https://reunir.unir.net/handle/123456789/19072 2026-03-10T07:42:14Z 2026-03-10T07:42:14Z Díaz Gutiérrez, Yesid Cueva-Lovelle, Juan Manuel Candia Herrera, Diana Carolina https://reunir.unir.net/handle/123456789/19132 2026-03-06T12:25:21Z

Security Model for the Internet of Things, Through Blockchain Díaz Gutiérrez, Yesid; Cueva-Lovelle, Juan Manuel; Candia Herrera, Diana Carolina Due to the proliferation of computer crimes related to information vulnerability handled by people and entities and evidenced in attacks of financial, commercial, personal and even family nature; a need has been identified to implement, security strategies and protocols in each and every one of these areas, which make possible the effective protection of the integrity and privacy of data. Regarding this, there are protection schemes such as cryptography and reliable time stamping which undoubtedly have managed to partially solve this problem by attacking structural and crucial points. However, the evolution in the technology field has been currently represented in the fourth industrial revolution and its context towards 4.0 technologies and smart industries; various technologies have been positioned in the emerging and disruptive categories, among which the Internet of Things (IoT) stands out. This technology has become the target of multiple computer attacks, due to the processes of Extraction, Transformation, Loading and Transmission of large volumes of data. Alongside its widespread connection to the Internet, it’s become a strategic target for such attacks. A possible alternative solution to this situation is blockchain, which allows information to be public and stored in different blocks, which makes it easier to guarantee the integrity of information based on the following aspects: • Identification of the attacked and / or compromised information, which can be marked as invalid information. • Public report of the attack. • Information backup in another block to facilitate its recovery. In this regard, it is important to highlight that these functional and technological characteristics offered by the blockchain, facilitate the management of information and its integrity. However, it is necessary and essential to previously guarantee the structure of the information generated; as some processes of Business Intelligence (BI), such as the Extraction, Transformation and Load scheme (ELT), would be of great relevance and support during the development of this procedure. Submitted by Angela María Porras Ruiz (angela.porras@unir.net) on 2026-03-06T12:25:21Z No. of bitstreams: 1 Security Model for the Internet of Things, Through Blockchain.pdf: 450339 bytes, checksum: 6744590dde281fb739e033db5658a521 (MD5); Made available in DSpace on 2026-03-06T12:25:21Z (GMT). No. of bitstreams: 1 Security Model for the Internet of Things, Through Blockchain.pdf: 450339 bytes, checksum: 6744590dde281fb739e033db5658a521 (MD5)

Vijayan, Rajilal Manathala Granelli, Fabrizio Umamakeswari, A. https://reunir.unir.net/handle/123456789/19131 2026-03-06T12:20:20Z

An Adaptive Framework for Resource Allocation Management in 5G Vehicular Networks Vijayan, Rajilal Manathala; Granelli, Fabrizio; Umamakeswari, A. Vehicle-to-everything (V2X) communication is crucial in vehicular networks, for enhancing traffic safety by ensuring dependable and low latency services. However, interference has a significant impact on V2X communication when channel states are changed in a high mobility environment. Integration of next generation cellular networks such as 5G in V2X communication can solve this issue. Also, successful resource allocation among users achieves a better interference control in high mobility scenarios. This work proposes a novel resource allocation strategy for 5G cellular V2X communication based on clustering technique and Deep Reinforcement Learning (DRL) with the aim of maximizing systems energy efficiency and MVNO’s profit. DRL is used to distribute communication resources for the best interference control in high mobility scenarios. To reduce signalling overhead in DRL deployments, the proposed method adopted RRH grouping and vehicle clustering technique. The overall architecture is implemented in two phases. The first phase addresses the RRH grouping and vehicle clustering technique with the objective of maximising the energy efficiency of the system and the second phase addresses the technique of employing DRL in conjunction with bidding to optimise MVNO’s profit. Second phase addresses the resource allocation which is implemented in two level stage. First level addresses the bidding of resources to BS using bidding and DRL techniques and the second level addresses the resource allocation to users using Dueling DQN technique. Through simulations, the proposed algorithm's performance is compared with the existing algorithms and the results depicts the improved performance of the proposed system. Submitted by Angela María Porras Ruiz (angela.porras@unir.net) on 2026-03-06T12:20:20Z No. of bitstreams: 1 An Adaptive Framework for Resource Allocation Management in 5G Vehicular Networks.pdf: 1076396 bytes, checksum: 3309ad5626d22e3cdb0b884466631f03 (MD5); Made available in DSpace on 2026-03-06T12:20:20Z (GMT). No. of bitstreams: 1 An Adaptive Framework for Resource Allocation Management in 5G Vehicular Networks.pdf: 1076396 bytes, checksum: 3309ad5626d22e3cdb0b884466631f03 (MD5)

Sadesh, S. Thangaraj, Rajasekaran Pandiyan, P. Priya, R. Devi Naveen, Palanichamy https://reunir.unir.net/handle/123456789/19130 2026-03-06T12:15:38Z

Identification of Monkeypox Disease Based on MpoxNet and Swin Transformer Models Using Mobile Application Sadesh, S.; Thangaraj, Rajasekaran; Pandiyan, P.; Priya, R. Devi; Naveen, Palanichamy Humankind is still reeling from the devastating impact of the Covid-19 pandemic, yet another looming threatis the potential global spread of the monkeypox virus. While monkeypox may not pose the same level of lethality or contagion as COVID-19, its significant spread across countries is cause for concern. Already, outbreaks have been reported in 75 nations worldwide. Despite sharing clinical characteristics with smallpox, including lesions and rashes, monkeypox symptoms are frequently mistaken for those of other poxviruses such as chickenpox and cowpox. Consequently, accurate early diagnosis of monkeypox by healthcare professionals remains challenging. Automated monkeypox identification using Deep Learning (DL) techniques presents a promising avenue for addressing this challenge. In this study, a modified deep convolutional neural network (DCNN) model named MpoxNet is proposed for the identification of monkeypox disease. The performance of MpoxNet is evaluated against established DCNN models, including ResNet50, VGG16, VGG19, DenseNet121, DenseNet169, Xception, InceptionResNetV2, and MobileNetV2. This study addresses the pressing challenge of monkeypox identification by proposing MpoxNet. With the aim of enhancing early detection and containment efforts, MpoxNet's performance is evaluated against established DCNN models across two distinct datasets: MSLD and MSID Dataset. Results reveal MpoxNet's superior test accuracy of 94.82% on the MSLD Dataset, surpassing other models. However, evaluation on the MSID Dataset highlights variations in performance, emphasizing the influence of dataset characteristics. Additionally, the introduction of the Swin Transformer model demonstrates exceptional performance on the MSLD and the MSID Dataset and, achieving an accuracy of 98%. These findings underscore the importance of considering diverse datasets and leveraging advanced techniques for robust monkeypox detection systems. Integration of MpoxNet with a mobile application offers a promising solution for rapid and precise monkeypox disease detection, providing valuable insights for future research and real-world deployment strategies to effectively combat the global spread of monkeypox. Submitted by Angela María Porras Ruiz (angela.porras@unir.net) on 2026-03-06T12:15:38Z No. of bitstreams: 1 Identification of Monkeypox Disease Based on MpoxNet and Swin Transformer Models Using Mobile Application.pdf: 783729 bytes, checksum: c12c03c73eb88128014adcf20f2337ba (MD5); Made available in DSpace on 2026-03-06T12:15:38Z (GMT). No. of bitstreams: 1 Identification of Monkeypox Disease Based on MpoxNet and Swin Transformer Models Using Mobile Application.pdf: 783729 bytes, checksum: c12c03c73eb88128014adcf20f2337ba (MD5)

Elmannai, Hela Saleem, Nasir Bourouis, Sami Alkanhel, Reem Ibrahim https://reunir.unir.net/handle/123456789/19129 2026-03-06T10:42:33Z

Self-Supervised Attentive Feature Learning for Alzheimer’s Disease Detection Elmannai, Hela; Saleem, Nasir; Bourouis, Sami; Alkanhel, Reem Ibrahim Alzheimer’s disease (AD) is a progressive neurodegenerative disorder that leads to memory loss and a decline in cognitive abilities. It primarily affects older adults and is the most common cause of dementia. Using deep learning, models can analyze brain imaging scans to detect specific patterns and biomarkers associated with the disease. Supervised learning models achieve high accuracy rates, but they require a large amount of data sets and labelled medical images. Self-supervised learning can achieve high accuracy rates with fewer training data. This study proposes a self-supervised attentive feature learning network (SSA-Net) for classifying Alzheimer’s disease. The proposed approach leverages self-supervised learning and attention mechanisms to enhance the accuracy and reliability of the classifying model. We employ ResNet-50, incorporating attentive activation, which replaces the ReLU activation, improving the ability of the neural model to focus on the most relevant features in the input medical images. We use SimCLR (Simple Framework for Contrastive Learning of Visual Representations) with the ResNet-50 backbone as a self-supervised learning framework that effectively learns high-quality visual representations in brain MRI (Magnetic Resonance Imaging) scans without labelling. We used the Kaggle Alzheimer’s classification dataset (KACD) containing brain MRI scans for training and testing. Experimental results on the KACD dataset show that the proposed attentive self-supervised ResNet50 reached 99.7% classification accuracy compared to the traditional ResNet50 with 98.1% accuracy. Evaluation metrics show the effectiveness of the proposed SSA-Net for the efficient classification of Alzheimer’s disease. Submitted by Angela María Porras Ruiz (angela.porras@unir.net) on 2026-03-06T10:42:32Z No. of bitstreams: 1 Self-Supervised Attentive Feature Learning for Alzheimer’s Disease Detection.pdf: 893051 bytes, checksum: 97b61d8ee0b13f3aea3f086ae205f8ea (MD5); Made available in DSpace on 2026-03-06T10:42:33Z (GMT). No. of bitstreams: 1 Self-Supervised Attentive Feature Learning for Alzheimer’s Disease Detection.pdf: 893051 bytes, checksum: 97b61d8ee0b13f3aea3f086ae205f8ea (MD5)

Fatima, Mamuna Khan, Muhammad Attique Shaheen, Saima Kadry, Seifedine Alqahtani, Omar Alouane, M. Turki-Hadj https://reunir.unir.net/handle/123456789/19128 2026-03-06T10:36:41Z

Attention Embedded Residual Bottleneck CNN Architecture for Breast Cancer Diagnosis in Ultrasound Images Fatima, Mamuna; Khan, Muhammad Attique; Shaheen, Saima; Kadry, Seifedine; Alqahtani, Omar; Alouane, M. Turki-Hadj Breast cancer (BrC) stands as the predominant cancer among women, resulting in a substantial global mortality toll each year. Early detection plays a pivotal role in diminishing mortality rates. Manual diagnosis of BrC is time-intensive, intricate, and prone to errors, emphasizing the necessity for an automated system for timely detection. Various imaging methods have been investigated, underscoring the crucial need for accurate detection to prevent unwarranted treatments and biopsies. Recent years have witnessed substantial exploration and enhancement in the application of DL for efficiently processing medical images. This study aiming to create an effective and resilient DL framework for BrC detection and classification. The steps are contrast enhancement and augmentation, a hybrid CNN network ‘BrC-DeepRBNet’ is introduced that is built from scratch and incorporates several design elements including residual blocks, bottleneck architecture, and a self-attention mechanism. This framework is employed to construct two networks, one comprising of 107 layers and the other with 149 layers. Moreover, the network capitalizes on the benefits offered by batch normalization (BN) and group normalization (GN), utilizes ReLU and leaky ReLU as activation functions, and integrates Max pooling layer into its architecture in a series of residual-bottleneck blocks. Further, for feature fusion horizontal approach is used and optimization is done using generalized normal distribution optimization (GNDO). The selected features are further classified using neural network classifiers. The introduced framework achieved the highest classification accuracy at 97.05% with publicly available BUS dataset. A detailed ablation study is presented that demonstrates the superior performance of the presented approach, surpassing various pre-trained models (i.e. AlexNet, InceptionV3, ResNet50, and ResNet101) and existing BrC detection and classification techniques. Submitted by Angela María Porras Ruiz (angela.porras@unir.net) on 2026-03-06T10:36:41Z No. of bitstreams: 1 Attention Embedded Residual Bottleneck CNN Architecture for Breast Cancer Diagnosis in Ultrasound Images.pdf: 1319325 bytes, checksum: 4835fb493fbddb081c3c59346e1bdd4c (MD5); Made available in DSpace on 2026-03-06T10:36:41Z (GMT). No. of bitstreams: 1 Attention Embedded Residual Bottleneck CNN Architecture for Breast Cancer Diagnosis in Ultrasound Images.pdf: 1319325 bytes, checksum: 4835fb493fbddb081c3c59346e1bdd4c (MD5)

Sheela, A. Jeba Krishnamurthy, M. https://reunir.unir.net/handle/123456789/19127 2026-03-06T10:30:32Z

Ensemble Diabetic Retinopathy Severity Classification Framework With Optimized VGG16, Resnet, and Inception Features Sheela, A. Jeba; Krishnamurthy, M. Background problem: Diabetic Retinopathy (DR) is characterized by high glucose levels in the blood, which can lead to permanent vision loss and microvascular complications. Various deep learning techniques for DR analysis tend to be more complex and may experience delays in delivering accurate results, thereby limiting their application in clinical settings. Implementing real-time predictionand severity analysisof DR can address this problem by providing real-time diagnostic insights based on DR severity levels. Aim: So, this paper is intended to offer a new DR detection and severity classification model with the highranking-based ensemble learning approach. Methodology: The preprocessed and segmented images are utilized in the feature extraction processusing ensemble architecture which incorporated VGG16, Resnet, and Inception to get three sets of features. The optimal features are selected using an Adaptive Scavenger-Based Dingo Optimization Algorithm (AS-DOX) to achieve the efficient classification of DR severity. The optimization constraint stake place in the HighRanking-Based Deep Ensemble Learning (HR-DEL) model helps to enhance the efficacy of classification for the offered approach. The simulation analysis provides enhanced performance with the accurate classification of the designed DR severity classification approach by comparing it with other baseline methods. Result: From the result analysis, the offered method achieves 96.6 % accuracy and sensitivity rate. Moreover, it achieves a 90.52% precision rate. Conclusion: Thus, the designed DR severity classification model attains better performance, and also it is utilized for early detection of DR severity. Submitted by Angela María Porras Ruiz (angela.porras@unir.net) on 2026-03-06T10:30:32Z No. of bitstreams: 1 Ensemble Diabetic Retinopathy Severity Classification Framework With Optimized VGG16, Resnet, and Inception Features.pdf: 1425672 bytes, checksum: 1c1ae01ff8659faf7cdf408c333e09ad (MD5); Made available in DSpace on 2026-03-06T10:30:32Z (GMT). No. of bitstreams: 1 Ensemble Diabetic Retinopathy Severity Classification Framework With Optimized VGG16, Resnet, and Inception Features.pdf: 1425672 bytes, checksum: 1c1ae01ff8659faf7cdf408c333e09ad (MD5)

García Navarro, Alejandro L. Koneva, Nataliia Hernández, José Alberto Sánchez-Macián, Alfonso https://reunir.unir.net/handle/123456789/19115 2026-03-04T16:46:41Z

On the Use of Large Language Models at Solving Math Problems: A Comparison Between GPT-4, LlaMA-2 and Gemini García Navarro, Alejandro L.; Koneva, Nataliia; Hernández, José Alberto; Sánchez-Macián, Alfonso In November 2022, ChatGPT v3.5 was announced to the world. Since then, Generative Artificial Intelligence (GAI) has appeared in the news almost daily, showing impressive capabilities at solving multiple tasks that have surprised the research community and the world in general. Indeed the number of tasks that ChatGPT and other Large Language Models (LLMs) can do are unimaginable, especially when dealing with natural text. Text generation, summarisation, translation, and transformation (into poems, songs, or other styles) are some of its strengths. However, when it comes to reasoning or mathematical calculations, ChatGPT finds difficulties. In this work, we compare different flavors of ChatGPT (v3.5, v4, and Wolfram GPT) at solving 20 mathematical tasks, from high school and first-year engineering courses. We show that GPT-4 is far more powerful than ChatGPT-3.5, and further that the use of Wolfram GPT can even slightly improve the results obtained with GPT-4 at these mathematical tasks. Submitted by Angela María Porras Ruiz (angela.porras@unir.net) on 2026-03-04T16:46:41Z No. of bitstreams: 1 On the Use of Large Language Models at Solving Math Problems.pdf: 416406 bytes, checksum: ec1a652ce78b2e3a3d30e6fb54f768e6 (MD5); Made available in DSpace on 2026-03-04T16:46:41Z (GMT). No. of bitstreams: 1 On the Use of Large Language Models at Solving Math Problems.pdf: 416406 bytes, checksum: ec1a652ce78b2e3a3d30e6fb54f768e6 (MD5)

Wan, Xiuan Fang, Yuchun Wu, Jiahua Pan, Shouyong https://reunir.unir.net/handle/123456789/19081 2026-02-25T16:34:26Z

Geometrics Assisted Rubbing Generation and Semantics Enhanced Detection for Small and Dense OBI Character Wan, Xiuan; Fang, Yuchun; Wu, Jiahua; Pan, Shouyong Character detection is essential for subsequent Oracle Bone Inscription (OBI) research. However, the lack of labeled data and the complexity of small and dense OBI characters are the main difficulties in OBI detection research. In this paper, we propose a framework for rubbing generation that can automatically build up largescale rubbing samples with verisimilar scenarios to noisy wild OBI through geometric and morphological construction combined with style transferring. Moreover, we propose a semantic-enhanced detection model aiming at small and dense OBI through the fusion of multi-resolution feature maps with the enriched feature in the YOLOv5s backbone. We introduce the higher resolution and the Soft-NMS into the proposed OBI detection model to solve the overlapping of small and dense OBI characters. The augmented dataset improves the performance of benchmark object detection models in the real OBI detection task when sufficient data is lacking. Furthermore, the proposed OBI detection model can provide easy and preferable access to OBI detection even with a small number of labeled data and obtain preferable results. Experiments ascertain the effectiveness of the proposed OBI generation framework and the proposed OBI detection model. Submitted by Angela María Porras Ruiz (angela.porras@unir.net) on 2026-02-25T16:34:26Z No. of bitstreams: 1 Geometrics Assisted Rubbing Generation and Semantics Enhanced Detection for Small and Dense OBI Character.pdf: 2212538 bytes, checksum: 09e6fef4d982a81652653756e6970bff (MD5); Made available in DSpace on 2026-02-25T16:34:26Z (GMT). No. of bitstreams: 1 Geometrics Assisted Rubbing Generation and Semantics Enhanced Detection for Small and Dense OBI Character.pdf: 2212538 bytes, checksum: 09e6fef4d982a81652653756e6970bff (MD5)

Li, Xin Zhou, Xuan Xu, Sheng https://reunir.unir.net/handle/123456789/19080 2026-02-25T16:27:48Z

Three Dimensional Tree Modeling Based on the Skeleton Path Optimization and Geometrical Shapes Li, Xin; Zhou, Xuan; Xu, Sheng Nowadays, the 3D individual tree reconstruction has played a significant role in the phenotypic study of trees. This paper proposes a new automatic method for extracting skeletons of individual trees and reconstructing 3D models. Firstly, the Euclidean clustering is performed to obtain center points of candidate branch regions. Then, the initial skeletons of LiDAR point clouds are obtained by slicing clusters in three dimensions. Secondly, skeleton points are completed by the proposed branch tracking. Then, the radius of the branches is accurately estimated from the branches. Thirdly, optimal points are interpolated in appropriate directions to refine skeletons of individual trees. Then, the Laplacian algorithm is conducted for smoothing branches. After that, optimal geometric shapes are formulated to reconstruct the final 3D tree models. Experimental results show that the average accuracy of our individual tree models is up to 97.49%, which shows a promising algorithm in 3D tree reconstructions. Submitted by Angela María Porras Ruiz (angela.porras@unir.net) on 2026-02-25T16:27:48Z No. of bitstreams: 1 Three Dimensional Tree Modeling Based on the Skeleton Path Optimization and Geometrical Shapes.pdf: 1783238 bytes, checksum: 78e5941b5dd4a843db7d0c416d3cf06d (MD5); Made available in DSpace on 2026-02-25T16:27:48Z (GMT). No. of bitstreams: 1 Three Dimensional Tree Modeling Based on the Skeleton Path Optimization and Geometrical Shapes.pdf: 1783238 bytes, checksum: 78e5941b5dd4a843db7d0c416d3cf06d (MD5)

Wu, Xiaoming Cao, Yu Wang, Yu Li, Bing Yang, Haitao Raja, S. P. https://reunir.unir.net/handle/123456789/19079 2026-02-25T16:21:22Z

Posture Estimation of Curve Running Motion Using Nano-Biosensor and Machine Learning Wu, Xiaoming; Cao, Yu; Wang, Yu; Li, Bing; Yang, Haitao; Raja, S. P. Curve running is a common form of training and competition. Conducting research on posture estimation during curve running can provide more accurate training and competition data for athletes. However, due to the unique nature of curve running, traditional posture estimation methods neglect the temporal changes in athlete posture, resulting in a decrease in estimation accuracy. Therefore, a posture estimation method for curve running motion using nano-biosensor and machine learning is proposed. First, the motion parameters of humans are collected by nano-biosensor, and the posture coordinates are obtained preliminarily. Second, the posture coordinates are established according to the human motion parameters, and the curve running posture data is obtained and filtered to obtain more accurate data. Finally, the Bayesian network in machine learning is used to continuously track the posture, and a nonlinear equation is established to fuse the posture angle obtained by the sensor and the posture tracked by the Bayesian network, to realize the posture estimation of curve running motion. The results show that the proposed estimation method has a good motion posture estimation effect, and the hip joint estimation error, knee joint estimation error and ankle joint estimation error are all less than 5°, and the endpoint displacement estimation offset rate is less than 2%. It can realize accurate motion posture estimation of curve running motion, and has important application value in the field of track training. Submitted by Angela María Porras Ruiz (angela.porras@unir.net) on 2026-02-25T16:21:22Z No. of bitstreams: 1 Posture Estimation of Curve Running Motion Using Nano Biosensor and Machine Learning.pdf: 508410 bytes, checksum: 9f8033f27a6485bc48c75b4c6ca909f9 (MD5); Made available in DSpace on 2026-02-25T16:21:22Z (GMT). No. of bitstreams: 1 Posture Estimation of Curve Running Motion Using Nano Biosensor and Machine Learning.pdf: 508410 bytes, checksum: 9f8033f27a6485bc48c75b4c6ca909f9 (MD5)

Lin, Kuan-Cheng Lin, Ya-Hsuan Chen, Ya-Hsuan https://reunir.unir.net/handle/123456789/19078 2026-02-25T16:15:50Z

A Realtime Classroom Assessment System for Analysis of Students’ Evaluation of Teaching Through a Deep Learning and Emotional Contagion Mechanism Lin, Kuan-Cheng; Lin, Ya-Hsuan; Chen, Ya-Hsuan Student evaluations of teacher performance are often derived from end-of-semester assessments, significantly impacting the authenticity of teaching evaluations but failing to provide real-time feedback. In addition, teachers' emotional states affect student performance, including in terms of learning motivation and classroom participation, which reflect the students' emotional state. This teacher-student emotional contagion mechanism focuses on the interaction of teacher-student emotions and can be used to observe the quality of instructional performance. Therefore, automatically detecting teacher-student emotional interaction and then providing real-time class satisfaction feedback can provide teachers with a more effective basis for adjusting classroom content. This research proposes an end-to-end classroom real-time teaching evaluation system based on automatic facial-emotion recognition, which can accurately detect and directly analyze the emotions of students and teachers in streaming frames. The system consists of two parts: First, a YOLO model based on deep learning approaches is used to automatically detect the emotional states of teachers and students during the teaching process; Then, combining the emotional contagion mechanism with the teaching evaluation scale, teaching satisfaction can be predicted using a Long Short-Term Memory (LSTM) model to output a classroom satisfaction score within a fixed period. Further analysis of the testing dataset confirms that the model has a high reliability in predicting teaching satisfaction. Research results show the proposed system can achieve an emotional recognition accuracy rate of 98.1% for teachers and 99.5% for students based on the emotion datasets. Further development could potentially provide teachers with strategies to improve classroom teaching effectiveness, better understand students' emotions and learning motivation, and improve learning outcomes. Submitted by Angela María Porras Ruiz (angela.porras@unir.net) on 2026-02-25T16:15:50Z No. of bitstreams: 1 A Realtime Classroom Assessment System for Analysis.pdf: 410131 bytes, checksum: 7244389b1751d418174f081e546ef794 (MD5); Made available in DSpace on 2026-02-25T16:15:50Z (GMT). No. of bitstreams: 1 A Realtime Classroom Assessment System for Analysis.pdf: 410131 bytes, checksum: 7244389b1751d418174f081e546ef794 (MD5)

García Navarro, Alejandro L. Koneva, Nataliia Hernández, José Alberto Sánchez-Macián, Alfonso https://reunir.unir.net/handle/123456789/19077 2026-02-25T16:07:52Z

ChatGPT, Generative AI, Mathematical Problems, Wolfram Mathematica García Navarro, Alejandro L.; Koneva, Nataliia; Hernández, José Alberto; Sánchez-Macián, Alfonso In November 2022, ChatGPT v3.5 was announced to the world. Since then, Generative Artificial Intelligence (GAI) has appeared in the news almost daily, showing impressive capabilities at solving multiple tasks that have surprised the research community and the world in general. Indeed the number of tasks that ChatGPT and other Large Language Models (LLMs) can do are unimaginable, especially when dealing with natural text. Text generation, summarisation, translation, and transformation (into poems, songs, or other styles) are some of its strengths. However, when it comes to reasoning or mathematical calculations, ChatGPT finds difficulties. In this work, we compare different flavors of ChatGPT (v3.5, v4, and Wolfram GPT) at solving 20 mathematical tasks, from high school and first-year engineering courses. We show that GPT-4 is far more powerful than ChatGPT-3.5, and further that the use of Wolfram GPT can even slightly improve the results obtained with GPT-4 at these mathematical tasks. Submitted by Angela María Porras Ruiz (angela.porras@unir.net) on 2026-02-25T16:07:52Z No. of bitstreams: 1 On the Use of Large Language Models at Solving Math Problems.pdf: 416406 bytes, checksum: ec1a652ce78b2e3a3d30e6fb54f768e6 (MD5); Made available in DSpace on 2026-02-25T16:07:52Z (GMT). No. of bitstreams: 1 On the Use of Large Language Models at Solving Math Problems.pdf: 416406 bytes, checksum: ec1a652ce78b2e3a3d30e6fb54f768e6 (MD5)

Jeon, Gwangil Ahmed, Imran Han, Sanghoon https://reunir.unir.net/handle/123456789/19076 2026-02-25T16:01:01Z

A Sustainable Deep Learning Paradigm for Reliable Energy Prediction in Next-Gen Consumer Electronics Jeon, Gwangil; Ahmed, Imran; Han, Sanghoon In the rapidly evolving consumer electronics landscape, the imperative for sustainable energy solutions necessitates the development of accurate energy prediction methodologies. Traditional energy prediction models often fall short in accounting for the dynamic characteristics of renewable energy sources, particularly wind and solar. This limitation is pronounced in consumer electronics, where precise energy forecasting is pivotal for achieving optimal device performance and energy efficiency. To address this gap, we present a sustainable deep learning paradigm using Long Short-Term Memory (LSTM) networks to capture the complex temporal patterns inherent in renewable energy data. This paper introduces a novel and sustainable deep learning approach that significantly enhances energy prediction accuracy within the context of next-generation consumer electronics. By leveraging the capabilities of an LSTM-based model, we utilize an extensive dataset comprising hourly records of wind and solar energy production from the French grid since 2020. Our research addresses the inherent challenges in precise energy prediction, a cornerstone for efficient energy management and consumption optimization in emerging technology ecosystems. Through comprehensive data preprocessing, feature engineering, and rigorous training, the LSTM model demonstrates exceptional proficiency, achieving an impressive 82% accuracy in predicting energy production. This underscores its efficacy in capturing intricate temporal relationships and patterns within renewable energy data, facilitating its integration into next-generation consumer electronics. Our proposed paradigm addresses a critical need and paves the way for a future where accurate energy prediction fuels eco-conscious technology. In conclusion, this study contributes to a more sustainable energy landscape by advancing the development of reliable and efficient energy prediction methodologies for the evolving realm of next-generation consumer electronics. Submitted by Angela María Porras Ruiz (angela.porras@unir.net) on 2026-02-25T16:01:01Z No. of bitstreams: 1 A Sustainable Deep Learning Paradigm for Reliable Energy Prediction in Next Gen Consumer Electronics.pdf: 572480 bytes, checksum: 15dbecabf4998c3cebec7cfc3700902d (MD5); Made available in DSpace on 2026-02-25T16:01:01Z (GMT). No. of bitstreams: 1 A Sustainable Deep Learning Paradigm for Reliable Energy Prediction in Next Gen Consumer Electronics.pdf: 572480 bytes, checksum: 15dbecabf4998c3cebec7cfc3700902d (MD5)

Wang, Xiaoliang Zhou, Lian Li, Kuan-Ching Zheng, Shiqi Fan, Huijing https://reunir.unir.net/handle/123456789/19075 2026-02-25T15:46:42Z

IAtraj: Multi-Modal Trajectory Prediction Through Contextual Information Spatio-Temporal Interaction and Awareness Wang, Xiaoliang; Zhou, Lian; Li, Kuan-Ching; Zheng, Shiqi; Fan, Huijing Accurately and feasibly predicting the future trajectories of autonomous vehicles is a critically important task. However, this task faces significant challenges due to the variability of driving intentions and the complexity of social interactions. These challenges primarily arise from the need to understand one’s driving behaviors and model the interaction information of the surrounding environment. A substantial amount of research has been focused on integrating interaction information from the surrounding environment, mainly using raster images or High-Definition maps (HD maps). However, the real-time update of environmental maps and the high computational cost associated with processing interaction information using compatible technologies such as vision have become limiting factors. Additionally, ineffective simulation and modeling of real driving scenarios, coupled with inadequate understanding of contextual environmental information, result in lower prediction accuracy. To overcome these challenges, we propose a multi-modal trajectory prediction model based on sequence modeling namely IAtraj, incorporating multiple attention mechanisms, focuses on the three critical elements in real traffic scenarios: the target agent’s historical trajectory, effective interactions with neighboring vehicles, and lane supervision and retention strategies. To better model these elements, we design modules for Temporal Interaction (TI), Spatial Interaction (SI), and Lane Awareness (LA). Through extensive experiments conducted on the publicly available nuScenes dataset, IAtraj exhibits outstanding performance, successfully addressing the challenges of temporal dependencies in trajectory sequences and the representation of scene changes. Finally, comprehensive ablation experiments validate the effectiveness of each significant module, reinforcing the reliability and robustness of IAtraj in dealing with complex traffic scenarios. Submitted by Angela María Porras Ruiz (angela.porras@unir.net) on 2026-02-25T15:46:42Z No. of bitstreams: 1 IAtraj Multi-Modal Trajectory Prediction Through Contextual Information Spatio-Temporal Interaction and Awareness.pdf: 1423128 bytes, checksum: 60d5f9e813b3a6ff8100670b5134b89d (MD5); Made available in DSpace on 2026-02-25T15:46:42Z (GMT). No. of bitstreams: 1 IAtraj Multi-Modal Trajectory Prediction Through Contextual Information Spatio-Temporal Interaction and Awareness.pdf: 1423128 bytes, checksum: 60d5f9e813b3a6ff8100670b5134b89d (MD5)

Iqbal, Muhammad Faisal Buland Ullah, Aman Alhomoud, Ahmed Hussain, Tariq Attar, Razaz Waheeb Ouyang, Jianquan Alnfiai, Mrim M. Hatamleh, Wesam Atef https://reunir.unir.net/handle/123456789/19073 2026-02-25T15:27:47Z

A 2D Clustering Based Hotspot Identification Approach for Spatio-Temporal Crime Prediction Iqbal, Muhammad Faisal Buland; Ullah, Aman; Alhomoud, Ahmed; Hussain, Tariq; Attar, Razaz Waheeb; Ouyang, Jianquan; Alnfiai, Mrim M.; Hatamleh, Wesam Atef This research introduces a method for predicting where crimes will occur based on clustering activity in the area. Hotspots, or locations with a disproportionately high number of crimes, are located by a combination of spatial and temporal grouping methods employed by this strategy. Crime forecasting models use these hotspots to predict where crimes will occur. The approach's efficacy is tested using actual crime data, and it successfully predicts future crimes in high-crime zones. Law enforcement agencies can use the proposed method to protect the public better, and it shows promise as a tool for crime prediction. Academic research into the topic of foreseeing criminal behavior is a newer development. Researchers in this discipline have discovered that criminal behavior has universal patterns. These patterns may help law enforcement agencies plan for criminal activities. Predictive policing, hotspot analysis, and geographical profiling are examples of when crime forecasting has been useful. Several aspects of the census, such as the average yearly income and literacy rate, are related to the prevalence of crime in a certain area. Indicators of potentially criminal behavior, these characteristics may be seen as markers. This investigation aims to discover if any clues can be gleaned from past criminal behavior that may be utilized to forecast future criminal behavior. Using machine learning and 2-D Hotspot analysis, we propose a method for the spatiotemporal prediction of criminal activity within the scope of this study. Clustering is a method used in 2-dimensional hotspot analysis. Methods of modern categorization, both with and without hotspot analysis, are used to evaluate the suggested model's efficacy. It is found that the model that incorporates hotspot analysis performs better than the one that does not. Submitted by Angela María Porras Ruiz (angela.porras@unir.net) on 2026-02-25T15:27:47Z No. of bitstreams: 1 A 2D Clustering Based Hotspot Identification Approach for Spatio-Temporal Crime Prediction.pdf: 561649 bytes, checksum: 3df958bf45fed378586ac4711cc20fc4 (MD5); Made available in DSpace on 2026-02-25T15:27:47Z (GMT). No. of bitstreams: 1 A 2D Clustering Based Hotspot Identification Approach for Spatio-Temporal Crime Prediction.pdf: 561649 bytes, checksum: 3df958bf45fed378586ac4711cc20fc4 (MD5)