereby maintaining the stability of the grid frequency and voltage. Experimental data show that the LSTM model performs well in terms of prediction, with a mean square error as low as 0.0021 and a mean absolute error of 0.037. When the wind speed fluctuates at low speeds, the dispatch strategy stabilizes within 5 seconds; when the fluctuation is high, the adjustment time and error increase. The energy storage system responds well to low fluctuations; but although it responds quickly to high fluctuations, it has reduced stability, highlighting the advantages and disadvantages of each strategy under different conditions.

Keywords: new energy stability, LSTM algorithm, power system dispatch, energy storage system optimization, load management

Cite As

C. Chen, L. Wang, C. Deng, J. Jiang, S. Zhong, "Stability Analysis and Optimization of Power System in New
Energy Grid Connection Control", Engineering Intelligent Systems, vol. 33 no. 6, pp. 621-630, 2025.

Intelligent Logistics Scheduling Algorithm in Dynamic Traffic Environment

2025-12-12T22:41:31-08:00

With the increasingly heavy traffic conditions, the dynamic traffic environment has gradually become a serious problem for the scheduling of logistics vehicle transportation. Traditional heuristic scheduling algorithms are mostly based on static scheduling, lacking adaptability to dynamic environments that involve, for instance, traffic accidents and natural disasters, and failing to fully consider multiple issues such as cost and time for optimization, resulting in low scheduling efficiency. This study used the multi-objective optimization of MAPPO (Multi-Agent Proximal Policy Optimization) and the dynamic information extraction advantages of LSTM (Long Short-term Memory) to study intelligent logistics scheduling under dynamic traffic environments. First, the study matched the traffic monitoring data and Google maps API (Application Programming Interface) data of similar time with the logistics distribution data according to the timestamp, and combined the geocoding of the Geopy library to match the geographic location and records according to the nearest matching method. Subsequently, a mapping relationship between traffic section ID and delivery route section was established, and urban traffic system data and real-time traffic data were linked to each delivery record. Then, the LSTM model was used to capture the dynamic information of the traffic environment, generate predicted traffic flow and congestion conditions, and finally input the traffic flow, speed and other states predicted by LSTM into the MAPPO algorithm model to assist the logistics vehicle intelligent body to dynamically adjust route selection and other scheduling according to traffic conditions. The experiment was based on data from the urban traffic system of the Shenzhen Traffic
Management Center and a logistics center from June to December 2023, and intelligently dispatched logistics vehicles in a dynamic environment. The results showed that in peak traffic flow, the dispatch efficiency of MAPPO-LSTM reached 87.5%, an increase of 3.5% compared to the MAPPO algorithm. The overall satisfaction score reached a high 14 points. Experiments show that the MAPPO-LSTM algorithm has good adaptability to dynamic traffic environments, greatly improves scheduling efficiency, and provides efficient guarantees for intelligent logistics transportation.

Keywords: dynamic traffic environment, intelligent logistics scheduling, MAPPO algorithm, LSTM model, scheduling efficiency

Cite As

X. Wu, X. Du, Y. Liu, "Intelligent Logistics Scheduling Algorithm in Dynamic Traffic Environment", Engineering Intelligent
Systems, vol. 33 no. 6, pp. 631-644, 2025.

Deep Learning-Driven Tourism Data Mining and Fuzzy Prediction Methods

2025-12-12T22:49:19-08:00

Tourism is a dynamic industry influenced by a variety of factors, including uncertainty about tourist behavior, seasonal variations, and the impact of external environmental factors. With the development of big data technology and artificial intelligence, the tourism industry has begun to seek more efficient ways to understand and predict tourists’ behavioral patterns and their impact on the tourism economy. However, traditional prediction models often struggle to accurately capture the ambiguity and uncertainty in tourism data, especially when confronted with complex time series data and multimodal data. This study proposes an innovative deep learning-driven tourism data mining framework that incorporates fuzzy logic to improve the accuracy of predictions regarding tourist numbers and expected revenue. By employing a deep learning architecture with multimodal fusion, the framework is able to handle both structured and unstructured data and utilize the attention mechanism to highlight key features. The integration of fuzzy logic further enhances the model’s ability to handle uncertainty and fuzzy information. The experimental results show that the proposed method performs better than the traditional baseline model under a variety of evaluation metrics, especially in dealing with seasonal variations and uncertainty. Specifically, in terms of tourist number prediction, the MSE, RMSE and MAE of the proposed method are 75.42, 8.68 and 5.23, respectively, while in terms of revenue prediction, these metrics are 1023.21, 31.99 and 23.42, respectively, which are significantly lower than the baseline models, such as
linear regression, support vector machine, random forests, and long and short-term memory networks. In addition, the proposed method shows high stability when dealing with seasonal variations in different months, which can provide reliable forecasting support for tourism companies throughout the year.

Keywords: deep learning, tourism data, data mining, fuzzy prediction

Cite As
1International Education School, Guangzhou College of Technology and Business, Guangzhou 510000, China
2School of Foreign Language and International Trade, Guangdong Polytechnic, Foshan 528000 China
3Macau University of Science and Technology, Macau 999078, China
4School of Tourism and Leisure Management, Fujian Business University, Fuzhou 350012, China
Xiaochen Li1, Jiayu Wu2, Yuqing Zhang 1,3,? and Yanling Xiao4

X. Li, J. Wu, Y. Zhang, Y. Xiao, "Deep Learning-Driven Tourism Data Mining and Fuzzy Prediction
Methods", Engineering Intelligent Systems, vol. 33 no. 6, pp. 645-655, 2025.

Relationship between Industry-University-Research Collaborative Education Model Combined with Kohonen Network and the Improvement of Students’ Multiple Intelligences

2025-12-12T22:56:44-08:00

The purpose of this study is to explore the effect of applying the Kohonen network to the Industry-University-Research (IUR) collaborative education model. An improved system based on the Kohonen network is designed and applied to the educational intervention of 200 students, and the students’ multiple intelligences in eight dimensions are evaluated. The results show that: (1) After implementing this model, students’ scores for multiple intelligences are significantly improved. The average score for logical mathematics intelligence increased from 65.3 to 74.5, an improvement of 14.1%. Interpersonal intelligence and spatial intelligence scores increased by 12.1% and 11.9% respectively. The scores for language intelligence, body movement intelligence, music intelligence, introspection intelligence and natural observation intelligence increase by 10.9%, 11.4%, 10.1%, 11.1% and 10.4% respectively. (2) The Kohonen network had an obvious impact on data analysis. By means of visualization, the Kohonen network
divides students into four groups through self-organizing clustering. The 41 students in Cluster 1 performed well in all types of intelligence, with a score of 83.2 for music intelligence and 80.3 for body movement intelligence. The overall score of 25 students in Cluster 2 is low, and their logical and mathematical intelligence is only 59.9. The intelligence scores of 91 students in Cluster 3 are relatively balanced, but lower than those in Cluster 1, and their language intelligence is 65.7. The 82 students in Cluster 4 are outstanding in spatial intelligence and limb movement intelligence, which are 69.8 and 71.9 respectively. These results show that the Kohonen network can effectively classify students’ multiple intelligences levels and provide data support for further educational intervention. In addition, this model has significantly improved students’ innovation skills and team cooperation
ability. More than 80% of students’ feedback suggested that this model helps them to better understand and apply what they have learned. The Kohonen network as a means of fostering multiple intelligences has demonstrated its great potential for application in the field of education, supports the strong integration of theory and practice, and provides a strong incentive for future education reform.

Keywords: Kohonen network; Industry-University-Research Collaborative Education Model; multiple intelligences; logical mathematical intelligence; average score

Cite As

Y. Wang, W. Zhang, G. Li, S. Ding, "Relationship between Industry-University-Research
Collaborative Education Model Combined with Kohonen Network and the Improvement
of Students’ Multiple Intelligences", Engineering Intelligent Systems, vol. 33 no. 6, pp. 655-666,
2025.

Yan Wang?, Weili Zhang, Guanhong Li and Shuming Ding
Medical College, Zhengzhou Institute of Industrial Application Technology, Zhengzhou 451150, China

Innovative Application of Cultural Heritage Education Resources Combined with Fuzzy Algorithm in Modern History Teaching

2025-12-12T23:05:48-08:00

This study utilizes a fuzzy algorithm to explore the innovative application of cultural heritage educational resources in modern history teaching, optimize the integration path of educational resources, improve teaching outcomes, and integrate traditional culture with modern education. With the acceleration of globalization, the protection and survival of cultural heritage have become an important subject of modern education. In modern history teaching, course content needs to incorporate rich cultural heritage resources as a key way to foster students’ cultural identity and historical consciousness. This study uses a fuzzy algorithm for the integration and application of cultural heritage educational resources. It optimizes the allocation of teaching resources by intelligent means, and improves the quality of education. Firstly, it analyzes the characteristics of cultural heritage educational resources, points out their importance and uniqueness in modern history teaching, and addresses many challenges posed by the integration of cultural heritage educational resources, particularly in terms of providing diversified educational content and individualized teaching needs. Then, based on the fuzzy comprehensive evaluation model and the fuzzy clustering algorithm, an educational resource optimization model was constructed. Through methods such as data collection, questionnaire surveys, and interviews, the teaching data of educational institution A was obtained, and fuzzy algorithms were used to scientifically evaluate and optimize educational resources. The results show that the fuzzy algorithm can effectively deal with the uncertainty in the teaching process, provide personalized learning path recommendation, and effectively integrate cultural heritage resources and
modern education technology through data analysis using the weighted average method. The results show that the application of the fuzzy algorithm in the integration of educational resources improves the accuracy of resource allocation and helps to create personalized learning paths. For model verification, the cross-validation method is adopted to determine the effectiveness and accuracy of the model, and confirm the feasibility of applying a fuzzy algorithm in the field of education. The study also found that the model has limitations in terms of data dependence and real-time update, and future research needs to optimize the adaptability and operability of the algorithm. By combining the fuzzy algorithm with cultural heritage education, this study provides a new approach to the optimization of educational resources, a feasible scheme for the integration of cultural heritage resources in modern history teaching, and promotes the innovative development of cultural heritage education.

Keywords: fuzzy algorithm; cultural heritage education; modern history teaching

Cite As

L. Zhong, "Innovative Application of Cultural Heritage Education Resources Combined with Fuzzy Algorithm
in Modern History Teaching", Engineering Intelligent Systems, vol. 33 no. 6, pp. 667-680, 2025.

Combine the network model to conduct dynamic monitoring and prediction of the mental health status of students

2025-12-12T23:11:23-08:00

In order to meet the needs of students’ mental health management, this study designed and implemented a dynamic mental health monitoring and prediction system based on big data. Based on multi-dimensional data, a multi-source data collection framework is constructed, and a deep neural network model is combined to accurately classify and predict the trend of students’ mental health state. The system has advantages in terms of classification accuracy, operational efficiency and data processing ability, enabling educational administrators to detect mental health risks in a timely manner. The system designed real-time data acquisition, index analysis, state evaluation and intervention suggestions and other functional modules, forming a closed-loop monitoring and feedback system. The differences in classification performance and the lack of generalization ability are also
identified, and optimization paths, minority class sample enhancement, data diversity expansion and system performance optimization are proposed. It provides scientific tools and implementation paths for students’ mental health management, and promotes the development of the field of dynamic monitoring of mental health.

Keywords: mental health; big data analysis; dynamic monitoring

Cite As

N. Li, D. Kang, H. Ran, "Combine the network model to conduct dynamic monitoring and prediction of the
mental health status of students", Engineering Intelligent Systems, vol. 33 no. 6, pp. 681-688, 2025.

Classroom Teaching Behaviour Analysis and Teaching Quality Evaluation Design Based on Deep Learning

2025-12-12T23:17:52-08:00

With the rapid development of educational technology, the analysis of students’ classroom behavior has been an important focus of educational research. Traditional classroom monitoring methods rely on actual observation and cannot provide accurate and comprehensive analysis. Multimodal data analysis combined with deep learning technology provides a new way to solve this problem. This paper proposes a multimodal classroom behavior analysis system based on video, audio and sensor data, which aims to accurately identify and predict the behavior of teachers and students through deep learning models to improve the quality of classroom teaching. The system includes a data acquisition module, data preprocessing module, feature extraction module, behavior analysis and prediction module, and result feedback module. First, classroom behavior data is collected
through video, audio and sensors, and the data is preprocessed by denoising, normalization and other operations. Then, a convolutional neural network (CNN) is used to extract image features from video, the MFCC method is used to extract spectral features from audio, and the LSTM model is used to extract time series features from sensor data. Then, the system uses deep neural network (DNN) for behavior classification and LSTM to predict learning status and teaching quality. Finally, the analysis results are fed back to teachers and education managers in the form of reports. Experimental results show that the system performs well in terms of behavior classification and regression tasks, with higher accuracy, precision and F1 score than
traditional models, and the system has stable performance in different tasks. This research not only provides new ideas for classroom behavior analysis, but also provides decision support tools for educational administrators.

Keywords: multimodal data, deep learning, behavior analysis, classroom monitoring, educational technology

Cite As

Y. Lian, F. He, S. Wang, "Classroom Teaching Behaviour Analysis and Teaching Quality Evaluation Design Based
on Deep Learning", Engineering Intelligent Systems, vol. 33 no. 6, pp. 689-699, 2025.

Design and Implementation of Red Culture Education experience System Based on VR Technology

2025-12-12T23:23:30-08:00

Red culture is an important heritage of Chinese revolutionary history, and contains rich educational value. Traditional modes of red culture education lack interaction and immersion, and therefore cannot fully meet the needs of today’s learners. In order to create a more vivid and effective means of conveying red culture, this study combined virtual reality technology to design and achieve an immersive educational experience system. The integration of technology and education enables learners to improve their knowledge mastery, strengthen their emotional identification, and facilitate the communication of red culture. Based on constructionist learning theory, immersion theory and multimedia learning theory, a system architecture is constructed that integrates scenario-immersion experience, dynamic content adaptation and user interaction optimization. The system development involves data collection, scenario modeling and algorithm design, ultimately building a red culture education platform suitable for different learning scenarios. The results show that the system improves the user’s knowledge mastery and emotional recognition, and optimizes the learning experience and ease-of-operation. It demonstrates the broad applicability of virtual reality technology in red culture education.

Keywords: virtual reality technology; red culture education; Immersive learning

Cite As

C. Liu, "Design and Implementation of Red Culture Education experience System Based on VR Technology",
Engineering Intelligent Systems, vol. 33 no. 6, pp. 701-710, 2025.

Application of Fuzzy Clustering Algorithm in the Analysis of Students’ Learning Styles on International Education Platforms

2025-12-12T23:28:13-08:00

This paper discusses the analysis and application of middle school students’ learning styles in a cross-border education platform, facilitated by a fuzzy clustering algorithm. With the increasing popularity of online education, it has become important to personalize the course content and teaching approach according to students’ individual learning styles. How to do so in order to improve learning outcomes has become a core issue in the education field. For this study, student learning data from multiple cross-border education platforms was collected, and a fuzzy clustering algorithm was applied to classify students’ learning behaviors. “Five main learning styles were identified: visual, auditory, kinesthetic, reading/writing, and multimodal , in which learners flexibly combine multiple strategies rather than relying on a single mode. These learning styles reflect the differences
in students’ preferences during the learning process, demonstrating the importance of designing personalized learning paths. The research results show that the fuzzy clustering algorithm can cope with the diversity and complexity of students’ learning behavior, accurately identify each student’s learning style by calculating the degree of membership, and provide a basis for a personalized teaching strategy design for the platform. The evaluation results demonstrate that the proposed fuzzy clustering algorithm enhances not only the classification accuracy but also the robustness of learning style identification, ensuring consistent and reliable outcomes under varying data conditions. This study provides new ideas and methods whereby cross-border education platforms can analyze students’ personalized learning, and provides theoretical basis and practical guidance for educators and platform developers wishing to implement personalized teaching.

Keywords: fuzzy clustering algorithm; cross-border education; learning style; personalized teaching; data analysis

Cite As

Y. Hu, "Application of Fuzzy Clustering Algorithm in the Analysis of Students’ Learning Styles on
International Education Platforms", Engineering Intelligent Systems, vol. 33 no. 6, pp. 711-719, 2025.

Badminton Action Recognition Model Combining Artificial Neural Network Algorithm and DTW Algorithm

2025-12-12T23:34:02-08:00

Badminton is a sport that is susceptible to the influence of the racket and the player’s attire, and it is difficult to recognize the action without the help of 3D-capturing technology. To accurately recognize the technical movements involved in badminton, this study proposes a badminton action recognition model that combines an artificial neural network algorithm and a dynamic time warping algorithm. The badminton player’s posture can be estimated using the improved OpenPose algorithm, which can accurately capture the player’s key skeletal nodes. Based on this, the mobile network V3 network architecture and dynamic time warping algorithm are used to recognize the action. The results revealed that the mobile network V3 model achieved significant performance in recognizing badminton movements, with an accuracy of 0.987, which was significantly higher than the results obtained with the visual geometric group network, the residual network and the mobile network V2 model. Moreover, the total number of parameters was significantly decreased by 4.7M. In addition, the precision of the model was improved by an average of 12.27%, outperforming the other three network models. The dynamic time warping algorithm also performed well in evaluating badminton technical movements. The results of the evaluation were significantly improved by introducing the weighting values. In the smash and split score, the score based on dynamic time warping algorithm was 84, while the score after introducing the weights reached 86, which was closer to the scoring value of the domain experts. The results demonstrated that the mobile network V3 model with the dynamic time warping algorithm used in the study was able to achieve accurate recognition
of badminton movements with high computational efficiency. The study provides an efficient badminton action recognition and evaluation method, which helps to improve the scientific basis and effectiveness of athletes’ training, and also provides a new technical means for sports analysis.

Keywords: Badminton action recognition; MobileNetV3; Improved OpenPose algorithm; DTW algorithm; Joint points

Cite As

Y. Zhong, J. Yang, W. Guo, "Badminton Action Recognition Model Combining Artificial Neural Network
Algorithm and DTW Algorithm", Engineering Intelligent Systems, vol. 33 no. 6, pp. 721-732, 2025.

Risk Assessment of Corporate Financial Internal Control and Audit Matters: Data Mining

2025-12-12T23:38:10-08:00

Internal financial audits within enterprises are beneficial for planning the operations of enterprises and enhancing their market competitiveness. In this study, the random forest (RF) algorithm was combined with the back-propagation neural network (BPNN) algorithm to improve financial risk recognition and auditing accuracy. The RF algorithm was used to screen the main financial risk indicators, and the BPNN algorithm was employed to identify the financial risk. The RF algorithm calculated the importance of financial risk indicators in the simulation experiment. Then, the performance of the RF algorithm, traditional BPNN algorithm, and the improved BPNN algorithm was compared. The results showed that the RF algorithm effectively screened out the characteristics of important financial risk indicators. Compared with the single RF algorithm and the single BPNN algorithm, the
improved BPNN algorithm had better recognition accuracy and shorter recognition time.

Keywords: financial risk, audit, data mining, random forest

Cite As

S. Ren, "Risk Assessment of Corporate Financial Internal Control and Audit Matters: Data Mining",
Engineering Intelligent Systems, vol. 33 no. 6, pp. 733-737, 2025.