https://scj.kashanu.ac.ir/ Soft Computing Journal en daily 1 Fri, 22 May 2026 00:00:00 +0330 Fri, 22 May 2026 00:00:00 +0330 https://scj.kashanu.ac.ir/article_113810.html Software comprehension plays an important role during its improvement and maintenance process. Software modularization is a key activity for recovering the software architecture, which improves software understanding. Since the software modularization problem is NP-hard, meta-heuristics such as evolutionary algorithms (EAs) are usually used to solve it. EAs are problem-dependent, and they also require considerable space and time. Recently, the use of hyper-heuristic approaches growing to obtain more generality. This paper proposes an iterated local search (ILS) strengthened by a Q-learning-based hyper-heuristic for software modularization that overcomes the limitations of EAs.  In the proposed algorithm, two main components of ILS, i.e., perturbation and local search components, are intelligently selected using a Q-learning-based hyper-heuristic in each iteration. The performance of the proposed algorithm is evaluated on eleven real-world software systems of small and medium sizes. The results of the experiments demonstrate that the proposed ILS produces modularizations that have higher or equal quality compared to the quality of the modularizations obtained by selected algorithms. https://scj.kashanu.ac.ir/article_114251.html Cloud computing involves a variety of technologies, including networking and virtualization, to meet the new needs of users, but is vulnerable to many security threats. To provide the necessary level of security in cloud computing, decision-making on the type and number of security modules used by cloud service users and then paying the relevant fees is of particular importance. Game theory, with the ability to model the behavior of users and attackers of a supervisor and analyze the possible strategy and profitability of each, can suggest a suitable strategy for investing in the security modules of a virtual machine. In our previous work, we used game theory to analyze the decision to invest in one of the security modules for each of the actors. The purpose of this paper is to study the effect of the three parameters "different investment costs", "probability of success of the attack on the user" and "probability of success of the attack on the supervisor" and to make an appropriate decision in this situation. Based on the simulation results, it can be said that given the different values of the probability of a successful attack on a supervisor, a predetermined investment can lead to a proper Nash equilibrium. In general, at low costs or in the case of increasing the cost of investing in security, the user tends to constantly change his strategy and provide the desired security conditions. The results also show that as the probability of a successful attack on a user not investing in security increases, so does the security investment cost. https://scj.kashanu.ac.ir/article_112800.html Edge computing is presented as a new model with a focus on near-resource data processing and processing to address Internet of Things(IoT) needs and localize computing needs, increase power for emergency response time, increase scalability and reduce energy costs, and control privacy and data protection. One of the main challenges in using edge computing is the quality of data obtained from multiple sources. Different sources in the vast and heterogeneous IoT environment receive and send inconsistent and conflicting data from the same phenomenon. This creates a strong need to identify and correct the collected data. In this research, a two-step approach for identifying and correcting the data of each source and then identifying the conflicts between the data of different data sources and merging the sources and the final correction of the data is presented. In the first step, the identification and correction of defective information are performed based on the confidence interval and estimated data. The second step is to measure the conflict and fusion of data, which is created to calculate the degree of conflict in different data sources based on fuzzy measures and calculate the degree of validity of each data source. The proposed approach provides good results on the various types of data conflicts. Based on the simulation results with accuracy, sensitivity, specificity, and F-score criteria, the proposed approach has a good performance and in all conflicts, accuracy, and sensitivity show more than 75%, specificity more than 72%, and F-score criterion more than 73%. https://scj.kashanu.ac.ir/article_114258.html Anomaly detection in multivariate time series has been an active research area due to its widespread application in various fields. Window-based methods are popular in the anomaly detection domain. These methods identify anomalous windows rather than specific anomalous points, even if not all points within the window are anomalies. It is a critical limitation of window-based methods. We propose an unsupervised sliding window-based model for detecting anomalies in multivariate time series to address this limitation. Our model employs a sliding mechanism to iterate through the input time series multiple times and utilizes a consensus function to aggregate different window anomaly scores. This mechanism facilitates the discovery of more anomalous subsequences, even if they are not precisely confined within a specific window. To evaluate the performance of the proposed method, several experiments on synthetic and real-world datasets, including SKAB and MSL, with multiple indices. The results confirm the superiority of the proposed method. The method achieves an Fscore of 0.902 for SKAB and 0.620 for MSL, which are twice as good as the results achieved by other methods. https://scj.kashanu.ac.ir/article_114266.html A neurodevelopmental disorder recognized by insufficiency in social communication and repetitive behaviours is expressed as an autism spectrum disorder (ASD). One of the most useful tools for diagnosing autism is the use of electroencephalography (EEG) signals because these signals accurately represent the brain's function. The recorded EEG of each person contains a lot of information that is very difficult to study and check visually. The main goal of machine learning algorithms is to train the machine in such a way that it finally has a diagnosis close to that of the human brain. This paper evaluates the appropriate strategies for further exploitation of deep learning capabilities in the feature extraction block of autism diagnosis without using classical feature extraction methods. In this research, a convolutional neural network (CNN) structure is used to check the available data in order to extract features. Classification has been done with five machine learning classifiers, covering support vector machine (SVM), linear discriminant analysis (LDA), decision tree (DT), simple Bayes classification (GNB), and random forest (RF). The accuracy obtained from the use of classifiers with SVM methods is 100%, LDA 82%, DT 80.5%, GNB 100%, and RF 100%. The proposed idea of convolutional neural networks for feature extraction and classification with different machine learning methods has provided high-accuracy results that are equal to the other amazing methods for autism diagnosis. https://scj.kashanu.ac.ir/article_114424.html In this paper, a classification model based on ensemble approach is used for credit scoring of the bank costumers. The proposed method is based on the bagging scheme for the support vector machines classifier. First, the data set is divided into several subsets via the bootstrap method and the support vector machines classifier is implemented on each subsets. Then the final model is made by voting among all of the classifiers. The proposed method has many advantages for implementation, including reduction of computational costs. Two credit data sets are used to show the efficiency and applicability of the present method. https://scj.kashanu.ac.ir/article_114425.html In this paper, by consideringthe notion of ideals in BL-algebras,we introduce some special ideals whichare related to a subset of a BL-algebraand derive some new relations and resultsabout them. We also define theconcepts of ⊙-derivation for BL-algebrasand obtain some related results. Finally,we investigate the connection betweenthese functions and BL-algebras https://scj.kashanu.ac.ir/article_114426.html Attention deficit hyperactivity disorder is a neurodevelopmental disorder that typically begins in early childhood and poses significant challenges during school years. This disorder is characterized by impulsive behaviors, inattention, and difficulties with concentration. Early diagnosis and prompt treatment can effectively manage this condition. Accurate diagnosis of ADHD can be achieved through the precise analysis of electroencephalography signals. This article proposes a brain modeling approach using a cellular neural network in various frequency bands to diagnose ADHD. Firstly, the inter-area connections in the brains of individuals with hyperactivity are estimated by assessing the spectral coherence function between channels. Subsequently, the intra-area connections are obtained using a cellular neural network. The results obtained indicate that the intra and inter-area connections in the central, frontal, and parietal regions of the brains of individuals with hyperactivity differ from those of normal individuals in the beta and gamma frequency bands. Consequently, it can be inferred that the presence of disparities in intra and inter-area connections between the brains of individuals with ADHD and normal individuals results in distinct brain functionality within these two groups. https://scj.kashanu.ac.ir/article_114444.html This work presents a complete all-graphene SRAM chip design. The SRAM requires analog and digital sub-circuits, each having different design criteria. On the other hand, the electrical parameters of a GNRFET device are strongly related to geometry. In this study, we built a complete graphene-based SRAM chip and then proposed a new approach to optimize the GNRFET’s physical design which fulfills SRAM requirements for HOLD, READ, and WRITE operations. The effect of geometric and process parameters such as chirality, channel length, and width are investigated on the characteristics of an SRAM cell based on GNRFET. Analysis of power consumption, delay, and SNM results, indicate that adjustable parameters of GNRFETs can have significant effects on SRAM cell performance, and our approach is very effective in parameter optimization. Using optimized GNRFETs, a full-circuit SRAM chip is designed and analyzed. The noise margin test of the SRAM cell shows 188mV HSNM, and 240mV WSNM, while standby and leakage currents were 5, and 20 times smaller. https://scj.kashanu.ac.ir/article_114563.html In this article, the behavior of event-based and rule-based systems is modeled using Hierarchical Fuzzy Petri nets (HFPN). In such systems, a large number of rules with fuzzy variables can lead to increase complexity in deducing behavior. So far, several FPN methods have been presented for these systems. In this paper, we present an HFPN leading to a reduction in the number of arcs, places and transitions \hl{as well as a reduction in the ML language code on Petri-net arcs and the increase of the code constructiveness}. Finally, we applied our method for modeling and reasoning a secure water treatment system against burst pipe attack.In this article, the behavior of event-based and rule-based systems is modeled using Hierarchical Fuzzy Petri nets (HFPN). In such systems, a large number of rules with fuzzy variables can lead to increase complexity in deducing behavior. So far, several FPN methods have been presented for these systems. In this paper, we present an HFPN leading to a reduction in the number of arcs, places and transitions \hl{as well as a reduction in the ML language code on Petri-net arcs and the increase of the code constructiveness}. Finally, we applied our method for modeling and reasoning a secure water treatment system against burst pipe attack. https://scj.kashanu.ac.ir/article_114671.html In recent decades, with the advancement of information and communication technology, the need for organizing and optimizing informational and communicational processes among individuals and various resources has gained increased importance. One effective approach in this field is the use of recommender systems. These systems can significantly enhance user experience either offline, by preparing a list of the best recommendations, or online, in real-time. One of the challenges of recommender systems is achieving a balance between the speed of delivering recommendations and their quality. This paper presents the development and evaluation of a recommender algorithm named "Newpixie," which aims to provide recommendations based on common interests among users. The Newpixie algorithm utilizes a scalable, bipartite graph comprising pins and boards, where each board contains a number of related pins on a specific topic. In the offline phase of the algorithm, strong associations between pins are extracted using association rules in data mining and applied as virtual edges in the graph. In the online phase, biased random walks on the strong links between pairs of items within the boards are employed to generate suitable recommendations for users. The performance of the Newpixie algorithm is evaluated against the baseline pixie method based on three evaluation metrics. Experimental results indicate that the Newpixie algorithm improves recommendation speed by approximately 17%, recommendation quality by about 34%, and the recall@k metric in link prediction by roughly 20% compared to the baseline pixie method. https://scj.kashanu.ac.ir/article_114872.html The COVID-19 pandemic has underscored the critical challenges faced by healthcare systems worldwide, particularly in meeting the escalating demand for resources such as ICU beds, specialized care, and medical equipment. This shortfall has resulted in significant loss of life, highlighting the urgent need for accurate and timely diagnosis to optimize patient outcomes and reduce healthcare costs. In response to these challenges, our research focuses on developing a machine learning system capable of predicting whether patients will require ICU admission or can be managed remotely at home during peak periods of demand. Leveraging a novel two-dimensional reduction approach that combines evolutionary algorithms, Pattern Discovery, and machine learning techniques, we aim to streamline patient-collected data to train predictive models capable of forecasting ICU needs and remote care requirements. By providing healthcare systems with the ability to anticipate patient needs during critical phases of the pandemic, our predictive model empowers healthcare providers to allocate resources more effectively, optimize patient care delivery, and mitigate the impact of healthcare crises. The results of our experimental evaluation demonstrate the promising potential of our approach in addressing the pressing challenges posed by the COVID-19 pandemic and similar public health emergencies. https://scj.kashanu.ac.ir/article_114875.html With the widespread use of the internet and the development of wireless networks that transfer large data streams, the importance of assessing and controlling the quality of communication links in wireless networks has gained significant attention. By predicting link quality, energy consumption of network nodes and the overall stability of the network can be improved. One category of methods used for predicting the quality of wireless links is machine learning techniques. This paper examines the performance of ensemble methods, a type of supervised machine learning approach that has previously received less focus in the context of wireless link quality prediction. Additionally, due to the advantages of unsupervised methods that can be trained on unlabeled datasets, the performance of the k-means algorithm is also evaluated. The results show that ensemble algorithms are highly effective in predicting the quality of communication links in wireless networks. Among the ensemble methods, Gradient Boosting achieved the best performance with an F1 score of 95.79, while the k-means method demonstrated superior performance in the recall metric, achieving a value of 96.47 compared to other methods. https://scj.kashanu.ac.ir/article_114927.html Abstract: A brain-computer interface is a hardware and software communication system through which the user will be able to control computers and external devices using only their brain activities.Signal processing algorithm is the most important part of a brain-computer interface and includes the steps of data acquisition, preprocessing or signal amplification, feature extraction and classification.The aim of this research is to design the signal processing algorithm of a brain-computer interface and also to improve its performance using noise reduction methods.Considering the importance of feature extraction and classification steps, we must choose appropriate methods in these steps.First, brain-computer system, signal processing algorithm and human nervous system and brain, electroencephalogram signal have been investigated. Then the pre-processing step and noise reduction techniques, the feature extraction step and the classification step and different classifiers with their applications and characteristics have been introduced. Finally, a new method based on channel selection using the placement of electrodes has been presented, which reduces noise and significantly increases the performance of the algorithm, and the use of this method increases the accuracy of the system. https://scj.kashanu.ac.ir/article_115032.html This paper presents a numerical method for solving fuzzy delay nonlinear Volterra-Hammerstein integral equations using Legendre wavelets. The importance of this class of equations is highlighted by their application to modeling epidemic problems, which represent a special case. After introducing preliminary definitions related to fuzzy equations and the basic characteristics of Legendre wavelets, we present the parametric form of nonlinear fuzzy delay Volterra-Hammerstein integral equations, which are essentially a system of nonlinear delay integral equations in a non-fuzzy state. We then employ Legendre wavelets, the collocation method, and the Gauss-Legendre quadrature rule to transform the integral equation into a system of algebraic equations that can be solved. Furthermore, we provide a detailed convergence analysis of the proposed method. The accuracy of the method is demonstrated through several numerical examples, with results compared to those obtained using Bernoulli and B-spline wavelet methods. These comparisons confirm the accuracy and efficiency of the presented method. https://scj.kashanu.ac.ir/article_115033.html Analyzing user behavior on social networks to understand psychological traits is of growing importance. Structuring data extracted from users' concurrent activity across various platforms as a multiplex network can provide valuable insights. However, conventional approaches are limited by their reliance on explicit relationships, which are often private, and their use of single-source data. Therefore, this paper proposes a novel two-stage framework for personality type prediction. Its primary innovation is the use of an implicit relationship network, extracted from the fusion of multi-source, multi-modal data include text, image, and location, instead of relying on explicit friendship links. In the first stage, community structures in the multiplex network are identified using a multi-objective evolutionary approach (MOEA/D-TS), and then the implicit relationship network is constructed using a link prediction method (CLPES). In the second stage, the resulting network is leveraged as an information source within a machine learning framework to predict user personality types. Evaluations demonstrate the proposed framework's significant superiority. The final model, using a Random Forest classifier with the fusion of all three data sources, achieved an average Macro-F1 score of 0.673. This represents an average performance improvement of 23% compared to single-source models. This finding confirms the value of analyzing implicit relations and fusing multi-source data for a deeper understanding of user personality. https://scj.kashanu.ac.ir/article_115127.html Cancer remains a major challenge in modern medicine, and effective treatment requires accurate modeling of its complex processes. Various approaches have been developed to model cancer progression and treatment, including cellular automata, differential equations, and agent-based models. Differential equations are widely used for cancer growth and treatment response, but they often oversimplify biological complexities and face challenges in parameter estimation. Cellular automata help simulate cancer at the cellular level, though they may not fully capture biological mechanisms. Agent-based models, while insightful, demand significant computational resources. Game theory has emerged as a valuable tool for understanding strategic interactions between cancer cells and their environment, offering insights into tumor evolution and treatment resistance. By modeling cancer progression as an evolutionary competition among cell types, game theory-based models can predict cancer dynamics and help design treatment strategies that lead to better patient outcomes. This approach enhances the understanding of cancer progression and offers potential for creating more effective therapies by integrating experimental findings with mathematical modeling. https://scj.kashanu.ac.ir/article_115167.html In this paper, a Tabu Search-based algorithm is proposed to find the l-clique metric generator set in connected graphs. The algorithm aims to minimize the size of a set of vertices that can uniquely identify the l-cliques in the graph. This problem, as a generalization of the classical metric dimension, is NP-hard and requires heuristic optimization methods due to its computational complexity. The proposed algorithm uses a tabu list and generates neighboring sets to search for an optimal solution, and through an evaluation function, it minimizes the number of vertices with identical metric codes. The results demonstrate that this approach is highly effective in identifying group structures in complex networks, with potential applications in network analysis and data mining. https://scj.kashanu.ac.ir/article_115168.html Using a qualitative approach, this study compared the experiences of students and teachers in traditional algebraic thinking classrooms with those in constructivist classroom discussion approaches implemented in experimental school teaching. Content analysis was used from a qualitative perspective, combined with social constructivism and situated algebraic thinking theories, to interpret students' algebraic thinking and development. The research results indicated differences in the group of students exposed to the constructivist algebraic thinkingal environment, particularly in their independent skills. However, the study also highlighted several challenges, such as time management, understanding classmates' conversations, writing to convey their thinking, and an increased workload for students. The Fuzzy structural equations-Based content discussion classes employed innovative methods to create new knowledge and provided more opportunities for students to develop their ideas. This environment also fostered a more social/collective/adaptive form of knowledge by continuously evaluating information provided by students to inform teaching practices. Overall, the findings suggest that students who participated in innovative discussion approaches and Fuzzy structural equations-Based content had richer algebraic thinking experiences due to their active involvement during training. https://scj.kashanu.ac.ir/article_115169.html With the rapid advancement of autonomous and connected vehicle technology, positioning accuracy has become a critical requirement in navigation systems. Positioning systems such as Global Navigation Satellite System and Inertial Navigation System face challenges in urban environments, including Global Navigation Satellite System signal blockage and the accumulation of temporal errors in Inertial Navigation System. Standard Kalman Filter and adaptive methods like the Adaptive Kalman Filter and Innovation-Based Adaptive Kalman Filter are unable to effectively reduce the impact of outlier data and prevent filter divergence under Global Positioning System signal disturbances. This paper proposes an improved method called the Innovation-Based Adaptive Kalman Filter optimized with a Genetic Algorithm, which consists of two stages: (1) using Improved Adaptive Kalman Filter to adaptively update the measurement noise covariance matrix online and detect outlier data through the Chi-squared test; (2) further optimizing the measurement noise covariance matrix using a Genetic Algorithm. This approach reduces the impact of outlier data and enhances filter stability against environmental noise variations. Simulation results demonstrate that the Improved Innovation Adaptive Kalman Filter with Genetic Algorithm method outperforms Kalman Filter, Adaptive Kalman Filter, and Improved Adaptive Kalman Filter in reducing positioning errors and maintaining filter stability in challenging urban environments. These improvements indicate the capability of Improved Innovation Adaptive Kalman Filter with Genetic Algorithm to enhance positioning accuracy and preserve filter stability in integrated navigation systems. https://scj.kashanu.ac.ir/article_115344.html The classification accuracy of datasets heavily depends on their features. The presence of irrelevant and redundant features in a dataset can lead to a reduction in classification accuracy. Identifying and remov-ing such features is the main purpose of feature selection problem, which is an important step in the data science lifecycle. The aim of the Wrapper feature selection method is to reduce the number of selected features (SF) while improving the classification accuracy by optimizing a set of features. Feature selec-tion is a challenging and computationally expensive problem that falls under the NP-complete category, so it requires computationally efficient algorithms to solve it. The Artificial Rabbits Optimization (ARO) is a biologically inspired optimization technique that mimics the unique and intelligent foraging tactics of rabbits in nature. This paper proposed a new feature selection method based on the ARO meta-heuristic algorithm, called the memory artificial rabbits optimization (MARO), to improve its performance for solving feature selection problems. The proposed MARO method is tested on a standard benchmark da-taset and compared with four state-of-the-art feature selection algorithms. The results show the effec-tiveness of the proposed MARO algorithm in searching for an optimal subset of features. https://scj.kashanu.ac.ir/article_115345.html Heart diseases are one of the important factors that threaten the health of the society. Accurate and early recognition of these diseases improves the possibility of optimal intervention and treatment and can improve the ability to increase survival and reduce disability. In recent years, with the advancement of technology and the development process of electronic medical systems, the use of ECG signals as a non-destructive and non-invasive method to diagnose heart diseases has increased. In this article, the combination of convolutional neural network (CNN) and fuzzy logic is used to automatically detect heart disease from ECG signals. The purpose of combining the CNN and fuzzy logic is that the system can deal with cognitive uncertainties more like humans and have the possibility of processing uncertain and incorrect information. The proposed model was tested on the MIT-BIH arrhythmia dataset and the results show that the proposed model, with 97.54% accuracy, shows better performance than other methods. https://scj.kashanu.ac.ir/article_115346.html Machine reading comprehension (MRC) is a long-standing topic in natural language processing (NLP). The MRC task aims to answer a question based on the given context. Recently studies focus on multi-hop MRC which is a more challenging extension of MRC, which to answer a question some disjoint pieces of information across the context are required. Due to the complexity and importance of multi-hop MRC, a large number of studies have been focused on this topic in recent years, therefore, it is necessary and worth reviewing the related literature. This study aims to investigate recent advances in the multi-hop MRC approaches based on 34 studies from 2018 to 2024. In this regard, first, the multi-hop MRC problem definition will be introduced, then 34 models will be reviewed in detail with a strong focus on their multi-hop aspects. They also will be categorized based on their main techniques. Finally, a fine-grain comprehensive comparison of the models and techniques will be presented. https://scj.kashanu.ac.ir/article_115347.html Action Quality Assessment (AQA), a prominent and rapidly growing field in computer vision, focuses on developing automated and objective methods to evaluate the correctness of actions and the level of skill demonstrated in videos. Its diverse applications in sports, healthcare, industrial production, and other emerging domains have attracted significant research attention. Despite remarkable progress, there remains a strong need for a comprehensive and systematic review to consolidate fragmented knowledge and identify future research priorities. In this systematic review, following the standard Kitchenham methodology, 100 relevant studies were selected and analyzed. The field of AQA has evolved from foundational research toward fine-grained, multimodal, generalizable, and multitask approaches. Furthermore, emerging research trends such as continual learning, self-supervised learning, and explainable AI systems—particularly neuro-symbolic approaches—play a pivotal role in providing transparent and actionable feedback. This review offers a holistic perspective on various aspects of the field, including a systematic examination of methods, benchmark datasets, evaluation metrics, existing challenges, and future research directions. Its primary objective is to provide a valuable reference for both newcomers and experienced researchers, facilitating subsequent studies and guiding future advancements in AQA. https://scj.kashanu.ac.ir/article_115348.html The spreading of artificial intelligence, especially machine learning applications in the medical field, has opened that possibility of raising the precision and effectiveness of diagnosis and treatment of diseases. One of the important applications is the facilitation of the detection of the patient's response to the treatment by cardiac resynchronization therapy, a treatment method for patients with heart failure in which the function of the heart is improved by the coordinated stimulation of the ventricles. One of the main challenges in this area is class imbalance among the numbers of responders and non-responders to the treatment, which lowers the accuracy of the classification models. This research proposes a hybrid method based on Synthetic Minority Over-sampling with generalized class and ensemble learning. To improve the quality of synthetic samples, the Crow Search Algorithm has been used as a metaheuristic method and the Genetic Algorithm has been used for dimensionality reduction and efficient feature selection. For the classification stage, two ensemble learning models of Gradient Boosting and Random Forest were implemented. The utilized dataset includes 60 initial features, which have been reduced to 41 optimized selected features using the genetic algorithm. The criterion of optimality was the maximization of model accuracy in the identification of non-responders. The performance of the proposed method with these 41 selected features showed the average harmonic with the value of 89.07% and the accuracy of the non-response class of 93.59%. The results of this study indicated the combination of optimization, oversampling, and ensemble learning methods could effectively increase the identification accuracy of the non-responder patients of cardiac resynchronization therapy and thus provide assistive data-driven medical decision-making. https://scj.kashanu.ac.ir/article_115356.html The estimation of unknown boundary conditions in inverse parabolic partial differential equations (IPDEs), such as the Fisher equation, presents significant challenges due to the ill-posed nature and sensitivity to noise of these problems. Traditional methods often require strong prior assumptions or initial guesses, limiting their general applicability and accuracy. In this paper, we propose a robust hybrid numerical framework that integrates a fully implicit finite difference scheme with the parameter-free Secretary Bird Optimization Algorithm (SBOA) to address inverse Fisher equation problems (IFEPs) without prior knowledge of the unknown boundary function. The SBOA algorithm, inspired by the predator-prey dynamics of secretary birds, is employed to efficiently minimize the discrepancy between numerical solutions and noisy observation data, enabling precise recovery of the unknown boundary condition. Numerical experiments conducted on benchmark IFEPs demonstrate that the proposed method achieves outstanding precision, with relative errors as low as 0.07%, and consistently outperforms nine state-of-the-art metaheuristic algorithms in both accuracy and convergence speed. The algorithm also exhibits strong stability under varying grid sizes and noise levels, with solutions typically obtained within seconds on standard computing hardware. These results affirm the effectiveness and reliability of the SBOA-based framework as a powerful and scalable tool for solving complex inverse problems in computational science and engineering. https://scj.kashanu.ac.ir/article_115429.html Abstract: In this study, a binary version of the Ivy-inspired algorithm (IVY) was designed and evaluated to solve the 0–1 Knapsack Problem. The Knapsack Problem is a classical combinatorial optimization problem with applications in resource allocation, scheduling, and project planning. Due to the exponential complexity of its search space, efficient and effective algorithms are required to solve it. The BiIVY version combines a directed growth mechanism, a penalty function, and a specialized repair algorithm, enabling it to find higher-quality solutions with fewer iterations compared to conventional algorithms. Algorithm parameters were determined based on previous studies and experimental trials. The performance of BiIVY was evaluated on 25 standard datasets (L1–L25) and compared with Binary Flower Pollination Algorithm (BFPA) and Binary Sine-Cosine Algorithm (BSCA). Friedman mean-rank test results indicate that BiIVY generally outperforms the other algorithms in terms of total profit and iteration count, providing optimal or near-optimal solutions with fewer iterations, which demonstrates the effectiveness and capability of the proposed algorithm in solving complex combinatorial problems. https://scj.kashanu.ac.ir/article_115430.html The challenge of data classification by insufficient labeled data can be solved by domain adaptation techniques and leveraging external knowledge. However, most of these techniques lose robustness in noisy environments where the labels and features become corrupted. Aiming to model the indiscernibility and vagueness in domain adaptation, the present paper introduces a twin model for domain adaptation that combines the quarter-sphere support vector machine data description (QS-SVM) with a new fuzzy rough set-based weighting approach. The proposed model learns two small hyperspheres per domain regarding binary classification by solving two linear equations rather than one Quadratic Programming Problem (QPP), unlike standard QS-SVM. Consequently, the time complexity is reduced by this strategy. The Benefit of the fuzzy rough set is that only the high-confidence samples influence the adaptation and classification results of the hyperspheres. The strength of the proposed model is that after constructing and training the source domain classifiers, accessibility to the source domain data is not required, and the existence of only the source domain hyperspheres is sufficient. Also, the proposed fuzzy rough set-based sample weighting method ensures that the minority classes that are often underrepresented in the dataset are not overlooked when constructing the model. The effectiveness of the proposed model has been compared to the state-of-the-art methods on fifteen tasks taken from two benchmark datasets. The experimental results demonstrate the superiority of the proposed model over state-of-the-art ones in terms of classification accuracy and computational time. Besides, the noise analysis proves the robustness of the proposed model.