Sana'a University Journal of Applied Sciences and Technology

Forensic entomology is a branch of forensic science that utilizes insects to assist in criminal investigations. Building on the foundational principles of forensic entomology, this study investigates of entomofauna’s role in decomposing rabbit carcasses ( Oryctolagus cuniculus L.), across varied environments. Using a consistent collection and analysis protocol, we monitored the decomposition process and insect activity in three different environments in Dhamar City, Yemen. Two carcasses were outdoors: one in a botanical garden and another in an open agricultural area. The third was indoors in a laboratory at the Faculty of Agriculture at Thamar University. The research was conducted over two periods: September and November 2023, as well as October 2024. Urban environments, including both botanical gardens and controlled indoor settings, typically exhibit higher temperatures and lower humidity, which accelerates decomposition more than in agricultural sites. The dry stage was reached by day 30 in urban environments, whereas in agricultural areas, it took approximately 52 days. A total of 1270 adult insect specimens were collected, with outdoor habitats exhibiting the highest abundance. Chrysomya albiceps (Calliphoridae) was the dominant species at all sites in the early phases of decomposition, confirming its forensic usefulness in determining postmortem intervals (PMIs). This study describes insect succession resulting from carrion decomposition in both indoor and outdoor habitats in Dhamar City, Yemen. It establishes baseline data essential for estimating the post-mortem interval (PMI) in the locality.

Many factors influence the efficiency of a classification model in Machine Learning, including the dataset’s size and the type of Machine Learning (ML) technique utilized in the classification process. Primarily, the accuracy varies among different machine learning methods. This paper develops a model that analyses and measures the impact of Tax Dataset Augmentation on the Accuracy of Machine Learning Models. The paper compares the performance of different models based on the most common machine learning techniques, namely: DT, RF, SVM, and ANN (MLP). Based on three datasets provided by Yemen’s Tax Authority. The first dataset contains 1083 records, while the second dataset is identical to the previous one, but with nearly five times the number of records, following data preprocessing, which resulted in approximately 5000 additional records. The last dataset is the same as the original dataset, but it has been duplicated nearly ten times, resulting in almost 10,000 records. The dataset partitioning technique utilized k-fold validation using the three datasets. Results show that the performance of ML classifiers such as ANN (MLP), DT, RF, and SVM is affected by dataset augmentation in terms of accuracy, recall, precision, and F-score. Results also show that the performance varies among the first three techniques; however, the SVM Classifier yields the lowest results. In general, despite some techniques leading to overfitting, it is found that most machine learning models utilizing tax datasets with five times the duplicates produced better outcomes than those using the original dataset. These findings provide practical guidance for tax authorities in selecting robust machine learning models under limited data availability and highlight risks associated with naïve dataset expansion.

This study addresses a critical knowledge gap by presenting the first comprehensive 3D static geological model of the heavy-oil-bearing Upper Qishn Clastic Member in the Al Roidhat Field, Yemen. The model integrates data from eight wells. The workflow encompassed structural modeling, stratigraphic zonation into three reservoir units (S1, S2, S3), facies modeling using Sequential Indicator Simulation (30 realizations), and petrophysical property modeling using Sequential Gaussian Simulation. Probabilistic volumetric assessment was based on 100 stochastic realizations. Uncertainty analysis yields a P50 Stock-Tank Oil Initially in Place (STOIIP) of 56.81 million barrels (P10–P90 range: 44.40–73.39 MMSTB). The S3 unit exhibits the best reservoir quality (62.8% sand, 0.15 porosity, 47mD permeability), while the S1 unit shows good quality (48.1% sand, 0.14 porosity, 60 mD permeability). Yet, the S1 unit contains 58% of the STOIIP because it lies entirely within the oil column, whereas a significant portion of S3 is submerged below the oil-water contact. Together, S1 and S3 account for approximately 85% of the total resource. This pioneering 3D model quantifies reservoir heterogeneity and structural controls, establishing an essential digital foundation for the field. It demonstrates that structural position can override intrinsic rock quality in governing hydrocarbon distribution, and it clearly identifies the S1 and S3 units as priority targets for future development and enhanced oil-recovery planning in this challenging heavy-oil reservoir.

Clustering is a fundamental task in unsupervised learning, aiming to group similar data points while maximizing inter-group dissimilarity. Traditional clustering algorithms often struggle with large-scale datasets containing mixed data types (numerical and categorical). This Research introduces an incremental clustering algorithm that incorporates a novelty criterion to identify only interesting patterns from big data. The novelty criterion functions as a distance metric, measuring deviation between instances and labeling an instance as novel if its deviation exceeds a user-defined threshold. Unlike conventional methods such as k-means, the proposed approach does not require predefining the number of clusters; instead, it dynamically determines them during execution. This flexibility enhances its applicability to complex datasets. The algorithm was implemented and evaluated using public datasets, Through experimentation it became clear that the proposed algorithm achieved the same high accuracy of 81.5% as the DBSCAN algorithm. and However it surpassed it in speed, taking only 22 seconds compared to 100 seconds. This means it is more efficient the same level of accuracy but in significantly less time, making it a better option for practical applications.

In this paper, we introduce novel concepts that represent unique contributions to the framework of supra soft topological spaces. First, we define and characterize the notion of feeble α-supra open soft subsets, establishing their fundamental properties. Through illustrative examples, we demonstrate the relationships between this class of soft subsets and existing generalizations of supra open soft sets. Then, we put forward the concepts of interior, closure, frontier, and accumulation operators induced by the class of feeble α-supra open soft and α-closed subsets, deriving their key properties and establishing fundamental relationships through some formulas

The rapid evolution of 5G networks introduces unprecedented high-speed communication and massive connectivity, but it also intensifies security challenges. Ensuring confidentiality, integrity, and availability of data requires robust encryption mechanisms. This study evaluates the performance and overhead of Advanced Encryption Standard (AES), Rivest–Shamir–Adleman (RSA), and Elliptic Curve Cryptography (ECC) in 5G networks using simulation-based analysis. Metrics considered include latency (ms), throughput (%), encryption/decryption time per MB, and computational cost on UE-class hardware, providing quantitative comparison across typical 5G scenarios (high mobility eMBB, URLLC, and mMTCIoT devices).

Key findings: AES achieves low latency and high throughput for bulk data; ECC provides secure, lightweight key exchange; RSA is suitable only for session key establishment due to higher computational overhead. Overall, hybrid AES+ECC provides the best tradeoff between security and performance in realistic 5G environments.

 This study aimed to identify the most significant factors distinguishing between diabetes and non-diabetes, using linear discriminant analysis LDA. A sample of 368 individuals (225 diabetes mellitus, 143 non- diabetes mellitus) was analyzed based on data collected through via personal interviews and Al-Thawrah General Hospital in the capital Sana’a records, for the year 2025. The stepwise discriminant analysis led to the building of a statistically significant model. Results revealed that the variables (x₅) body mass index BMI, (x₁₀) Cholesterol, ( x₆) Gout, (x₉) Blood pressure, and ( x₂) Age, had the highest discriminative power, as indicated by standardized coefficients and Wilks’ Lambda values (p < 0.05). These variables significantly contributed to classifying cases, while the remaining variables did not demonstrate a significant effect. The overall correct classification rate for diabetes mellitus DM reached 87.3%, when the overall classification accuracy of the model reached 89.9%, confirming its strong predictive power.

تم تنفيذ البحث خلال عامي 2023-2024م في مختبر المؤسسة التشاركية للدراسات والبحوث والنشر وكلية الزراعة والأغذية والبيئة جامعة صنعاء، في بيئة متحكم فيها، وبهدف دراسة مدى إمكانية إيجاد تباينات وراثية في صنف من البن اليمني باستخدام المواد المطفرة. استخدم في هذه الدراسة بذور صنف محلي "عديني"، في تصميم تجريبي عشوائي كامل في أربع معاملات هي الشاهد، نقع البذور بمحلول ازيد الصوديوم بتركيز 0.1 مولار، معاملة البذور بمستخلص بذور نبات الحرمل بتركيز 150جرام/لتر، وغمس جذور بادرات البن بعد الانبات بمستخلص الحرمل، وجميع نباتات هذه المعاملة الرابعة ماتت خلال الأسابيع الأولى من غمس جذور البادرات في مستخلص الحرمل مما يوشر الى أهمية استخدام تراكيز وفترات تغميس اقل. النتائج بينت ان استخدام المواد الكيميائية المطفرة كان لها دور في احداث تغيرات معنوية في سرعة ونسبة انبات البذور، وفي طول البادرات وعدد وطول الجذور،والتحليل لمنحنيات التكرارات أوضحت تأثر توزيع التكرارات للمجموعة النباتية الطبيعية بشكل واضح في المعاملات بالمواد المطفرة مقارنة بالمجموعة الطبيعية للشاهد بالنسبة لمجموع الصفات. وفي المرحلة الثالثة سجلت التحويرات المظهرية التي ظهرت على الأوراق الفلقية وعلى الأوراق الحقيقية النامية على العقدة الأولى، وقد تم تمييز سبعة اشكال من التحويرات تمثلت في التغاير في الأوراق الفلقية وفي الأوراق الحقيقية. النتائج المشجعة هذه تشير بوضوح الى امكانية استخدام الطفرات كإدة لتوسيع التباين الوراثي في البن اليمني.

The Schrödinger operator with a combined singular, confining, and logarithmic potential arises in diverse physical contexts, from quantum dots and Bose-Einstein condensates to effective field theories. Despite its broad relevance, a rigorous spectral analysis of this operator family has remained elusive due to the simultaneous presence of three challenging features: a 1/x² singularity at the origin, an unbounded logarithmic term, and harmonic confinement at infinity.

We introduce and rigorously analyze the one-parameter family of operators:

H = -d²/dx² + ν(ν + 1)/x² + ω²x² + β ln x,

where ν > 0, β ∈ R, and ω > 0, acting on L²(0,∞).

Self-adjointness is established through a combination of Weyl's limit-point/limit-circle classification and the KLMN theorem for quadratic forms, circumventing the failure of standard Kato-Rellich perturbation theory.

The spectrum is shown to be purely discrete and bounded below, with simple eigenvalues that depend analytically on the logarithmic coupling β.

The central result of this work is the sharp asymptotic expansion of the eigenvalues for large quantum numbers:

En = 2ωn + β ln n + γ0 + β/(2n) + O(1/n²),

where

γ0 = ω(ν + 3/2) + β(ln√(2/ω) - 1/2).

This formula is derived explicitly from the Langer-corrected WKB quantization condition.

The expansion reveals a distinctive logarithmic growth that modifies the standard linear Weyl law, yielding the spectral counting function:

N(E) = E/(2ω) - (β/(2ω)) ln E + O(1).

To complement the asymptotic analysis, we construct the exact Green's function and prove the Hilbert-Schmidt property of the resolvent.

Numerical diagonalization using a logarithmic grid confirms the asymptotic formula with relative errors below 0.02% for all n ≥ 5, validating both the analytic predictions and the underlying WKB methodology.

These results provide the first complete spectral characterization of Schrödinger operators with logarithmic potentials in the presence of singular and confining terms, with potential applications to few-body quantum systems, spectral geometry, and the analysis of trace formulae.

Telecommunications operators are under increasing pressure to use Artificial Intelligence (AI) to keep up with the recent technologies and stay ahead in the service provision advancement. However, within organizations that are not adequately prepared for AI adoption, many of these initiatives fail to gain traction. The study explores how existing academic research and practical frameworks can conceptualize organizational AI readiness in project-based telecom contexts. It synthesizes six peer-reviewed academic studies on AI readiness and digital transformation and four telecom-relevant AI readiness and maturity frameworks from industry bodies. Using a hybrid Technology–Organization–Environment (TOE) and People Process–Technology-in-use–Data (PPTD) lens, the 80 readiness indicators derived from the academic and practical resources were coded. Then, analyzed them through reflexive thematic analysis and identified nine interrelated AI readiness themes: strategic AI vision and business alignment; leadership, culture and change climate; human capital and AI skills; data assets, governance and protection; technology and infrastructure backbone; AI solutions, lifecycle and experimentation; organizational processes, governance and project operations; value realization and performance impact; and external ecosystem, regulation and partnerships. These themes are integrated into a conceptual diagnostic framework tailored to project management in telecom organizations and motivated by the challenges of communication service providers in contexts such as Yemen. The resulted framework addresses the gap of fragmented academic and practical perspectives. This offers telecom project management offices (PMOs) a conceptualizing mechanism to reflect on AI readiness, identify potential gaps and prioritizing interventions. In the same time, it provides a basis for future empirical validation and cross-sector comparison.

This paper presents a literature review on the application of deep reinforcement learning (DRL) in cybersecurity. It focuses on key domains, including intrusion detection, adaptive cyber defense, multi-agent coordination, and automated penetration testing. A total of 18 peer-reviewed studies published between 2022 and 2025 were selected through a structured review process and analyzed using performance metrics such as accuracy, precision, recall, and F1-score.The results indicate that multi-agent DRL approaches generally outperform single-agent models in dynamic attack environments. Hybrid DRL models that integrate deep learning techniques, such as convolutional and recurrent neural networks and attention mechanisms, show improved detection accuracy and adaptability. DRL-based penetration testing methods also demonstrate the ability to autonomously explore vulnerabilities and optimize attack strategies. However, challenges remain, including limited generalization to real-world scenarios, high computational costs, low interpretability, and the lack of standardized datasets. Addressing these issues can enable the development of more adaptive, efficient, and reliable cybersecurity systems.