Aymen I. Zreikat
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Aymen I. Zreikat has a total of 66 co-authors. Here are the top co-authors by number of shared publications:
- Ersin Elbasi
College of Engineering and Technology, American University of the Middle East, Egaila 54200, Kuwait
10 shared publications - Ahmad Abadleh
Computer Science Department, College of Information Technology, Mutah University, Karak 61710, Jordan
5 shared publications - Ahmed Younes Shdefat
College of Engineering and Technology, American University of the Middle East, Egaila, Kuwait
5 shared publications - Elda Cina
College of Engineering and Technology, American University of the Middle East, Egaila 54200, Kuwait
5 shared publications - Ahmet Ercan Topcu
College of Engineering and Technology, American University of the Middle East
4 shared publications - Bassam A. Y. Alqaralleh
College of Business Administration, American University of the Middle East, Egaila 54200, Kuwait
4 shared publications
Latest Publications
- Open Access
Evidence-Aware Human-in-the-Loop LLM Review for Requirements-to-Planning Decisions
Large language models can produce fluent requirements refinements and planning artifacts while still leaving information unresolved for implementation, testing, or planning commitment. This paper presents ReqPlan-Eval, an evidence-aware human-in-the-loop architecture that connects NFR disagreement, weak-word cues, planning-relevant ambiguity, and role-specialized hypotheses to inspectable planning-support records and configurable review routes. The empirical study evaluates the principal mechanisms and role-based routing signals on separate datasets; it does not constitute an end-to-end evaluation of the complete pipeline on a common set of requirements. A held-out 500-requirement NFR diagnostic achieved exact-match accuracy of 0.716, micro F1 of 0.702, and quality accuracy of 0.874. Pattern-aware arbitration increased weak-word specificity from 0.572 to 0.676 and review precision from 0.678 to 0.724, while component/goal gating increased ambiguity specificity from 0.380 to 0.908 and F1 from 0.748 to 0.855; both mechanisms lost recall. In Experiment 3, four role-specialized outputs were compared with a model-seeded reference reviewed and adjudicated by three human reviewers. On the 20 public stories, role-level exact agreement for validation_needed was 0.825, and a two-or-more-vote policy reviewed 12 stories and captured 12 of 14 reference positives without routing any of the six reference negatives. Open-ended planning artifacts showed very low exact normalized-item overlap with the reference for tasks, acceptance criteria, and test ideas, so no semantic-agreement claim is made for those artifacts. Seven external participants provided initial face-validity evidence for selective review and human control. The results support the evaluated component mechanisms and selective review routing under the frozen configurations, but they do not establish end-to-end workflow effectiveness, model invariance, better planning decisions, or industrial effectiveness.
Open AccessEvidence-Aware Human-in-the-Loop LLM Review for Requirements-to-Planning Decisions
Large language models can produce fluent requirements refinements and planning artifacts while still leaving information unresolved for implementation, testing, or planning commitment. This paper presents ReqPlan-Eval, an evidence-aware human-in-the-loop architecture that connects NFR disagreement, weak-word cues, planning-relevant ambiguity, and role-specialized hypotheses to inspectable planning-support records and configurable review routes. The empirical study evaluates the principal mechanisms and role-based routing signals on separate datasets; it does not constitute an end-to-end evaluation of the complete pipeline on a common set of requirements. A held-out 500-requirement NFR diagnostic achieved exact-match accuracy of 0.716, micro F1 of 0.702, and quality accuracy of 0.874. Pattern-aware arbitration increased weak-word specificity from 0.572 to 0.676 and review precision from 0.678 to 0.724, while component/goal gating increased ambiguity specificity from 0.380 to 0.908 and F1 from 0.748 to 0.855; both mechanisms lost recall. In Experiment 3, four role-specialized outputs were compared with a model-seeded reference reviewed and adjudicated by three human reviewers. On the 20 public stories, role-level exact agreement for validation_needed was 0.825, and a two-or-more-vote policy reviewed 12 stories and captured 12 of 14 reference positives without routing any of the six reference negatives. Open-ended planning artifacts showed very low exact normalized-item overlap with the reference for tasks, acceptance criteria, and test ideas, so no semantic-agreement claim is made for those artifacts. Seven external participants provided initial face-validity evidence for selective review and human control. The results support the evaluated component mechanisms and selective review routing under the frozen configurations, but they do not establish end-to-end workflow effectiveness, model invariance, better planning decisions, or industrial effectiveness.
- 2 June 2026
- Published by Springer Nature
- p. 73-89
Scope of AI in Future Farming
- Open Access
- 27 April 2026
- Published by Public Library of Science (PLoS) in PLOS ONE
- Vol. 21 (4) , e0346128
MedSpectralNet: A lightweight convolutional neural network architecture for multi-modal image classification
Medical image classification requires models that effectively capture both fine-grained local patterns and global anatomical structures while maintaining computational efficiency for clinical deployment. Although state-of-the-art models such as MedMamba utilize State-Space Models (SSMs) to balance accuracy and efficiency, their sequential operations limit parallelism and increase runtime. To overcome these limitations, we propose MedSpectralNet, a lightweight Convolutional Neural Network (CNN) architecture that approximates self-attention with linear complexity to efficiently extract multi-frequency features. The model introduces a dual-stream feature extractor that processes global and local information in parallel, and a ContextGate block that adaptively fuses multi-scale representations. MedSpectralNet is evaluated across six benchmark datasets from MedMNIST (including BloodMNIST, BreastMNIST, DermaMNIST, PneumoniaMNIST, OrganCMNIST, and OrganSMNIST), MedSpectralNet achieves an average accuracy of 93.7% on OrganCMNIST and 98.0% on BloodMNIST, showing 1–4.3% relative accuracy gains when compared to larger transformer-based models. Importantly, it delivers this performance with only 8.5 million parameters, representing approximately 60% fewer parameters than MedMamba-T, which requires 14.5 million parameters. MedSpectralNet has also achieved high AUC values up to 0.999 across multiple classes, demonstrating state-of-the-art accuracy with substantially reduced computational cost and improved parallelization, which makes MedSpectralNet well-suited for real-time and resource-constrained classification-based medical applications.
- 27 April 2026
- Published by Public Library of Science (PLoS) in PLOS ONE
- Vol. 21 (4) , e0346128
Open AccessMedSpectralNet: A lightweight convolutional neural network architecture for multi-modal image classification
Medical image classification requires models that effectively capture both fine-grained local patterns and global anatomical structures while maintaining computational efficiency for clinical deployment. Although state-of-the-art models such as MedMamba utilize State-Space Models (SSMs) to balance accuracy and efficiency, their sequential operations limit parallelism and increase runtime. To overcome these limitations, we propose MedSpectralNet, a lightweight Convolutional Neural Network (CNN) architecture that approximates self-attention with linear complexity to efficiently extract multi-frequency features. The model introduces a dual-stream feature extractor that processes global and local information in parallel, and a ContextGate block that adaptively fuses multi-scale representations. MedSpectralNet is evaluated across six benchmark datasets from MedMNIST (including BloodMNIST, BreastMNIST, DermaMNIST, PneumoniaMNIST, OrganCMNIST, and OrganSMNIST), MedSpectralNet achieves an average accuracy of 93.7% on OrganCMNIST and 98.0% on BloodMNIST, showing 1–4.3% relative accuracy gains when compared to larger transformer-based models. Importantly, it delivers this performance with only 8.5 million parameters, representing approximately 60% fewer parameters than MedMamba-T, which requires 14.5 million parameters. MedSpectralNet has also achieved high AUC values up to 0.999 across multiple classes, demonstrating state-of-the-art accuracy with substantially reduced computational cost and improved parallelization, which makes MedSpectralNet well-suited for real-time and resource-constrained classification-based medical applications.
- Open Access
Evaluating Machine Learning Classifiers in Detecting Cyberattacks
This study aims to develop a machine learning model that can accurately detect cyberattacks. We compare the performance of Support Vector Machine (SVM), Logistic Regression (LR), and Random Forest (RF) in predicting cyberattacks. Furthermore, we investigate whether using Information Gain Attribute Evaluation (IGAE) for feature selection improves the performance of the algorithms. This work provides a clear comparison of the algorithms and shows the most suitable one for classifying cyberattacks. In addition, this study combines LR and RF using a voting classifier along with IGAE and compares its performance with that of the rest of the algorithms. We investigate whether combining algorithms increases the accuracy of the results. The results show that the most accurate algorithm is RF, followed by LR and SVM. Contrary to initial expectations, the findings further indicate that the application of IGAE marginally reduces algorithm accuracy across the tested classifiers, suggesting that feature selection through information gain is not universally beneficial in cyberattack detection tasks. These findings contribute to the growing body of knowledge on effective machine learning methodologies for cybersecurity applications.
Open AccessEvaluating Machine Learning Classifiers in Detecting Cyberattacks
This study aims to develop a machine learning model that can accurately detect cyberattacks. We compare the performance of Support Vector Machine (SVM), Logistic Regression (LR), and Random Forest (RF) in predicting cyberattacks. Furthermore, we investigate whether using Information Gain Attribute Evaluation (IGAE) for feature selection improves the performance of the algorithms. This work provides a clear comparison of the algorithms and shows the most suitable one for classifying cyberattacks. In addition, this study combines LR and RF using a voting classifier along with IGAE and compares its performance with that of the rest of the algorithms. We investigate whether combining algorithms increases the accuracy of the results. The results show that the most accurate algorithm is RF, followed by LR and SVM. Contrary to initial expectations, the findings further indicate that the application of IGAE marginally reduces algorithm accuracy across the tested classifiers, suggesting that feature selection through information gain is not universally beneficial in cyberattack detection tasks. These findings contribute to the growing body of knowledge on effective machine learning methodologies for cybersecurity applications.
- 28 February 2026
- Published by World Scientific Pub Co Pte Ltd in International Journal of Software Engineering and Knowledge Engineering
- Vol. 36 (11) , 1647-1671
A Lightweight Real-Time Detection and Recovery Framework for Deauthentication Attacks in WPA3-SAE Networks
Wireless networks operating under WPA3-SAE protocols face continuous threats from deauthentication attacks which compromise their reliability and availability. Weaknesses in Wi-Fi management frames allow attackers to force legitimate users offline, resulting in denial-of-service (DoS) conditions. This paper presents the Intrusion Detection and Recovery of Deauthentication Frames (IDR-DF) framework which provides real-time detection and recovery capabilities for WPA3 networks. The timing-based anomaly detection system of IDR-DF detects harmful deauthentication patterns to start automatic recovery procedures which enable clients to maintain network connectivity without interruptions. The framework is implemented as a Python tool that runs on Linux in a WPA3 testbed. In a WPA3-SAE testbed, IDR-DF achieved 100% detection of baseline unicast deauthentication attacks with less than 2% false positives and an average client recovery time of 1.2[ ]s. Under more advanced multi-client spoofing and timing-jitter attacks, the detection engine maintained a 96.3% detection rate. The proposed IDR-DF system provides enhanced adaptability together with minimal computational overhead and seamless WPA3 Management Frame Protection (MFP) integration. The experimental findings validate the proposed framework as an effective solution to protect next-generation wireless networks against real-world threats.
- 28 February 2026
- Published by World Scientific Pub Co Pte Ltd in International Journal of Software Engineering and Knowledge Engineering
- Vol. 36 (11) , 1647-1671
A Lightweight Real-Time Detection and Recovery Framework for Deauthentication Attacks in WPA3-SAE Networks
Wireless networks operating under WPA3-SAE protocols face continuous threats from deauthentication attacks which compromise their reliability and availability. Weaknesses in Wi-Fi management frames allow attackers to force legitimate users offline, resulting in denial-of-service (DoS) conditions. This paper presents the Intrusion Detection and Recovery of Deauthentication Frames (IDR-DF) framework which provides real-time detection and recovery capabilities for WPA3 networks. The timing-based anomaly detection system of IDR-DF detects harmful deauthentication patterns to start automatic recovery procedures which enable clients to maintain network connectivity without interruptions. The framework is implemented as a Python tool that runs on Linux in a WPA3 testbed. In a WPA3-SAE testbed, IDR-DF achieved 100% detection of baseline unicast deauthentication attacks with less than 2% false positives and an average client recovery time of 1.2[ ]s. Under more advanced multi-client spoofing and timing-jitter attacks, the detection engine maintained a 96.3% detection rate. The proposed IDR-DF system provides enhanced adaptability together with minimal computational overhead and seamless WPA3 Management Frame Protection (MFP) integration. The experimental findings validate the proposed framework as an effective solution to protect next-generation wireless networks against real-world threats.
- Open Access
- 25 February 2026
- Published by MDPI AG in Preprints.org
Coexistence of 5G-NR High Band/Mid-Band/Low-Band Technologies with Performance Evaluation
Wireless communications face both opportunities and challenges due to the coexistence of 5G New Radio (NR) high-band, 5G mid-band, and 5G low-band technologies. Each technology uses both licensed and unlicensed spectrum to operate in separate frequency bands. For example, 5G NR uses the high-band of 24+ GHz, the mid-band of 2-6 GHz, or the low-band of less than 2 GHz, including the 5 GHz band via Licensed-Assisted Access (LAA). With the use of sophisticated coexistence mechanisms and optimization techniques, this 5G coexistence scenario in shared spectrum can be effectively managed. These strategies are essential for boosting network capacity, reducing latency, and ensuring fair spectrum use across different wireless technologies. This work provides a comprehensive system-level evaluation of multi-band coexistence and offloading strategies under realistic deployment assumptions. The simulation results confirm the effectiveness of the proposed model, showing that spectrum sharing and coexistence among these technologies deliver scalable and robust performance in heterogeneous service environments. This approach enables efficient load balancing across the entire network and highlights the need for additional features to achieve further performance gains.
- 25 February 2026
- Published by MDPI AG in Preprints.org
Open AccessCoexistence of 5G-NR High Band/Mid-Band/Low-Band Technologies with Performance Evaluation
Wireless communications face both opportunities and challenges due to the coexistence of 5G New Radio (NR) high-band, 5G mid-band, and 5G low-band technologies. Each technology uses both licensed and unlicensed spectrum to operate in separate frequency bands. For example, 5G NR uses the high-band of 24+ GHz, the mid-band of 2-6 GHz, or the low-band of less than 2 GHz, including the 5 GHz band via Licensed-Assisted Access (LAA). With the use of sophisticated coexistence mechanisms and optimization techniques, this 5G coexistence scenario in shared spectrum can be effectively managed. These strategies are essential for boosting network capacity, reducing latency, and ensuring fair spectrum use across different wireless technologies. This work provides a comprehensive system-level evaluation of multi-band coexistence and offloading strategies under realistic deployment assumptions. The simulation results confirm the effectiveness of the proposed model, showing that spectrum sharing and coexistence among these technologies deliver scalable and robust performance in heterogeneous service environments. This approach enables efficient load balancing across the entire network and highlights the need for additional features to achieve further performance gains.
- Open Access
- 16 February 2026
- Published by Springer Nature in Scientific Reports
- Vol. 16 (1)
An intelligent healthcare framework for hepatocellular carcinoma diagnosis based on aggregated learners from biomedical data utilising explainable artificial intelligence
In recent days, biomedical data mining and machine learning (ML) technologies have transformed the healthcare sector, which utilises cutting-edge medical innovative tools to develop effective decision support systems for disease diagnosis and health informatics. Liver cancer (LC) is a major contributor to the global cancer problem. Incidence rates of this disease have improved in several countries in the past decades. As the main histological kind of LC, hepatocellular carcinoma (HCC) constitutes the large majority of LC diagnoses and deaths. HCC is one of the primary reasons for cancer occurrence and fatality. Initial diagnosis of HCC remains the main aim in improving the poor diagnosis of this type of LC. Recognising HCC at an initial stage is frequently related to improved treatment possibilities for patients with small and symptomless tumours. Several artificial intelligence (AI) methods are considered advanced methods for processing and handling composite multimodal data ranging from repetitive clinical variables to higher-resolution medical images. This paper presents a Hepatocellular Carcinoma Diagnosis based on an Aggregated Learners Utilising Explainable Artificial Intelligence (HCDAL-XAI) model from biomedical data. The primary purpose of the HCDAL-XAI model is to deliver an accurate detection model for initial diagnosis and efficient treatment of HCC using progressive methods. Initially, the data pre-processing step uses min-max normalisation. Furthermore, the HCDAL-XAI model employs an ensemble of a sparse autoencoder (SAE), gated recurrent unit (GRU), and deep belief network (DBN) for the classification process. Lastly, the explainable AI (XAI) model employs SHapley Additive exPlanations (SHAP) to enhance the reliability of AI methods by making their decision-making processes understandable to humans. The comparison analysis of the HCDAL-XAI methodology portrayed a greater accuracy value of 98.18% over existing models under the HCC dataset.
- 16 February 2026
- Published by Springer Nature in Scientific Reports
- Vol. 16 (1)
Open AccessAn intelligent healthcare framework for hepatocellular carcinoma diagnosis based on aggregated learners from biomedical data utilising explainable artificial intelligence
In recent days, biomedical data mining and machine learning (ML) technologies have transformed the healthcare sector, which utilises cutting-edge medical innovative tools to develop effective decision support systems for disease diagnosis and health informatics. Liver cancer (LC) is a major contributor to the global cancer problem. Incidence rates of this disease have improved in several countries in the past decades. As the main histological kind of LC, hepatocellular carcinoma (HCC) constitutes the large majority of LC diagnoses and deaths. HCC is one of the primary reasons for cancer occurrence and fatality. Initial diagnosis of HCC remains the main aim in improving the poor diagnosis of this type of LC. Recognising HCC at an initial stage is frequently related to improved treatment possibilities for patients with small and symptomless tumours. Several artificial intelligence (AI) methods are considered advanced methods for processing and handling composite multimodal data ranging from repetitive clinical variables to higher-resolution medical images. This paper presents a Hepatocellular Carcinoma Diagnosis based on an Aggregated Learners Utilising Explainable Artificial Intelligence (HCDAL-XAI) model from biomedical data. The primary purpose of the HCDAL-XAI model is to deliver an accurate detection model for initial diagnosis and efficient treatment of HCC using progressive methods. Initially, the data pre-processing step uses min-max normalisation. Furthermore, the HCDAL-XAI model employs an ensemble of a sparse autoencoder (SAE), gated recurrent unit (GRU), and deep belief network (DBN) for the classification process. Lastly, the explainable AI (XAI) model employs SHapley Additive exPlanations (SHAP) to enhance the reliability of AI methods by making their decision-making processes understandable to humans. The comparison analysis of the HCDAL-XAI methodology portrayed a greater accuracy value of 98.18% over existing models under the HCC dataset.
- Open Access
Evaluating LLMs for Source Code Generation and Summarization Using Machine Learning Classification and Ranking
The recent use of large language models (LLMs) in code generation and code summarization tasks has been widely adopted by the software engineering community. New LLMs are emerging regularly with improved functionalities, efficiency, and expanding data that allow models to learn more effectively. The lack of guidelines for selecting the right LLMs for coding tasks makes the selection a subjective choice by developers rather than a choice built on code complexity, code correctness, and linguistic similarity analysis. This research investigates the use of machine learning classification and ranking methods to select the best-suited open-source LLMs for code generation and code summarization tasks. This work conducts a comparison experiment on four open-source LLMs (Mistral, CodeLlama, Gemma 2, and Phi-3) and uses the MBPP coding question dataset to analyze code-generated outputs in terms of code complexity, maintainability, cyclomatic complexity, code structure, and LLM perplexity by collecting these as a set of features. An SVM classification problem is conducted on the highest correlated feature pairs, where the models are evaluated through performance metrics, including accuracy, area under the ROC curve (AUC), precision, recall, and F1 scores. The RankNet ranking methodology is used to evaluate code summarization model capabilities by measuring ROUGE and BERTScore accuracies between LLM code-generated summaries and the coding questions used from the dataset. The study results show a maximum accuracy of 49% for the code generation experiment, with the highest AUC score reaching 86% among the top four correlated feature pairs. The highest precision score reached is 90%, and the recall score reached up to 92%. Code summarization experiment results show Gemma 2 scored a 1.93 RankNet win probability score, and represented the highest ranking reached among other models. The phi3 model was the second-highest ranking with a 1.66 score. The research highlights the potential of machine learning to select LLMs based on coding metrics and paves the way for advancements in terms of accuracy, dataset diversity, and exploring other machine learning algorithms for other researchers.
Open AccessEvaluating LLMs for Source Code Generation and Summarization Using Machine Learning Classification and Ranking
The recent use of large language models (LLMs) in code generation and code summarization tasks has been widely adopted by the software engineering community. New LLMs are emerging regularly with improved functionalities, efficiency, and expanding data that allow models to learn more effectively. The lack of guidelines for selecting the right LLMs for coding tasks makes the selection a subjective choice by developers rather than a choice built on code complexity, code correctness, and linguistic similarity analysis. This research investigates the use of machine learning classification and ranking methods to select the best-suited open-source LLMs for code generation and code summarization tasks. This work conducts a comparison experiment on four open-source LLMs (Mistral, CodeLlama, Gemma 2, and Phi-3) and uses the MBPP coding question dataset to analyze code-generated outputs in terms of code complexity, maintainability, cyclomatic complexity, code structure, and LLM perplexity by collecting these as a set of features. An SVM classification problem is conducted on the highest correlated feature pairs, where the models are evaluated through performance metrics, including accuracy, area under the ROC curve (AUC), precision, recall, and F1 scores. The RankNet ranking methodology is used to evaluate code summarization model capabilities by measuring ROUGE and BERTScore accuracies between LLM code-generated summaries and the coding questions used from the dataset. The study results show a maximum accuracy of 49% for the code generation experiment, with the highest AUC score reaching 86% among the top four correlated feature pairs. The highest precision score reached is 90%, and the recall score reached up to 92%. Code summarization experiment results show Gemma 2 scored a 1.93 RankNet win probability score, and represented the highest ranking reached among other models. The phi3 model was the second-highest ranking with a 1.66 score. The research highlights the potential of machine learning to select LLMs based on coding metrics and paves the way for advancements in terms of accuracy, dataset diversity, and exploring other machine learning algorithms for other researchers.
- Open Access
- 10 February 2026
- Published by MDPI AG in Diagnostics
- Vol. 16 (4) , 530
Dynamic Mode Decomposition-Based Clustered Pattern Projection for Reliable Alzheimer’s Disease Detection from EEG
Background/Objectives: Detecting Alzheimer’s disease (AD) from normal aging using eyes-open (EO) EEG is challenging due to stimulus-driven nonstationarity and fragmented oscillatory responses. This study aims to determine whether prototype-based representations derived from Dynamic Mode Decomposition (DMD) can improve AD detection from EO photostimulation EEG. Methods: We propose a DMD-based framework termed DMD-based Clustered Pattern Projection (DMD-CPP). Segment-wise DMD representations were clustered to learn class-specific medoid prototypes, and each EEG epoch was encoded as a vector of cosine-similarity coordinates with respect to these prototypes. A linear SVM classifier was trained on the resulting DMD-CPP features and evaluated under strict leave-one-subject-out validation. Results: The DMD-CPP model achieved competitive classification accuracy and, importantly, enhanced margin-based reliability. In EO photostimulation, AD versus healthy control classification showed a pronounced improvement, characterized by wider and more asymmetric decision margins, particularly assigning low confidence to normal epochs misclassified as AD. Tasks involving frontotemporal dementia also showed improvement, although the effect was less pronounced than for AD. Conclusions: Clustering-based pattern projection has been shown to stabilize EEG dynamics and provide an interpretable, confidence-aware feature representation. These findings suggest that DMD-CPP offers a promising framework for reliable AD detection from EO EEG, where conventional spectral methods typically struggle.
- 10 February 2026
- Published by MDPI AG in Diagnostics
- Vol. 16 (4) , 530
Open AccessDynamic Mode Decomposition-Based Clustered Pattern Projection for Reliable Alzheimer’s Disease Detection from EEG
Background/Objectives: Detecting Alzheimer’s disease (AD) from normal aging using eyes-open (EO) EEG is challenging due to stimulus-driven nonstationarity and fragmented oscillatory responses. This study aims to determine whether prototype-based representations derived from Dynamic Mode Decomposition (DMD) can improve AD detection from EO photostimulation EEG. Methods: We propose a DMD-based framework termed DMD-based Clustered Pattern Projection (DMD-CPP). Segment-wise DMD representations were clustered to learn class-specific medoid prototypes, and each EEG epoch was encoded as a vector of cosine-similarity coordinates with respect to these prototypes. A linear SVM classifier was trained on the resulting DMD-CPP features and evaluated under strict leave-one-subject-out validation. Results: The DMD-CPP model achieved competitive classification accuracy and, importantly, enhanced margin-based reliability. In EO photostimulation, AD versus healthy control classification showed a pronounced improvement, characterized by wider and more asymmetric decision margins, particularly assigning low confidence to normal epochs misclassified as AD. Tasks involving frontotemporal dementia also showed improvement, although the effect was less pronounced than for AD. Conclusions: Clustering-based pattern projection has been shown to stabilize EEG dynamics and provide an interpretable, confidence-aware feature representation. These findings suggest that DMD-CPP offers a promising framework for reliable AD detection from EO EEG, where conventional spectral methods typically struggle.
- 29 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-5
Machine Learning-Based Beam Forming Prediction for Energy-Efficient 6G mmWave
Future 6G networks are expected to rely heavily on mmWave and Terahertz (THz) communication, where accurate beam alignment is essential but computationally expensive. This paper proposes a machine learning framework for predictive beamforming that significantly reduces alignment latency by forecasting the optimal beam indices based on contextual and RF-layer features. The paper utilizes two datasets: a context-based synthetic dataset for simulating user movement and positioning, and a DeepMIMO-inspired dataset for realistic wireless propagation modeling. The evaluation assesses three models, including Random Forest (RF), XGBoost, and a simulated LSTM, to determine their performance through Top-K accuracy and inference time measurements. The results demonstrate that XGBoost strikes an optimal trade-off between accuracy and latency, achieving 21% Top-5 accuracy while maintaining inference times below one millisecond. The proposed framework provides practical real-time beam prediction capabilities, which make it suitable for deployment at the edge and user side in ultra-dense 6G networks.
- 29 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-5
Machine Learning-Based Beam Forming Prediction for Energy-Efficient 6G mmWave
Future 6G networks are expected to rely heavily on mmWave and Terahertz (THz) communication, where accurate beam alignment is essential but computationally expensive. This paper proposes a machine learning framework for predictive beamforming that significantly reduces alignment latency by forecasting the optimal beam indices based on contextual and RF-layer features. The paper utilizes two datasets: a context-based synthetic dataset for simulating user movement and positioning, and a DeepMIMO-inspired dataset for realistic wireless propagation modeling. The evaluation assesses three models, including Random Forest (RF), XGBoost, and a simulated LSTM, to determine their performance through Top-K accuracy and inference time measurements. The results demonstrate that XGBoost strikes an optimal trade-off between accuracy and latency, achieving 21% Top-5 accuracy while maintaining inference times below one millisecond. The proposed framework provides practical real-time beam prediction capabilities, which make it suitable for deployment at the edge and user side in ultra-dense 6G networks.