Ebook: Machine Learning and Intelligent Systems

The International Conference on Machine Learning and Intelligent Systems (MLIS) is an annual conference aimed at the development and application of machine learning and intelligent systems, technologies that have found applications in all walks of life.

The 6th International Conference on Machine Learning and Intelligent Systems (MLIS 2024) will be held from 17 to 20 November 2024, in Kampar, Perak, Malaysia, with the support of the Faculty of Engineering and Green Technology (FEGT), Universiti Tunku Abdul Rahman, Malaysia. The plenary session of MLIS 2024 will include keynote speeches, invited speeches, poster presentations, and oral presentations. Among these, we are honored to have three distinguished keynote speakers: Prof. Rinat O. Esenaliev from the Department of Neurobiology, University of Texas, USA; Prof. Vladan Devedzic from the Faculty of Organizational Sciences at the University of Belgrade, Serbia; and Prof. Ankush Ghosh from the University Centre for Research & Development (UCRD), Chandigarh University, India.

This book contains the proceedings of the 6th International Conference on Machine Learning and Intelligent Systems (MLIS 2024). All papers were thoroughly reviewed by program committee members and peer reviewers, taking into account the breadth and depth of research topics that fall within the scope of MLIS. From approximately 70 submissions, 18 of the most promising and FAIA mainstream-relevant contributions were selected for inclusion in this book.

On behalf of the conference organizers, we would like to take this opportunity to express our sincere gratitude to the Guest Editor and the reviewers for their tremendous efforts and dedication to the conference; to the authors for their valuable contributions; and to all our colleagues at IOS Press for their support and tireless efforts toward the publication of the conference proceedings. We believe that with such excellent support and contributions, future MLIS Conferences will continue to reach new heights.

Guest Editor:

Prof. Jon-Lark Kim

Sogang University, South Korea

This research aims to enhance higher-order thinking Skills (HOTS) in undergraduates by exploring the integration of Challenge-Based Learning (CBL) and Personal Learning Environments (PLEs). Through a thorough review of existing literature, the study develops a comprehensive platform that combines the practical, real-world problem-solving nature of CBL with the customizable, digital features of PLEs. The platform is meticulously designed to support CBL within the framework of PLE, offering a wide array of resources, collaborative workspaces for peer interaction, real-time analytics for ongoing assessment, and a continuous cycle of reflection and feedback. The paper highlights the potential of this integration to significantly improve analytical and critical thinking skills among learners. While grounded in theoretical research, the study identifies a new area for future empirical research and practical application, advocating for further exploration to fully harness the potential of this innovative educational approach.

In this paper, the existence of solutions for fractional hybrid differential inclusions with tree-point boundary hybrid conditions is investigated:

^CD^p_0⁺(ξ(s)-m(s,ξ(s))/n(s,ξ(s))) ∈ L(s,ξ(s)),0<p≤2,

α₁(ξ(s)-m(s,ξ(s))/n(s,ξ(s)))_s=0+β₁(ξ(s)-m(s,ξ(s))/n(s,ξ(s)))_s=a = γ₁,

α₂^CD^q_0⁺(ξ(s)-m(s,ξ(s))/n(s,ξ(s)))_s=r+β₂^CD^q_0⁺(ξ(s)-m(s,ξ(s))/n(s,ξ(s)))_s=a = γ₂,0<r<a,

where D^p_0⁺ and D^q_0⁺ denotethe Caputo fractional derivative of order p,q respectively. 0<q≤1,α_i,β_i,γ_i,i=1,2, such that

α₁+β₁ ≠ 0, α₂r^1-q+β₂a^1-q ≠ 0, m ∈ C([0,a]×R,R),

n ∈ C([0,a]×R,R∖{0}), L ∈ [0,a]×R→P (R), is a multivalued map. By means of the multi-valued hybrid fixed point theorems, we present sufficient conditions for the existence of solutions for the fractional hybrid differential inclusions with three-point boundary hybrid conditions. An illustrative example is given to show the effectiveness of our main result. We generalize the single known results to the multi-valued ones.

In this paper, we investigate the following problem

^HD^{ϕ₁,η₁;ψ}_α⁺(^HD^{ϕ₂,η₂;ψ}_α⁺ξ(s)) ∈ L(s,ξ(s)),

ξ(α) = 0, ξ(β) = ∑^m_i=1ω_iξ(θ_i), s ∈ S = [α,β],

where ^HD^{ϕ₁,η₁;ψ}_α⁺, ^HD^{ϕ₂,η₂;ψ}_α⁺ denote the ψ-Hilfer fractional derivative of order ϕ₁, ϕ₂ respectively. ϕ₁, ϕ₂ ∈ (0,1), η₁, η₂ ∈ (0,1), ω_i ∈ R⁺, L is a multivalued map on [α,β]×R. By means of the multi-valued fixed point theorems, sufficient conditions for the existence of solutions for the ψ-Hilfer fractional differential inclusions with multi-point boundary conditions are presented. We give an example to show the effectiveness of the main theorem.

The art of music, integral to the human experience, has been present since ancient civilizations. Diverse musical styles emerge from various cultures, geographic locations, and historical periods. Nevertheless, creative individuals may unknowingly produce compositions resembling existing works, even without exposure to similar pieces. Consequently, the cultivation of evolutionary music is crucial for generating a broad spectrum of compositions less likely to evoke familiarity. Hence, music appreciation is significantly subjective from one person to another in society. With the advancement of today’s technology, evolutionary music can be created at ease using algorithmic composition methods from evolutionary algorithms. In this manuscript, we present a method for creating concise, non-binary genetic representations of music notes to generate short, monophonic melodies without harmonic accompaniment. This approach is specifically tailored for the preliminary stage of the research project. Genetic Algorithm (GA) is a metaheuristic about an evolutionary algorithm based on the natural selection processes, which are appropriate for the approach. Genetic Algorithms are versatile that can be combined with other algorithms to produce melodies, harmonies, chords, music structures, scales and others. The novelty of my proposed method is able to create monophony melodies with just solely genetic algorithm without having additional algorithms for the rhythms. Therefore, a literature survey is conducted to provide insight into the latest research using other machine learning methods for evolutionary music. Deep learning is a subset newer to machine learning that increases the accuracy of the results with more extensive data sets.

With universities continually producing data in their laboratories, effective data management has become increasingly crucial. To meet this demand, the application of laboratory information management systems (LIMS) is becoming increasingly widespread. LIMS is a complex computing system used to manage laboratory data. There are currently a variety of LIMS to choose from, but most LIMS use proprietary codes, so the development cost is very high. At the same time, existing LIMS models usually only focus on functional requirements and ignore the specific means of functional implementation, which leads to difficulties for users in management. In addition, since LIMS are very complex, they are often designed to meet the needs of specific laboratories, which makes their versatility and reusability very poor. Therefore, this paper presents the Collaborative Laboratory Management Model(CLMM), which introduces advanced management ideas based on the LIMS model, aiming to make the model focus on both function and technology. CLMM integrates workflows, and users can create and manage workflows themselves. This not only improves the flexibility of the model, but also increases reusability, which can meet the needs of different types of laboratories. With the development of the model, we hope to improve the management efficiency of laboratories by managing the data of university laboratory information, equipment, instruments, etc., and contribute to the complex management work of universities.

In this paper, we provide an overview of Human-Computer Interaction (HCI) in the context of Artificial Intelligence (AI), focusing on Machine Learning (ML), Deep Learning (DL), Artificial Neural Networks (ANN), and applications of these techniques in the field of Healthcare. This review highlights the critical role of the human factor in AI-driven systems with discussions on AI ethics and expectations. Artificial Intelligence techniques, including ML and DL enhance gesture and speech recognition in HCI, while ANN models are particularly effective for hand gesture recognition. In Human-Computer Interaction, AI techniques bring value and understanding to Healthcare in real life. Despite the benefits that Artificial Intelligence brings to HCI, challenges remain. The future promises new applications and perspectives in HCI, where AI with ML, DL, and ANN have an effective impact.

Addressing the anti-counterfeiting and verification issues of handwritten signatures, a traceable offline handwritten signature anti-counterfeiting and verification system is proposed. The system consists of four parts: a front-end mobile terminal application system, a database storage system, a backend management system, and a portable dynamic encoding stamp device. The system can achieve traceability of signatures, thereby more effectively preventing counterfeit signatures. The front-end mobile terminal application system can generate a unique two-dimensional code based on the relevant information of the signature file and send it to the database system for storage, while also sending it to the portable dynamic encoding stamp device. The portable dynamic encoding stamp device can receive the dynamic two-dimensional code from the front-end mobile terminal application system and produce a two-dimensional code imprint. The database system can store information such as signature file details, corresponding QR codes, and signature photos, which can be used for traceability queries and verification. The proposed system utilizes the SHA algorithm to generate unique codes and can generate corresponding QR code prints on a portable dynamic encoding device. When signing, a QR code seal is affixed, and during signature verification, the QR code on the signature file can be compared with the information in the database to verify the authenticity of the signature. When needed, signature photos stored in the database can also be used for further verification through machine learning. Compared with traditional signature verification methods, this method has better effectiveness.

Given the increasing demand for laboratory equipment management in universities, especially the increasingly complex equipment management, the traditional equipment management system can no longer meet the management needs of universities. Therefore, it is very important to optimize the university’s equipment management system. The allocation of laboratory equipment maintenance tasks in the laboratory equipment management system of universities is a very critical link. Its effective solution is crucial to ensure the normal operation of laboratory equipment and the reasonable allocation of maintenance resources. This study proposes a double coding adaptive genetic algorithm to optimize the allocation of laboratory equipment maintenance tasks in universities to achieve the optimal allocation of resources and minimize maintenance costs. The work allocation scheme is iteratively optimized by a dual-coding strategy and definition of adaptive crossover and mutation operators. The experimental results of this study show that the algorithm can find the approximate optimal task allocation scheme within a reasonable time, which improves the efficiency and accuracy of laboratory equipment maintenance. In addition, compared with the traditional allocation method, the algorithm in this paper shows stronger flexibility and robustness when dealing with large-scale complex problems.

ReLU nodes are utilized commonly in neural networks as they look and act like linear functions while providing nonlinearity. In spite of addressing the vanishing gradient problem, they can lead to the dying ReLU problem which can be detrimental in terms of convergence and generalization performance. This paper proposes antimatter networks, a new and simple solution to the dying ReLU problem which involves combining ReLU nodes with their inverse, negative ReLU nodes, together with activation swapping mechanisms. We tested the solution on six separate dataset-architecture combinations with the MNIST, CIFAR-10, and Flowers-16 datasets for Convolutional Neural Networks (CNN) and Multi-Layer Perceptron networks (MLP) and found that antimatter networks lead to consistent convergence and generalization improvements compared to networks solely consisting of ReLU or NReLU nodes.

Surface defect detection plays a pivotal role in ensuring product quality in industrial production, as defects like cracks, scratches, and dents can compromise product performance and durability. Traditional detection methods, such as manual inspection and Non-Destructive Testing (NDT), are limited by inefficiency, reliance on human expertise, and susceptibility to errors, which restrict their application in large-scale production. With advancements in artificial intelligence, deep learning models, particularly Convolutional Neural Networks (CNN) and Recurrent Neural Networks (RNN), have emerged as promising solutions for automated surface defect detection. This paper provides a comprehensive review of surface defect detection technologies, starting from traditional methods to modern deep learning-based techniques. The advantages and limitations of each approach are analyzed, highlighting key advancements in deep learning, including recent models like Faster R-CNN, Cascade R-CNN, and YOLOv4. Furthermore, challenges such as handling complex defects and improving detection accuracy in real-world industrial environments are discussed, along with potential directions for future research. Experimental evaluations using the Few Steels Classification (FSC) dataset demonstrate the effectiveness of modern detection methods in industrial applications, offering insights into enhancing defect detection systems.

This study explores sentiment analysis research framework based on the Mandarin social media dataset, focusing on the Transformer model. The paper first reviews the background of sentiment analysis, emphasizing the importance of this task in text classification and the relative lack of sentiment analysis in Mandarin Chinese. The study used three public datasets from the CSDN platform, including positive and negative reviews from different social media platform, explores how word order impacts sentiment classification in Mandarin. The study completed the experiment through four stages: preprocessing, text embedding, feature extraction, and sentiment classification, and used a pre-trained Transformer model for analysis. The results demonstrate the effectiveness of Transformer models for sentiment analysis with high accuracy on certain datasets, although challenges persist with specific data due to complexity. Future work aims to refine the model’s performance on diverse datasets and address limitations in sentiment feature extraction for Mandarin texts. The results confirm that there is still much room for improvement in Transformer models in improving sentiment classification in non-English languages such as Mandarin.

Conveying equipment is an important equipment for conveying bulk cargo materials in coal terminals, which has the characteristics of large volume, high efficiency, and low operating cost. The fault detection and disposal of conveying equipment during operation is an important guarantee for the normal operation of the terminal. This paper focuses on analyzing the types of faults such as conveyor belt deviation, slippage, longitudinal tearing, overload, spreading and stacking, abnormal temperature, and fire, puts forward the functional requirements of the safety intelligent inspection system of conveying equipment, and studies the composition of the safety intelligent inspection system of conveying equipment and the functions of its subsystems. The intelligent inspection system has been applied to the coal transportation system of Yulong Petrochemical Terminal to realize all-weather unmanned inspection operations during coal transportation and production operations, improve the safety management of coal transportation equipment, and strongly support the construction of intelligent port and safe port.

Current noise cancellation methods can barely deal with nonlinear signals in an effective way and the overall noise reduction performance is relatively poor. Aiming at these issues, we propose an adaptive active noise cancellation (ANC) method of high voltage reactors based on the denoising convolutional neural network (DnCNN). First, a feedforward adaptive ANC system was designed by considering the noise characteristics of high voltage reactors. Then, a noise control model was constructed and the weight coefficient of the filter was optimized by using the DnCNN network; meanwhile, the inverse waveform was reconstructed by adding spectrum analysis, thereby achieving effective control of noise signals. Additionally, the whale optimization algorithm was employed to solve the problem models of minimization of total radiation power of different sound sources, and the optimized location, amplitude, and phase angle of the secondary sound sources were determined. Based on LabVIEW, an online noise monitoring and ANC system was developed and used for experimental analysis of the proposed method. The results showed that the noise reduction performance was optimized (the overall noise reduction = 9 dB) when the number of secondary sound sources was 8 and the distance from the primary sound source was 1.0 m, suggesting that the proposed method can reliably reduce the noises of high voltage reactors.

Substations are subjected to multiple sound sources and complex environments and it is difficult to accurately extract the noises of high voltage reactors. In this study, we propose a noise extraction and evaluation method based on a multi-sound transmission array and deep learning for high voltage reactors. Firstly, the noise source was positioned on the basis of the reactor noise model by using the beamforming method, and the signal compensation was executed by combining the frequency domain-focused beamforming algorithm to guarantee high extraction accuracy of noise features. Then, echo was eliminated by using the improved local cepstrum, and the Deep Belief Network(DBN) evaluation model was improved by using genetic algorithms. Additionally, the processed sound pressure level signals were input into the improved DBN model to obtain the sound pressure level of the noises of high voltage reactors and achieve level evaluation. The proposed method was demonstrated experimentally by using the monitoring data of high voltage reactors in a 220 kV substation. The results showed that the proposed method can accurately locate the noise source, and the average percentage errors of the noise evaluation results in the daytime and at night were 2.7% and 3.5% respectively, both of which meet the requirements of practical application.

At present the world has entered the era of digital economy, digital technologies have already penetrated in people’s daily life and business’s activities. With the development of global economic integration and information technology, cross-border e-commerce industry has developed rapidly and has become a new form of Chia’s foreign trade. And at the same time cross-border e-commerce, as one of the digital manifestations of international trade, has been deeply impacted by digital technologies. This paper reviews of the present development status of China’s cross-border e-commerce in the era of digital economy, and also elaborates some of the characteristics of cross-border e-commerce in China. Given the present international political and economic situation the paper also points out the problems and challenges faced by China’s cross-border e-commerce. In an attempt to address those problems this paper proposes suggestions and countermeasures from such perspectives as technological innovation, legal legislation, government supervision.

Social condition, accompanied by booming economy and technology, has become increasingly complex, gradually increasing pressure on public security. Therefore, enhancing public security patrol performance has become important means for ensuring the security of lives and property of people. This study aimed to select a combination of public security patrol and three-dimensional (3D) path planning as the breakthrough point to explore the operating current condition and existing problems of the combination. 3D path planning under a simulation environment was implemented using bidirectional A*, rapidly exploring random tree (RRT) Extend, and RRT-Connect algorithms. Finally, the prospects of the combination of police patrol and 3D path planning were discussed to provide a new thought for the future development of police patrol.

The work presents the investigation of characteristics of the cavitation flow in a centrifugal pump using a combination of experimental and computational studies. In experimental studies, the head drop curve and pressure signals were dynamically measured. The numerical computations were implemented based on Zwart mass transfer cavitation model and a modified RNG k-ε turbulence model. The computational head drop curve showed good agreement with the experimental results. The computational results and experimental results revealed that the cavitation of a centrifugal pump could be divided into four types: non-cavitation, incipient cavitation, leading edge cavitation, and cloud cavitation. The main frequency of pressure in the pump of four cavitation types is the blade passing frequency (f_b). The larger the cavity is, the greater the pressure amplitude at the blade passing frequency (f_b). For the cloud cavitation, the cavity bubbles grow and collapse continuously at the edge of the cavity clouds, which led to a significant increase in the pulse quantity of pressure derivation.

With the rapid development of information technology, artificial intelligence technology gradually penetrates the field of education, providing new possibilities for the innovation of teaching mode. The traditional teaching mode often has problems such as single teaching resources, low student participation and insufficient cultivation of practical ability. To solve these problems, this paper proposes a practical teaching mode based on the concept of Human-Machine collaboration. It aims to harness the advantages of human and machine and improve teaching efficiency and quality. Taking the course “Python and Web Crawler” as an example, a teaching framework includes four modules: knowledge learning, practical training, evaluation feedback and personalized tutoring. An AI-assisted teaching system integrating the functions of knowledge base, adaptive learning, automatic evaluation, virtual practice and teacher assistance is also developed. The application of this model in practical teaching shows that the Human-Machine collaborative teaching model is effective. According to the experimental results of this research, firstly, the average learning time of students this semester has decreased by about 20% compared to the previous semester, but the average score of the final grade has improved by about 5%. Secondly, the students’ project works are better than in the previous semester in terms of code quality and functional completion, especially the improvement in code quality is the most obvious. Finally, based on ten questions, the overall evaluation of the students on the course this semester is significantly higher than that of the last semester. Therefore, we firmly believe that this research provides a useful exploration for the application of AI in education.