This paper presents an IoT-based software architecture designed to provide navigation assistance for visually impaired individuals, specifically addressing critical limitations inherent in existing systems, including cloud dependency, processing latency, and insufficient localized linguistic support. The architecture employs wearable smart glasses integrated with an ESP32CAM module, and utilizes edge-based artificial intelligence to enable real-time environmental perception and interpretation. Object detection employs a lightweight YOLOv8n model, complemented through intelligent data fusion, walkability assessment, alert prioritization, and predictive navigation algorithms, which enhance contextual awareness and decision-making. Seven coordinated algorithms govern the image preprocessing, detection, temporal fusion, path safety evaluation, voice synthesis, and Kalman-based adaptive routing. The system generates prioritized navigation cues via Kinyarwanda voice notifications, thereby minimizing the cognitive load and enhancing user accessibility. The proposed modular cloud-independent architecture demonstrates considerable potential for scalable real-world deployment, offering a practical and inclusive navigation solution for visually impaired users navigating diverse outdoor environments.
This work presents the ideas of Lukasiewicz fuzzy implicative filters and Lukasiewicz fuzzy positive implicative filters within the context of hoops. We provide a detailed characterization of the conditions under which a Lukasiewicz fuzzy set can be regarded as a Lukasiewicz fuzzy implicative filter and Lukasiewicz fuzzy positive implicative filter. Additionally, we examine the relationship between these filter types with Lukasiewicz fuzzy filter. Furthermore, the relationship between Lukasiewicz fuzzy implicative filter and Lukasiewicz fuzzy positive implicative filter is analyzed.
This paper explores key structural properties of elegant fuzzy labeling graphs, focusing on tree structures, edge classification, and vertex coloring. We begin by characterizing tree structures within this framework and propose a classification of edges based on their strength and connectivity to enhance structural analysis. A significant contribution of this work is the development of an efficient algorithm to determine the minimum fuzzy vertex coloring of a fuzzy labeling graph. Additionally, we introduce a novel approach for determining the fuzzy chromatic polynomial of a fuzzy labeling graph. As an application, we demonstrate how fuzzy vertex coloring can be effectively utilized in session allocation for course scheduling. These findings provide deeper insights into both the structural and computational aspects of fuzzy labeling graphs, contributing to further research in fuzzy graph theory.
Clustering with intuitionistic fuzzy C-means (IFCM) extends the classical fuzzy C-means (FCM) framework by introducing dual notions of membership and non-membership, controlled via a novel objective function Jm,'v. In IFCM, each data point xk is assigned both a membership degree & micro;zk and a non-membership degree nu zk for cluster i, with their sum bounded by one, yielding a hesitation margin pi zk = 1-& micro;zk-nu zk. We derive closed-form update rules for & micro;zk, nu zk, and cluster prototypes vz, and prove monotonic descent and global convergence of the algorithm under Zangwill's theorem. A detailed synthetic-data case study on a three-component Gaussian mixture illustrates IFCM's behavior: membership and hesitation surfaces visualize soft boundaries and uncertainty; sensitivity analysis shows how the fuzzifier m and weight gamma affect clustering validity; and initialization strategies (random, K-means++, IFS-based) are compared in terms of adjusted rand index (up to 0.70), normalized mutual information (up to 0.68), partition coefficient, and partition entropy. IFCM consistently outperforms classical FCM by better discriminating borderline points and explicitly quantifying the residual uncertainty, while converging reliably within 10 iterations. Our contributions include a unified objective formulation, a rigorous convergence proof, and practical guidelines for parameter selection. IFCM paves the way for robust soft-clustering applications in domains where modeling hesitation and non-membership are critical.
This paper introduces soft S-closed spaces, a new class of soft topological spaces. We establish a number of characterizations, including one involving soft filterbases, which establishes that a soft topological space is soft S-closed if and only if every soft filterbase has a soft saccumulation point. This study placed soft S-closedness in context with other related notions by exploring its relationship with soft compact, soft quasi-H-closed, and soft nearly compact spaces. In addition, the properties of S-closedness with respect to soft semicontinuous and soft irresolute mappings are also studied. Finally, we create a one-to-one correspondence for generated soft topologies, showing under which a topology is soft S-closed if and only if every classical space of the generated topology is S-closed.
High-dimensional time series classification often requires complex models, extensive feature engineering, or deep learning architectures. We propose a simple and training-free classification method that leverages a recent similarity measure called the Dimension Insensitive Euclidean Metric (DIEM), developed by Tessari and his colleague in 2024. Unlike traditional distance-based or learning-based approaches, our method directly compares input vectors with labeled instances using DIEM, without requiring model training or parameter tuning. To evaluate its effectiveness, we applied it to benchmark datasets with varying dimensionality. In particular, on the Olive Oil dataset, the method achieved an accuracy of 87.2% and was over 2,000 times faster than a Gramian angular field (GAF)-based CNN. On the Meat dataset, it experienced only a 1.7% point reduction in accuracy compared to GAF, while still being more than 1,400 times faster. Our results show that the proposed method not only matches or exceeds the accuracy of complex models but also offers significant computational advantages across a variety of classification tasks.
This paper introduces S-bipolar fuzzy semigroups, a unified extension of S-fuzzy semigroups that models systems with dual, conflicting, or uncertain information. S-bipolar fuzzy semigroups combine the negative and positive membership functions into an S-semigroup. In this, the algebraic foundations such as fuzzy ideals, normal subsemigroups, cosets, orders and intersection operations of S-bipolar fuzzy semigroups are developed. The study provides evidence that S-bipolar fuzzy semigroups are more expressive and detailed in their models compared to classical fuzzy semigroup forms.
This study investigates how, in the digital age of finance, high-frequency trading (HFT) stands at the forefront of leveraging technology for rapid trading decisions. With the increasing complexity of markets, integrating fuzzy logic into HFT offers a novel approach to address the uncertainties inherent in financial environments. This study adopts a multifaceted approach to assess the efficacy of fuzzy logic in HFT. Using financial data to benchmark a fuzzy-logic-based strategy against traditional HFT approaches, optimization tools were evaluated, and the fuzzy-logic-based trading algorithm was fine-tuned. This bridging of theoretical concepts with application ensures that the findings are relevant and applicable. The results revealed that the fuzzy-logic-based trading strategy exhibited consistent superiority over traditional HFT methods, particularly in volatile market scenarios. By dynamically adapting to market nuances, this strategy exhibits remarkable resilience and adaptability. The inclusion of optimization tools, such as genetic algorithms and neural networks, amplified the strategy's performance, yielding higher risk-adjusted returns. This study provides valuable insights into a potential paradigm shift in algorithmic trading. The demonstrated efficacy of the fuzzy-logic-based strategy, coupled with optimization techniques, indicates a promising avenue for future trading innovations.
This study addresses the global health issue of cardiovascular diseases by using machine learning techniques to improve early-stage prediction of heart diseases. K-nearest neighbors, naive Bayes, logistic regression, decision trees, support vector machines, and random forest were used as individual classifiers to be carefully evaluated and compared. Additionally, the efficacy of the stacking ensemble technique was examined. The data from the UCI Machine Learning Repository were preprocessed and examined using performance metrics (accuracy, precision, recall, F1-score, and AUC) and stratified cross-validation. The results indicate that while the suggested stacking classifier with a logistic regression meta-learner performed competitively (84.78%), the random forest attained the best accuracy (86.41%). These findings emphasize the value of model diversity and meta-learner selection, while highlighting the promise of ensemble approaches in clinical diagnosis. Future research will focus on advanced ensemble strategies, improved model interpretability, comprehensive parameter optimization, and the validation of larger and more diverse clinical datasets to enhance robustness and generalizability.
Artificial intelligence (AI) has seen an extensive boom in all fields, including media, commerce, technology, and healthcare. In spite of this proliferation, its predictions and analysis still create a sense of distrust among those who must use these results, particularly in critical fields such as healthcare, national security, and complex domains such as law and order. These black-box models generate trust issues among users. Explainable artificial intelligence has emerged as a set of methods and processes that allows users to understand and trust the results of AI models. This study focuses on how explainable artificial intelligence can provide transparent, interpretable, and reliable decision-making processes, making it a compelling alternative to traditional AI models in the healthcare industry. Additionally, a comprehensive review of the various applications of artificial intelligence in the medical field is presented for diagnosing complex diseases such as cancer, neurological problems, medical diagnoses, and pathology. This analysis, combined with the practical knowledge of clinical professionals, can significantly impact the functioning of our traditional healthcare system and prove to be a transformative, patient-centered approach to healthcare delivery.
Mathematical problem-solving with visual components remains a significant challenge for artificial intelligence (AI) systems. This is because conventional OCR pipelines often fail to extract critical structural cues such as axes, tick marks, and spatial relationships in diagrams or graphs. Despite recent advances in generative models, a cohesive framework that integrates visual perception with symbolic inference is still lacking. Therefore, this study proposes a lightweight hybrid pipeline that combines ColPali for vision-language understanding with open-source large language models (LLMs) such as LLaMA for symbolic reasoning. We evaluated our system on the MathVision dataset, leveraging its category-level statistics to measure the performance across diverse visual math tasks. Our method improves accuracy by up to 29.3% compared to standalone models, demonstrating the effectiveness of integrating visual structures into mathematical reasoning.
Fuzzy logic is a crucial topic that will bring fundamental changes to the medical field; it is a procedure to address uncertainty in cases with unclear medical images. This research focused on leveraging the features of fuzzy logic in efficiently analyzing the data and information contained in medical images and the ability to think and reason, owing to the facts and rules built into it, to obtain the required results. Skin diseases affect people worldwide. Providing healthcare and timely and accurate diagnoses of diseases are important. In the current era, artificial intelligence has emerged as a significant topic to analyze and diagnose skin diseases. This study examined eight medical images of patients with different skin diseases. Manual regions of interest were selected using the ImageJ system. Normal and diseased areas of equal size were selected, and their average colors (red, green, and blue) were measured. The values were then normalized. The difference between the normalized average color values of the normal and diseased areas was determined. Rules were proposed using fuzzy logic to analyze the results and identify the most distinctive color in the medical image.
Complex fuzzy distance measures play a vital role in analyzing intricate and high-dimensional data, whereas conventional distance measures often fail to capture the nuanced characteristics of information. In this study, we propose a novel distance measure along with a weighted variant based on complex fuzzy sets. The fundamental properties of the proposed measure were examined through union, intersection, and complementary operations within a complex fuzzy set framework. Furthermore, we developed a decision-making algorithm that utilized these distance measures to identify the optimal choice among multiple alternatives. To demonstrate the effectiveness of the proposed approach, a practical example is presented that highlights its application to real-world decision-making problems.
This study introduces alpha b-approximations, anew class of generalized rough sets derived from basic neighborhood systems. Unlike traditional approaches that require strict topological assumptions, this framework preserves Pawlak's principles while offering greater simplicity and applicability. Five operators were defined and analyzed, with proofs and examples confirming their superior accuracy. Economic experiments demonstrate up to 100% classification accuracy in identifying high-growth countries, whereas a MATLAB implementation validates the computational efficiency. Overall, the proposed approach provides a flexible and practical tool for precise decision-making with broad application potential.
Forecasting multi-attribute time series (MATS) data remains a challenging task owing to complex data patterns and various sources of uncertainty. To address this, this study proposes two advanced forecasting models, plithogenic-intuitionistic fuzzy soft set (P-IFSS) and plithogenicneutrosophic soft set (P-NSS), which extend conventional soft set frameworks by integrating the plithogenic concept. This integration enables a more effective representation of the degree of contradiction among the attributes, thereby improving the accuracy and robustness of the forecast. Using Indonesian bond-yield data as a case study, the performance of the proposed models was evaluated against traditional intuitionistic fuzzy soft set (IFSS) and neutrosophic soft set (NSS) models. The experimental results demonstrated that both P-IFSS and P-NSS consistently outperformed their non-plithogenic counterparts across multiple training periods and parameter settings. The findings highlight the potential of plithogenic soft set extensions as an effective framework for handling complex uncertainties in multi-attribute time-series forecasting, particularly in financial data analysis and decision-making applications.
This paper introduces the concepts and properties of complex fuzzy ideals (generalized biideals, bi-ideals, interior ideals, quasi-ideals, and (1, 2)-ideals) in semigroups, with proofs. Furthermore, we investigate the essential conditions for coincidence in the regular, intraregular, semisimple, weakly regular, or quasi-regular manner, in semigroups.
To the best of our knowledge, this study is the first attempt to predict the water level of the Gamcheon River, Korea, using the large language model (LLM)-based TimeGPT. Our research shows that TimeGPT outperforms existing time-series models, including SARIMAX and DLinear. TimeGPT provides better prediction accuracy for short-and long-term forecasts under Nash-Sutcliffe efficiency scores. TimeGPT effectively captures water level changes by integrating rainfall data. Our approach provides insights into water resource management and flood prediction.
Transportation systems face numerous challenges in cost optimization due to fluctuating and uncertain environmental factors, which reduce the effectiveness of traditional methods. To address these uncertainties, interval-valued picture fuzzy sets (IVPFSs) provide an effective framework for deriving optimized solutions to transportation problems. This study proposes a ranking approach for interval-valued picture fuzzy numbers (IVPFNs) and integrates it with the modified distribution (MODI) method to optimize transportation costs modeled directly as IVPFNs. Using the proposed ranking method in conjunction with the MODI approach, the 3 & times; 3 transportation problem achieved an optimal fuzzy cost of (137, 315, 511) with intervals of [0.05, 0.1], [0, 0.1], and [0.6, 0.7], which outperformed the interval-valued intuitionistic fuzzy transportation problem (IVIFTP) result of (145, 306, 520) with intervals of [0.2, 0.4] and [0.03, 0.07]. Similarly, for the 3 & times; 4 problem, the interval-valued picture fuzzy transportation problem (IVPFTP) yields an improved optimal fuzzy cost, further demonstrating the effectiveness of the method. The MODI-based solution, combined with the new ranking method, ensures optimality while capturing neutrality, thereby highlighting the advantages of IVPFN-based cost modeling under uncertainty.
The uncertainty and vagueness of datasets associated with real-world problems render them complex and unpredictable. Problems such as pattern recognition and image processing require advanced mathematical tools to capture the maximum uncertainty, and conventional measures are typically inadequate in achieving this goal. Associating information measures with a fuzzy environment can resolve the issue of vagueness and precisely quantify data. In this study, a novel picture fuzzy divergence measure is proposed. The proposed study discusses information measures for picture fuzzy sets and illustrates important properties of the proposed measures. The effectiveness of the proposed measure over existing measures is demonstrated in the image-processing and pattern-recognition problems discussed. Comparative analysis with existing information measures highlights the constructive performance of the proposed measure. This study contributes to the advancement of fuzzy divergence theory and opens new avenues for uncertainty modeling in various domains. In general, this study contributes to the understanding of the significant role of fuzzy environments in managing imprecise and vague patterns.
The measurement of semantic similarity between sentences in Indonesian still faces challenges, especially in capturing contextual meanings that cannot be represented by rule-based approaches such as the Kamus Besar Bahasa Indonesia (KBBI). This study aims to evaluate and enhance the semantic representation of Indonesian sentences using embedding-based representation learning methods. Five models were compared: TF-IDF, FastText, IndoBERT, pre-trained SBERT, and the proposed method SBERT fine-tuned using the triplet loss approach. The dataset consists of 3,000 sentence pairs constructed from 500 KBBI entries, each labeled as a synonym or non-synonym. The evaluation results show that the proposed model achieves the highest accuracy of 92% and an F1-score of 0.92, followed by IndoBERT (90%), pre-trained SBERT (82%), FastText (65%), and TF-IDF (51%). The integration of the triplet loss effectively optimizes the vector positioning of sentence representations in the semantic space, allowing sentences with similar meanings to be closer and those with different meanings to be further apart. These findings demonstrate that the proposed approach is capable of capturing complex semantic nuances and contextual variations in Indonesian, contributing significantly to the development of more accurate meaning-based models for various natural language processing tasks, such as semantic similarity measurement and text classification.
