2786 papers
This paper introduces Contextual Token Rotation Ensembles (CTRE), a novel tabular learning framework that enhances ensemble diversity and robustness by integrating Transformer-based contextual token representations with data-adaptive, interaction-aware feature rotations, outperforming classical methods particularly on high-dimensional and imbalanced datasets.
This paper proposes a hierarchical co-evolutionary multi-objective artificial raindrop algorithm (HCMOARA) that integrates Latin hypercube sampling, partitioned subpopulations with specialized search strategies, and an adaptive flow-factor update mechanism to effectively optimize makespan, energy consumption, and cost in flexible job-shop scheduling problems.
This study presents an integrated machine learning framework that utilizes NLP, dimensionality reduction, and clustering algorithms to analyze a university's research portfolio, revealing that the Faculty of Engineering and Natural Sciences dominates the budget and identifying significant opportunities for strategic realignment alongside a 33.4% high-risk project profile.
This paper proposes four new heuristic techniques, including BEFRA and BEQRA, that leverage underutilized edge frequency and quality information to significantly enhance the performance and robustness of Ant Colony Optimization algorithms for solving the Symmetric Traveling Salesman Problem.
This paper proposes a hybrid LSTM–LightGBM framework enhanced with SHAP for interpretable, real-time multiclass DDoS attack detection in 5G and IoT networks, demonstrating superior accuracy and robustness across the CICIDS2017, CIC-DIAD2024, and CICIoT2023 datasets.
This paper proposes and evaluates an AI-based fraud detection system specifically designed for subscription-based payment services, demonstrating that a hybrid XGBoost+Bi-LSTM architecture outperforms standalone models across multiple metrics on a synthetic dataset covering six distinct fraud types, while explicitly acknowledging the results as proof-of-concept evidence rather than validated real-world performance.
This paper introduces the Immunological Synapse Network (ISN), a novel deep-learning architecture inspired by T-cell recognition mechanisms that effectively detects persistent weak anomalies in time-series data by filtering out transient signals through kinetic proofreading, outperforming standard models on tasks defined by temporal continuity rather than instantaneous magnitude.
This feasibility study evaluates the performance of classical machine learning and deep learning models on Ghana's Controller and Accountant-General's Department financial reports, revealing that while fine-tuned XGBoost achieved perfect category classification, the subsequent ablation study demonstrated that keyword features significantly inflated these results, underscoring the critical need for rigorous feature validation before deploying such tools in public sector contexts.
This paper proposes Gdlcn–dsbo, a hybrid framework combining Graph-dependent Labelled Convolutional Networks (GDLCN) and Deep Scheduled Butterfly Optimization (DSBO) to optimize multi-objective cloud resource and task allocation, thereby enhancing system efficiency, reducing costs, and ensuring fairness in dynamic IoT environments.
This paper proposes and synthetically evaluates a candidate pattern language of seven design patterns for resilient data pipelines in resource-constrained small and medium-sized enterprises, demonstrating through failure-injection experiments that these patterns effectively address specific failure modes—such as duplicates, schema drift, and silent data loss—compared to standard baselines, while explicitly acknowledging the study's limitations as a prototype-based, non-field-validated contribution.