George Sammour

Imbalanced classification poses a persistent challenge in real-world decision-support systems, where the rare positive class is typically the focus of interest and predicted probabilities directly influence critical decisions. Many conventional oversampling methods rely on uniform interpolation between minority instances, often generating redundant or noisy samples that fail to capture the heterogeneous structure and local uncertainties inherent in real minority classes. This study introduces LSEA (Latent-Structure Entropy Allocation), a reliability-oriented approach that models the minority class using a Gaussian Mixture Model and employs hierarchical entropy, at both component and instance levels, to direct synthetic sample generation toward structurally ambiguous regions.Evaluated on ten KEEL benchmark datasets and two real-world applications (credit card fraud detection and breast cancer diagnosis), LSEA delivers precision-recall performance and probability calibration comparable to established baselines, including SMOTE-family methods, GMF-SMOTE, and DeepSMOTE. Statistical tests show no significant pairwise differences among the seven methods in most cases. Unlike deep generative alternatives, LSEA achieves these results without neural network training, GPU resources, or extensive hyperparameter tuning, while offering greater interpretability through its explicit entropy-guided allocation.These findings suggest that combining latent structure modeling with entropy-based prioritization provides a practical and reliable way to improve learning under class imbalance, particularly when calibration and computational efficiency matter.

Artificial Intelligence (AI) is increasingly reshaping recruitment processes by automating candidate screening, interviews, and hiring decisions. Despite its growing adoption, AI-enabled recruitment tools raise significant ethical and psychological concerns among job candidates, particularly regarding fairness, transparency, risk, and trust. Addressing a notable research gap concerning candidate perspectives, this study aims to investigate job candidates’ acceptance of AI-enabled recruitment tools, examining how perceived usefulness, ease of use, fairness, transparency, and risk influence their attitudes, mediated by trust and moderated by AI experience. Drawing on the Technology Acceptance Model (TAM), the Unified Theory of Acceptance and Use of Technology (UTAUT), and the Fairness, Accountability, Transparency, and Explainability (FATE) framework, the study introduces a socio-technical conceptual model to address the identified research gap. A quantitative survey was conducted with 143 job applicants in Jordan’s pharmaceutical sector. The data were analysed using SPSS multiple regression analysis to assess the relationships among the variables. The findings show that perceived fairness, transparency, usefulness, and trust significantly shape positive attitudes toward AI-enabled hiring. Whereas ease of use and risk perceptions are less influential, trust proved to be a vital mediator, particularly between job applicants' perceptions of fairness and transparency and their attitudes toward AI-enabled recruitment tools, while prior AI experience moderates the influence of transparency and risk perceptions. The study concludes that respondents placed greater importance on AI's ethical and relational characteristics than technical functionality, highlighting the importance of designing functionally efficient, but also ethically responsible, fair, trustworthy, and transparent AI-enabled recruitment tools to promote candidates’ acceptance and enhance responsible AI adoption in emerging labour markets.

Logistics companies possess collect large amount of data on the shipments they perform while at the same time facing a challenge to understand their complicated market better. They can extract useful market knowledge by using data mining technologies such as visualization and clustering. The detailed results of such big data analytics methods can also be monetized under certain circumstances. We studied the data on the transactions of a logistics company in the Middle East. K-Means clustering of their data proved to generate deeper insight into several clusters of customers having different profiles. The results propose a best fit model for the clustering. Since the clustering and visualization results are relevant, reliable and anonymous they fit the monetization criteria as well. Improved data driven marketing applications are possible for the customers.

Fraudsters can manipulate telecom regulatory systems to their advantage, and to the disadvantage of the telecom operator, in ways that are difficult to detect, trace, and prosecute. A subscriber whose network has been compromised will often refuse to pay large fraudulent charges, leaving the operator to cover the bill resulting in revenue losses. Moreover, attacks frequently happen over holidays and weekends, when networks are often less monitored closely. This calls for an intelligent automated solution for telecom fraud detection. Wangiri (Japanese term) telecom fraud also referred to as “one ring and cut fraud” relies on this single ring method for a quick way to make money. Missed calls from unknown callers entice subscribers to call back unknowingly premium numbers where they are deceived to stay on the line for as long as possible in an effort to inflate their bill. This paper proposes the use of varied ensemble classifiers to overcome the highly biased dataset and help make a more precise classification in an efficient and effective manner. Extreme Gradient Boosting algorithm was found to have the best results in terms of correctness and performance.

In recent decades, activity-based transportation models have gained growing attention, due to their strong foundation in behavioral theory and ability to model the response of individuals to travel demand management policies. Hence, researchers have become increasingly interested in analyzing and predicting individuals’ decisions about activity participation. This paper investigates the reliability and uncertainty of computational process activity-based models. The design of the scheduling process model is experimented with by introducing an alternative decision sequence. The results provide additional information to better understand the process model’s reliability and behavior. Furthermore, the findings show that the current sequence of decision steps in the process model in ALBATROSS achieves satisfactory work activity schedules. Finally, the study concludes that using a decision tree model achieves a better performance than using diverse data mining approaches.

As a result of the rapid increase of online shopping, and the movement of products in and out among various countries, governments made it mandatory to pay a certain fee referred to as customs to maintain each country's economy, based on some rules i.e. weight and price. This paper proposes a model to anticipate and inform the customer if his order is going to be subject to customs or not based on some pre-defined rules. The model is built using a software called SPSS and data mining techniques such as; Decision Tree (DT) and logistic regression. A comparison between the two data mining techniques is made showing that the DT results were better than the logistic regression in terms of accuracy, precision, recall and f-score.