Abstract— Crime prediction has gained increasing attention due to the growing availability of historical crime data and the need for data-driven decision-making in public safety. This study presents a comparative analysis of Long Short-Term Memory (LSTM) architectures for predicting the exact occurrence time of crimes based on temporal patterns. Three LSTM-based models are evaluated: Vanilla LSTM, Stacked LSTM, and Bidirectional LSTM.

The proposed approach integrates time-based features and lag features to capture temporal dependencies within crime data. Model performance is assessed using standard regression metrics, including Mean Absolute Error (MAE), Mean Squared Error (MSE), and Root Mean Squared Error (RMSE). Experimental results indicate that deeper LSTM architectures combined with temporal lag information improve prediction accuracy compared to the baseline model.

This study demonstrates the effectiveness of LSTM-based models for crime occurrence prediction and provides insights into selecting suitable deep learning architectures for time-series crime analysis, supporting the development of more reliable tools for proactive crime prevention.

Abstract—

The verification of Arabic handwritten signatures is an important research area in the fields of computer vision, machine learning, and deep learning. It is essential to distinguish between genuine and forged signatures due to its significance in preventing fraud and forgery, as well as in various legal contexts. Handwritten signatures vary due to several influencing factors, such as mood, writing surface, and writing instrument. In this research paper, three different datasets were utilized: the first dataset (Dataset1) is an English signature dataset known as CEDAR; the second dataset (Dataset2) is an Arabic signature dataset specifically created for this research; and the third dataset (Dataset3) combines the CEDAR dataset with the Arabic signatures. These datasets were utilized to develop a transfer learning model for signature recognition called MobileNet-V2, which is a deep neural network (DNN) model developed by Google, aimed at achieving a balance between accuracy and performance speed. It is primarily designed for image classification tasks in resource-constrained environments. Standard training and k-fold cross-validation (CV) methods were applied to evaluate the model's performance. The model achieved an accuracy of 97.74% on the CEDAR dataset, 98.49% on the Arabic dataset, and 99.49% on the combined dataset.

This paper presents a comprehensive study of modelling human body blockage (the most critical challenges in fifth-generation (5G)) effects on indoor millimetre wave (mmWave) communication links at 32.5 GHz, a key frequency for 5G networks. Through controlled experiments in a laboratory environment, we analyse signal attenuation as a human subject obstructs the line-of-sight (LOS) path between transmitter and receiver, recording received power at incremental positions. To model the observed phenomena, we propose a hybrid framework integrating deterministic and statistical components: (1) a modified Double Knife-Edge Diffraction (DKED) model with Gaussian-shaped blockage attenuation (20.8 dB peak at full blockage) and reflection-induced signal enhancement (−15.0 dB peak from nearby objects), and (2) a log-normal shadowing component (σ = 11.8 dB) capturing environmental randomness. Our results reveal strong agreement between simulations and measurements, achieving a mean absolute error of 3.2 dB and a correlation coefficient R² = 0.89. The analysis demonstrates that human-induced diffraction dominates near the LOS centre, while multipath reflections significantly alter signal strength at peripheral positions. We further derive practical guidelines for 5G network design, recommending a 44.4 dB link budget safety margin to account for combined blockage and shadowing effects. This work advances indoor mmWaves channel modelling by unifying physics-based diffraction analysis with empirical reflection characterization, the framework achieves strong experimental validation and offers actionable insights for 5G network design

The advent of 5G networks has revolutionized wireless communications by unlocking unprecedented data rates through millimeter-wave (mmWave) frequencies. However, the short wavelengths of mmWave signals (e.g., 32.5 GHz) make them highly vulnerable to obstructions, particularly human blockage, posing significant challenges for reliable link prediction and network planning. Existing models often oversimplify human-induced attenuation, limiting their accuracy in real-world scenarios. This work addresses this gap by proposing a cylindrical diffraction model to quantify human blockage effects at 32.5 GHz—the first application of such a model at this frequency. Through controlled experiments, we measured signal degradation as a human subject progressively blocked a 2-meter mmWave link, revealing a sharp decline in received power from −41.2 dBm (no blockage) to −69.7 dBm (full blockage). The cylindrical model demonstrated strong alignment with empirical trends, accurately capturing the nonlinear increase in attenuation as the human approached the line-of sight path. Notably, the model matched baseline measurements within 1.4 dB and predicted full-blockage loss within 7 dB of observed values, despite inherent simplifications. This study underscores the efficacy of cylindrical modelling for mmWave blockage prediction while highlighting critical refinements needed for practical deployment, such as incorporating material properties and antenna radiation patterns. By bridging theoretical and empirical insights, our work provides a foundational framework for enhancing 5G/6G network resilience in human-dense environments, ensuring robust performance for high-data-rate applications. 

This paper addresses the critical challenge of human-induced signal attenuation in millimeter-wave (mmWave) communications, a key concern for fifth-generation (5G) network reliability in indoor environments. Our study introduces a simplified model to quantify the impact of human body blockage on indoor communication links at a frequency of 32.5 GHz., a frequency relevant to 5G systems. The influence of nearby scattering objects is investigated through experimental measurements involving a human body. Key wave propagation phenomena, including diffraction, are considered for each scattering object. The Double Knife-Edge Diffraction (DKED) model is used to estimate the attenuation caused by the human body (to estimate blockage losses). Through controlled experiments with human subjects, we systematically analyze how scattering objects and body positioning influence signal propagation. The model's performance is validated by comparing simulation results with experimental data. The findings show that the proposed model effectively predicts signal attenuation in indoor environments, providing valuable insights for future studies on human presence effects in fifth-generation (5G) communication systems. Keywords: 5G, DKED, diffraction, human shadowing, millimeter-wave, blockage.

-This paper examines the effect of human body blockage on signal propagation (millimeter-wave (mmWave) signal propagation) in indoor environments links at 32.5 GHz (a critical frequency for fifth-generation (5G) network), with a particular focus on the diffraction effects caused by the human body, where diffraction is one of the important wave propagation mechanisms. In this study, measurements were taken to assess the effect of the human body as it moves between the transmitter and the receiver. To predict the signal attenuation, the principles of Fresnel diffraction were utilized, particularly emphasizing complex Fresnel integrals. Our results show that the received power varies significantly based on the person’s position, as diffraction loss highly depends on the body’s location. This study enhances our understanding of how human-induced diffraction, is critical for designing more reliable wireless networks. As the findings demonstrate that the proposed model effectively predicts signal attenuation in indoor environments and emphasizes the importance of accounting for human interference when optimizing communication systems, thus supporting the effective deployment of 5G technology. 

This paper investigates human-induced signal attenuation in indoor mm-wave communications at 32.5 GHz, a critical concern for 5G systems. Two distinct diffraction-based models are applied to the same indoor scenario to assess human blockage effects: one employs the Double Knife-Edge Diffraction (DKED) approach, and the other uses Fresnel diffraction principles with complex Fresnel integrals. Controlled experiments with a human subject moving between a transmitter (TX) and a receiver (RX) reveal that the DKED model consistently underestimates the received power by 2 6 dB, while the Fresnel diffraction approach underestimates it by 2–5 dB Based on the comparative results, the DKED model demonstrates higher accuracy in predicting signal attenuation, offering valuable insights for improving indoor 5G network performance

Abstract— Arabic language is characterized by its rich morphological structure, presenting unique challenges in Nat- ural Language Processing (NLP). The categorization of Arabic plurals is the subject of this study, which uses a trans- former-based model—more precisely, the pre-trained Arabic BERT architecture—and has never been studied previously. Given the complexities of Arabic language, particularly in pluralization which includes sound masculine, sound feminine, and irregular (broken) plurals, the research aims to enhance NLP capabilities in this area. By utilizing a dataset of 7,400 instances classified into four distinct categories, the study demonstrates the effectiveness of transfer learning in achieving high classification accuracy, with results indicating an accuracy of 97% across both validation and testing sets. Addition- ally, the model achieves high precision, recall, and F1-score metrics. A confusion matrix provides insights into classifica- tion performance, highlighting areas of misclassification. The findings underscore the potential of transformer models in overcoming the linguistic challenges posed by Arabic plural forms.

Abstract— This survey explores different machine translation methods utilized in various systems and platforms for commercial and research purposes. These methods play a vital role in enabling global communication, enhancing accessibility, supporting business and trade, fostering intercultural understanding, facilitating travel and tourism, aiding education, delivering fast and efficient translations, contributing to humanitarian aid efforts, promoting research and collaboration, and preserving language and culture. The survey aims to equip software developers and researchers interested in machine translation with valuable insights into these methods. Its objective is to help them improve translation quality with great accuracy by providing them with the necessary knowledge and understanding of these approaches. The papers utilized in this survey were obtained from Open Access Journals and online databases. All these methods are essential and can differ based on the specific context, available resources, and the quality of the translation required. To achieve optimal translation results, researchers and practitioners commonly employ a combination of various methods and techniques. 

Abstract—Arabic Named Entity Recognition (NER) serves as a crucial facet within Natural Language Processing, given the intricacies of the Arabic language. This survey consolidates the current landscape of Arabic NER, covering methodologies, challenges, and advance- ments. The review encompasses an in-depth analysis of diverse approaches, from rule-based systems to modern deep learning techniques, highlighting their effectiveness and limitations. It also addresses the specific challenges inherent to Arabic NER, such as dialectal variations and limited annotated data, while exploring recent advancements and their applications in sentiment analysis, information retrieval, and other domains. This survey aims to provide a comprehensive overview, catering to researchers, practitioners, and enthusiasts in the field of Arabic NER and NLP.