Forecasting financial time series is particularly challenging since market data is complicated and non-stationary, and it is necessary to identify both short-term momentum and long-term structural patterns. This work develops the Dynamic Group Fusion Enhanced Transformer (DGFET), a new approach that integrates adaptive feature group fusion and selective information processing. The suggested DGFET architecture has Group-FiLM adapters that use Dynamic Group Fusion techniques for adaptive feature transformation to manage market, fundamental, technical, and sentiment feature groups. We assess the model using four unique labeling strategies: short-horizon momentum (binary/ternary) and triple-barrier (binary/ternary), which represent various temporal horizons and forecasting methods. Our ablation analysis, conducted on a comprehensive EUR/USD dataset from 2010 to 2023 with 88 features, demonstrates that the proposed method consistently outperforms baseline LSTM and standard transformer models across all prediction objectives. The improved architecture has a higher overall performance, with an F1-macro score of 0.5356 and a ROC-AUC of 0.6612. It also works very well for short-horizon momentum binary classification (F1: 0.7219, ROC-AUC: 0.8105). The results show that adaptive feature fusion works better than traditional designs when combined with dynamic group selection. The best configurations depend on the specific prediction job. Our results underscore the imperative of task-specific architectural design in financial machine learning applications, especially for methodologies necessitating varied temporal horizons and prediction granularities.

This research aims to enhance portfolio selection by integrating machine learning regression algorithms for predicting stock returns with various risk measures. These measures include mean-value-at-risk (VaR) variance (Var), semi-variance mean-absolute-deviation (MAD) and conditional value-at-risk (C-VaR). Addressing gaps in existing literature. Traditional methods lack adaptability to dynamic market conditions. We propose a hybrid approach optimized by genetic algorithms. The study employs multiple machine learning models. These include Random Forest, AdaBoost XGBoost, Support Vector Machine Regression (SVR) K-Nearest Neighbors (KNN) and Artificial Neural Network (ANN). These models are used to forecast stock returns. Utilizing monthly data from the Tehran Stock Exchange, the results indicate that the genetic algorithm prediction model combined with mean-VaR, Var semi-variance and MAD, produces the most efficient portfolios. These portfolios offer superior returns with minimized risk compared to other models. This hybrid strategy provides a robust and efficient method for investors aiming to optimize returns while managing risk effectively. To implement this approach successfully it is crucial to balance investments. This involves both traditional and alternative asset classes, ensuring diversification. It also capitalizes on market opportunities. Regular review and adjustment of fund allocation are essential. Maintain an optimized strategy for maximum returns and minimal risk.

The combination of machine learning (ML) and neural networks (NN), specifically deep learning (DL), is making a big breakthrough to breast cancer studies. This scientometrics survey studies 200 highly cited publications to map the intellectual landscape and studies trends in this dynamic field. The survey discloses a dominant concentration on computer-aided diagnosis (CAD) systems using convolutional neural networks (CNNs) for the classification of breast cancer from different imaging modalities, including mammography, histopathology, ultrasound, and magnetic resonance imaging (MRI). Key survey directions identified include: (1) the development of comprehensive deep learning techniques for image-based detection and classification; (2) the application of transfer learning to resolve data scarcity; (3) the combination of multi-omics and clinical data for personalized prognosis and treatment prediction; and (4) the exploration of explainability and robustness in ML-driven clinical tools. This study synthesizes the methodological advancements, sheds light on the evolution from traditional machine learning to deep learning, and surveys the challenges associated with data heterogeneity, model interpretability, and clinical integration. By giving a structured overview of the seminal work and emerging paradigms, the study serves as a reference for graduate students and other interested parties to have a better understanding about the current state and future trajectories of AI in breast oncology.

The paper discusses the advanced simulated annealing metaheuristic approach to solving the difficult operating room (OR) scheduling issue. A detailed mathematical formulation for the multi-day OR scheduling problem is presented that takes into account patient urgency scores, surgeon compatibility, room capacity limits, and time restrictions. Realizing that exact methods would be computationally untenable for larger-sized instances, we propose a highly sophisticated simulated annealing algorithm that uses a new way of representing solutions, makes strategic neighborhood moves, and applies adaptive penalty functions. The algorithm shows solid performance over many different scales of problems and easily deals with instances that mixed-integer linear programming methods find prohibitive. Computational tests have shown that the method suggested secures high-quality solutions while keeping computational cost low, thus giving hospitals a useful tool for increasing OR scheduling efficiency.

At the edge of the network, the exponential increase of data produced along with the growing concerns over data privacy coming from regulations and society have all together triggered the rise of Federated Learning (FL) as the main approach in distributed machine learning (DML). Fed learning allows the model training to be performed on decentralized devices or data silos even without the raw data being transferred. Hence, FL is completely in line with the objectives of the privacy-preserving techniques. In this paper, we carry out a scientometric analysis on the 200 most cited papers, which are the first 200 papers at the intersection of "Distributed Machine Learning," "Federated Learning," and "Privacy-Preserving" published between 2020 and 2024, and the Scopus database is where they are indexed. The literature of publication trends, prominent authors and works, the thematic clusters, and research fronts that are changing are all systematically examined in this study; hence, the intellectual landscape of this fast developing field is mapped out. Our findings point to the existence of certain streams of research such as the algorithms with differential privacy being the mainstay, secure aggregation methods through the use of homomorphic encryption and multi-party computation, blockchain-based FL systems which ensure security and trust, and resource-efficient FL that supports IoT and edge computing. The results also show an area that is nearly enjoying a complete transformation as a result of the overpowering need to address the triad of model quality, data protection, and system efficiency. The review not only encourages researchers, and practitioners but also helps the policymakers by providing the current trend to which the key challenges can be identified and the future directions in privacy-preserving distributed intelligence anticipated.

The technological efficiency of Magnetic Resonance Imaging (MRI) and Computed Tomography (CT) services within eight Australian states and territories is focused on within this particular study. Data Envelopment Analysis (DEA) was applied with two inputs—the number of radiographers and scanners estimated—and one output—MBS (Medicare Benefits Schedule) services per 1,000 people. The efficiency was calculated with the help of four standard DEA models, i.e., CCR, BCC Input-Oriented, BCC Output-Oriented, and the Additive model. The result of the study was a counter-intuitive finding: most models identified the Northern Territory (NT), a large and sparsely populated region, as the most effective unit for MRI and CT services. On the contrary, states with larger populations such as Victoria, Queensland, and New South Wales were still in the poor relative performance rankings. The Australian Capital Territory (ACT) was also among the top but Tasmania had medium efficiency. The outcomes indicate that small and resource-limited systems can reach a very high technical efficiency by using the limited infrastructure extensively. Nevertheless, this is probably an outcome of the models' concentration on volume-based output and may not reflect performance dimensions such as case-mix complexity or equity of access, among others. The discoveries are a milestone for the choice of action of policymakers, pointing out that one should examine the factors causing inefficiency in the large and more complex state health systems.