Intelligent Systems with Applications最新文献

{"title":"An appraisal of backscatter removal and refraction calibration models for improving the performance of vision-based mapping and navigation in shallow underwater environments","authors":"Fickrie Muhammad ,&nbsp;Poerbandono ,&nbsp;Harald Sternberg ,&nbsp;Eka Djunarsjah ,&nbsp;Hasanuddin Z Abidin","doi":"10.1016/j.iswa.2025.200476","DOIUrl":"10.1016/j.iswa.2025.200476","url":null,"abstract":"<div><div>Vision-based mapping (VbM) is one of the fundamental origins of automation in remote and autonomous spatial data acquisitions. Complexity in obtaining accurate data arises when such a method is applied in the underwater environment. Non-uniform illumination and refraction distortion are the most common problems encountered in underwater VbM. This study addresses this by employing backscatter removal to enhance image clarity and a pinhole-axial (Pinax) camera model to adjust the refraction distortion. In particular, the methods are computed in the robot operating system (ROS), publishing the enhanced images as separated image nodes in real-time and enabling seamless integration to the VbM pipeline. It is argued that the proposed VbM-dedicated models can significantly improve the feature detection method and conformity of object positions underwater around the camera's motion. Simulation datasets are generated to evaluate the sensitivity to varying turbidity levels to test the method's sensitivity. Additionally, field experiments with GoPro 10 hardware in Pramuka Island Waters, Indonesia, offer real-world context for the study's relevance to distinct underwater circumstances. Furthermore, additional visual-inertial datasets quantify the overall performance, especially in retrieving metric positioning information. The research shows efficient backscatter removal improves feature detection robustness, especially in murky water conditions. Refraction correction eliminates the bowing effect from missing ground control points in underwater environments. The study is significant because it emphasizes how vital image enhancement and refraction calibration are to obtaining &lt;4 % trajectory error of VbM. Overall, the proposed VbM pipeline can maintain &lt;5 cm trajectory error compared to the standard VbM pipeline. The results highlight the need for a comprehensive strategy to advance underwater mapping and navigation technology to deliver accurate and dependable outcomes in various underwater situations.</div></div>","PeriodicalId":100684,"journal":{"name":"Intelligent Systems with Applications","volume":"25 ","pages":"Article 200476"},"PeriodicalIF":0.0,"publicationDate":"2025-01-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143136761","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Robust vector-weighted and matrix-weighted multi-view hard c-means clustering","authors":"Zhe Liu ,&nbsp;Sarah Aljohani ,&nbsp;Sijia Zhu ,&nbsp;Tapan Senapati ,&nbsp;Gözde Ulutagay ,&nbsp;Salma Haque ,&nbsp;Nabil Mlaiki","doi":"10.1016/j.iswa.2024.200470","DOIUrl":"10.1016/j.iswa.2024.200470","url":null,"abstract":"<div><div>With the rapid advancement of information technology, multi-view data has become ubiquitous, prompting extensive attention towards multi-view clustering algorithms. Despite significant strides, several challenges persist: (1) the prevalence of noise and outliers in real-world multi-view data often compromises the efficacy of clustering; (2) most existing multi-view clustering algorithms predominantly assess the overall contribution of each view, while neglecting the intra-view contributions. In this paper, we first propose a robust vector-weighted multi-view hard <span><math><mi>c</mi></math></span>-means (VW-MVHCM) clustering, drawing inspiration from the single-view alternative hard <span><math><mi>c</mi></math></span>-means. A distinctive feature of VW-MVHCM is the substitution of the conventional Euclidean norm with a non-Euclidean norm metric, enhancing its resilience to noise and outliers. Additionally, we introduce view weights to learn the contribution of each view in clustering. On this basis, we further propose a robust matrix-weighted multi-view hard <span><math><mi>c</mi></math></span>-means (MW-MVHCM) clustering, which assigns view-specific weights at the cluster level, allowing for more detailed intra-view contribution modeling. This matrix-weighted approach enables MW-MVHCM to dynamically capture the varying importance of each view across clusters, improving clustering performance. We design an optimization scheme to obtain the optimal results of VW-MVHCM and MW-MVHCM. Experimental results on benchmark datasets demonstrate that our proposed algorithms outperform existing multi-view clustering algorithms, showcasing their robustness and effectiveness in real-world scenarios.</div></div>","PeriodicalId":100684,"journal":{"name":"Intelligent Systems with Applications","volume":"25 ","pages":"Article 200470"},"PeriodicalIF":0.0,"publicationDate":"2025-01-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143136007","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Virtualization resource scheduling and optimization method based on swarm intelligent systems","authors":"Jun Zhao","doi":"10.1016/j.iswa.2024.200469","DOIUrl":"10.1016/j.iswa.2024.200469","url":null,"abstract":"<div><div>Efficient scheduling of virtualized resources can not only meet the service needs of users, but also achieve the optimal allocation of resources and the stable operation of the system. However, due to the dynamic and diversity of virtualized resources, the traditional scheduling methods have been difficult to meet the actual needs. Therefore, a virtual resource scheduling and optimization method based on Swarm Intelligence System (SIS) is proposed in this paper. The core idea of this method is to transform the Virtualized Resource Scheduling (VRS) problem into a multi-objective optimization problem, and use the particle swarm optimization algorithm of SIS to search for the optimal solution. By updating the speed and position of the particles, the scheduling scheme is optimized iteratively to maximize the utilization of resources and optimize the performance of the system. The experimental results show that the SIS-based virtual resource scheduling method can significantly improve the resource utilization and system performance while meeting the needs of users. Compared with other scheduling methods, this method has better adaptability and robustness, and provides a new solution for virtualization resource scheduling in cloud computing environment.</div></div>","PeriodicalId":100684,"journal":{"name":"Intelligent Systems with Applications","volume":"25 ","pages":"Article 200469"},"PeriodicalIF":0.0,"publicationDate":"2024-12-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143136754","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A hybrid machine learning framework by incorporating categorical boosting and manifold learning for financial analysis","authors":"Yuyang Zhao ,&nbsp;Hongbo Zhao","doi":"10.1016/j.iswa.2024.200473","DOIUrl":"10.1016/j.iswa.2024.200473","url":null,"abstract":"<div><div>The financial analysis is essential to evaluate and assess the financial behavior and risk during the financial activities. However, it is challenging to implement the financial analysis due to the complexity of financial features and their interaction mechanism. This study developed a hybrid machine-learning framework incorporating categorical boosting (CatBoost) and manifold learning for financial analysis. CatBoost was employed to capture the financial mechanism and characterize the complex and nonlinear relationship between the financial feature and the associated financial behavior. Manifold learning was utilized to select and extract the critical financial features. The developed framework was verified and illustrated by the synthetic datasets, which are based on the financial model for the loan evaluation. The overall accuracy of the CatBoost model increased from 81.5 % to 99.1 %, and the accuracy for predicting unapproved loans increased from 64 % to 98.88 %. The developed framework significantly improves the prediction accuracy of loan-approved status and characterizes the financial behavior and mechanism well. The developed hybrid framework distinguishes between various financial features and the associated loan-approved status. Based on the developed framework, it also found that credit score and annual income are the two essential features, and the contribution of other features is almost negligible. The developed framework revealed that a credit score of 500 and an annual income of 70,000 are critical thresholds for loan approval, as set by the financial analysis model used to generate the dataset. The results show that the developed framework could extract the financial features and capture the financial mechanism during the financial analysis. It provides a scientific, reasonable, and promising approach to financial analysis and understanding financial behavior.</div></div>","PeriodicalId":100684,"journal":{"name":"Intelligent Systems with Applications","volume":"25 ","pages":"Article 200473"},"PeriodicalIF":0.0,"publicationDate":"2024-12-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143136303","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Detecting new obfuscated malware variants: A lightweight and interpretable machine learning approach","authors":"Oladipo A. Madamidola,&nbsp;Felix Ngobigha,&nbsp;Adnane Ez-zizi","doi":"10.1016/j.iswa.2024.200472","DOIUrl":"10.1016/j.iswa.2024.200472","url":null,"abstract":"<div><div>Machine learning has been successfully applied in developing malware detection systems, with a primary focus on accuracy, and increasing attention to reducing computational overhead and improving model interpretability. However, an important question remains underexplored: How well can machine learning-based models detect entirely new forms of malware not present in the training data? In this study, we present a machine learning-based system for detecting obfuscated malware that is not only highly accurate, lightweight and interpretable, but also capable of successfully adapting to new types of malware attacks. Our system is capable of detecting 15 malware subtypes despite being exclusively trained on one malware subtype, namely the Transponder from the Spyware family. This system was built after training 15 distinct random forest-based models, each on a different malware subtype from the CIC-MalMem-2022 dataset. These models were evaluated against the entire range of malware subtypes, including all unseen malware subtypes. To maintain the system's streamlined nature, training was confined to the top five most important features, which also enhanced interpretability. The Transponder-focused model exhibited high accuracy, exceeding 99.8%, with an average processing speed of 5.7 µs per file. We also illustrate how the Shapley additive explanations technique can facilitate the interpretation of the model predictions. Our research contributes to advancing malware detection methodologies, pioneering the feasibility of detecting obfuscated malware by exclusively training a model on a single or a few carefully selected malware subtype and applying it to detect unseen subtypes.</div></div>","PeriodicalId":100684,"journal":{"name":"Intelligent Systems with Applications","volume":"25 ","pages":"Article 200472"},"PeriodicalIF":0.0,"publicationDate":"2024-12-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143136756","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Estimation of the compressive strength of ultrahigh performance concrete using machine learning models","authors":"Rakesh Kumar ,&nbsp;Divesh Ranjan Kumar ,&nbsp;Warit Wipulanusat ,&nbsp;Chanachai Thongchom ,&nbsp;Pijush Samui ,&nbsp;Baboo Rai","doi":"10.1016/j.iswa.2024.200471","DOIUrl":"10.1016/j.iswa.2024.200471","url":null,"abstract":"<div><div>The compressive strength of ultrahigh performance concrete (UHPC) is influenced by the composition, quality, and quantity of its constituent elements. Using traditional statistical methods, it is difficult for us to quantify the relationships between the technical properties of UHPC and the composition of the mixture because of their complexity and nonlinearity. This work aims to develop advanced prediction models for estimating UHPC compressive strength over a large spectrum of supplementary cementitious material combinations and aggregate sizes. The models trained on the UHPC mixture dataset with 15 input variables included the group method of data handling, recurrent neural networks, long short-term memory, and bidirectional long short-term memory (Bi-LSTM). These models routinely forecast UHPC compressive strength according to sensitivity analysis, external validation, and statistical performance measures. During testing, the Bi-LSTM model outperformed the other models, with an RMSE of 0.0482 and an R² value of 0.9464. These results maximize component selection by showing how effectively the Bi-LSTM model might reduce UHPC formulation development and lower the cost and testing time span.</div></div>","PeriodicalId":100684,"journal":{"name":"Intelligent Systems with Applications","volume":"25 ","pages":"Article 200471"},"PeriodicalIF":0.0,"publicationDate":"2024-12-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143136302","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Unsupervised domain adaptation with self-training for weed segmentation","authors":"Yingchao Huang ,&nbsp;Amina E. Hussein ,&nbsp;Xin Wang ,&nbsp;Abdul Bais ,&nbsp;Shanshan Yao ,&nbsp;Tanis Wilder","doi":"10.1016/j.iswa.2024.200468","DOIUrl":"10.1016/j.iswa.2024.200468","url":null,"abstract":"<div><div>Accurate crop and weed segmentation in varied field conditions is crucial for advancing automated weed management but remains challenging. Though promising, convolutional neural networks (CNNs) often experience performance drops when deployed in new field environments due to shifts between training and test data distributions. To address this limitation, we proposed a self-training framework using a teacher–student model that adapts CNNs for diverse agricultural contexts. Our method enhances generalization by co-training the student model on both the source domain and pseudo-labelled target domain generated by the teacher model, with teacher parameters updated via an exponential moving average of the student’s model. The main contributions of this work are as follows: (1) we simplified the self-training procedure by using all target predictions, skipping the selection phase, and applying local dynamic weights (LDW) for target pixels during co-training; (2) we optimized iteration by monitoring covariance fluctuations to avoid pseudo-label overfitting and reduced the impact of false labels; (3) we addressed class imbalance with dynamic class weights (DCW) to give more importance to minority classes; and (4) we formulated a loss function integrating both LDW and DCW into the soft intersection over union (softIoU), enhancing weed segmentation effectiveness. We evaluated our framework with the ROSE challenge dataset across eight adaptations involving varied plants, robots, and growth stages, achieving up to a 0.17 mean IoU improvement over popular methods like CycleGAN. Our approach demonstrated consistent performance across diverse agricultural environments, supporting its use in real-field inference.</div></div>","PeriodicalId":100684,"journal":{"name":"Intelligent Systems with Applications","volume":"25 ","pages":"Article 200468"},"PeriodicalIF":0.0,"publicationDate":"2024-12-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143136296","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Artificial intelligence in digital image processing: A bibliometric analysis","authors":"Erwin Yudi Hidayat ,&nbsp;Khafiizh Hastuti ,&nbsp;Azah Kamilah Muda","doi":"10.1016/j.iswa.2024.200466","DOIUrl":"10.1016/j.iswa.2024.200466","url":null,"abstract":"<div><div>This study presents a bibliometric analysis of artificial intelligence (AI) in digital image processing (DIP), analyzing 1063 publications from the Scopus database from 1998 to 2023. The field has seen significant growth, with an average annual growth rate of 16.24%, accelerating sharply between 2020 and 2023. The analysis emphasizes AI’s growing influence in healthcare and real-time image processing. China leads in publication volume, while the USA dominates in citation impact, underscoring the global and collaborative nature of AI-DIP research. Key institutions like the University of California and Tsinghua University, along with authors such as U. Rajendra Acharya, have made significant contributions to AI-driven healthcare diagnostics, highlighting the importance of interdisciplinary collaboration. High-impact journals, including IEEE Transactions on Medical Imaging, play a crucial role in advancing the field. However, this study relied on a targeted keyword search in Scopus, which may not capture all relevant research, particularly those using alternative terminologies or broader AI classifications. Additionally, challenges related to data privacy, bias, and transparency persist. Addressing these issues will be critical for the responsible development and application of AI-DIP technologies. This study offers valuable insights for future research and highlights key areas for continued exploration.</div></div>","PeriodicalId":100684,"journal":{"name":"Intelligent Systems with Applications","volume":"25 ","pages":"Article 200466"},"PeriodicalIF":0.0,"publicationDate":"2024-12-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143136295","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Hidden-layer configurations in reinforcement learning models for stock portfolio optimization","authors":"Patrick Kevin Aritonang,&nbsp;Sudarso Kaderi Wiryono,&nbsp;Taufik Faturohman","doi":"10.1016/j.iswa.2024.200467","DOIUrl":"10.1016/j.iswa.2024.200467","url":null,"abstract":"<div><div>In the rapidly evolving field of artificial intelligence and financial markets, efficient and adaptive portfolio management strategies are becoming increasingly critical. This study explores the impact of hidden-layer configurations in reinforcement learning models for stock portfolio optimization. Using a portfolio of 45 actively traded stocks in the Indonesian stock market, the performance of four reinforcement learning algorithms—Advantage Actor-Critic (A2C), Deep Deterministic Policy Gradient (DDPG), Proximal Policy Optimization (PPO), and Twin Delayed Deep Deterministic Policy Gradient (TD3)—is evaluated with zero, one, and two hidden layers. The results show that A2C and DDPG models perform effectively without hidden layers, with A2C achieving the highest Cumulative Return (CuR) and Annualized Return Rate (ARR) among all configurations. Adding hidden layers to A2C improved risk management, resulting in a lower Maximum Drawdown (MDD) and a higher Annualized Sharpe Ratio (ASR). DDPG exhibited consistently strong performance, with its zero hidden-layer model showing the highest ASR. Conversely, PPO underperformed across all configurations, with negative returns in the zero-layer setup and marginal improvements with added complexity. Introducing additional hidden layers improved TD3′s performance, enhancing risk-adjusted returns. These findings suggest that the effectiveness of hidden-layer configurations depends on the specific algorithm used. While A2C and DDPG benefit from increased complexity, simpler architectures may be more suitable for PPO and TD3. This study offers new insights into optimizing reinforcement learning models for stock portfolio management by adjusting hidden-layer structures to balance returns and risk.</div></div>","PeriodicalId":100684,"journal":{"name":"Intelligent Systems with Applications","volume":"25 ","pages":"Article 200467"},"PeriodicalIF":0.0,"publicationDate":"2024-12-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143136299","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Detecting unknown intrusions from large heterogeneous data through ensemble learning","authors":"Farah Jemili,&nbsp;Khaled Jouini,&nbsp;Ouajdi Korbaa","doi":"10.1016/j.iswa.2024.200465","DOIUrl":"10.1016/j.iswa.2024.200465","url":null,"abstract":"<div><div>The rapid expansion of data volumes, technological advancements, and the emergence of the Internet of Things (IoT) have heightened concerns regarding the detection of unknown intrusions based on singular sources of network traffic. This progression has led to the generation of vast and diverse datasets originating from various sources including IoT devices, web applications, and web services. Effectively discerning attacks within such a heterogeneous network traffic landscape necessitates the identification of underlying security behaviors, essential for developing an efficient analysis information system.</div><div>This paper aims to establish a comprehensive framework for network intrusion detection. The proposed methodology involves the synthesis of network features into a universal security database through the utilization of Term Frequency-Inverse Document Frequency Terms (TF-IDF) and semantic Cosine similarity. By amalgamating a diverse array of data flows, a set of universal features is generated, facilitating storage within the newly devised universal representation. Subsequently, Principal Component Analysis (PCA) is employed to reduce the dimensionality of the extensive universal security database while preserving essential information. Leveraging Ensemble Learning, a novel method is introduced for the detection of unknown attacks.</div><div>The efficacy of the developed database is evaluated using various Machine Learning algorithms, including Naïve Bayes, K-Nearest Neighbor, Logistic Regression, Decision Tree, and Random Forest. Furthermore, Ensemble Learning methods are assessed under two distinct scenarios. Experimental findings, conducted on datasets such as CICIDS 2017, NSL-KDD, and UNSW, demonstrate the universality, versatility, and effectiveness of the proposed approach, particularly in accommodating datasets with diverse structures.</div></div>","PeriodicalId":100684,"journal":{"name":"Intelligent Systems with Applications","volume":"25 ","pages":"Article 200465"},"PeriodicalIF":0.0,"publicationDate":"2024-12-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143136297","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}