Abstract

Contextual recommendation has become attainable through the massive amounts of contextual information generated by smartphones and Internet of Things (IoT) devices. The availability of a huge amount of contextual data paves the way for a revolution in recommendation systems. It overcomes the static nature of personalization, which does not allow the discovery of new items and interests, toward a contextualization of the user’s tastes, which are in constant evolution. In this paper, we proposed CoD-MaF, a Context-Driven Collaborative Filtering using Contextual biased Matrix Factorization. Our approach employs feature selection methods to extract the most influential contextual features that will be used to cluster the users using K-means algorithm. The model then performs a collaborative filtering based on matrix factorization with improved contextual biases to suggest relevant personalized recommendations. We highlighted the performance of our method through experiments on four datasets (LDOS-CoMoDa, STS-Travel, IncarMusic and Frappe). Our model enhanced the accuracy of predictions and achieved competitive performance compared to baseline methods in metrics of RMSE and MAE.

Abstract

Classification of imbalanced multi-class data is still so far one of the most challenging issues in machine learning and data mining. This task becomes more serious when classes containing fewer instances are located in overlapping regions. Several approaches have been proposed through the literature to deal with these two issues such as the use of decomposition, the design of ensembles, the employment of misclassification costs, and the development of ad-hoc strategies. Despite these efforts, the number of existing works dealing with the imbalance in multi-class data is much reduced compared to the case of binary classification. Moreover, existing approaches still suffer from many limits. These limitations include difficulties in handling imbalances across multiple classes, challenges in adapting sampling techniques, limitations of certain classifiers, the need for specialized evaluation metrics, the complexity of data representation, and increased computational costs. Motivated by these observations, we propose a multi-objective evolutionary induction approach that evolves a population of NLM-DTs (Non-Linear Multivariate Decision Trees) using the -NSGA-III (-Non-dominated Sorting Genetic Algorithm-III) as a search engine. The resulting algorithm is termed EMO-NLM-DT (Evolutionary Multi-objective Optimization of NLM-DTs) and is designed to optimize the construction of NLM-DTs for imbalanced multi-class data classification by simultaneously maximizing both the Macro-Average-Precision and the Macro-Average-Recall as two possibly conflicting objectives. The choice of these two measures as objective functions is motivated by a recent study on the appropriateness of performance metrics for imbalanced data classification, which suggests that the mAURPC (mean Area Under Recall Precision Curve) satisfies all necessary conditions for imbalanced multi-class classification. Moreover, the NLM-DT adoption as a baseline classifier to be optimized allows the generation non-linear hyperplanes that are well-adapted to the classes ‘boundaries’ geometrical shapes. The statistical analysis of the comparative experimental results on more than twenty imbalanced multi-class data sets reveals the outperformance of EMO-NLM-DT in building NLM-DTs that are highly effective in classifying imbalanced multi-class data compared to seven relevant and recent state-of-the-art methods.

Abstract

Abstract

Multi-output Regression (MOR) is a machine learning technique that aims to predict several values simultaneously. Some existing approaches addressed this problem by decomposing the MOR problem into separate single-target ones. However, in real-world applications, it is more advantageous to exploit the inter-target correlations in the prediction task. Some other approaches proposed simultaneous prediction but they are based on greedy algorithms and are prone to fall easily into local optima. In order to solve these issues, we propose a novel approach called Bi-level Evolutionary Model TreeChain Induction (BEMTCI) which is able to deal with multi-output datasets using a bi-level evolutionary algorithm. BEMTCI evolves a population of Model Tree Chains (MTCs) where each Model Tree (MT) focuses on the prediction of one single target. The upper-level explores different orderings of the MTs of each MTC to find the best chaining order which is able to express the relationships among the output variables. A further optimization is performed in the lower-level of BEMTCI which concerns the linear models at the leaves of the MTs. The experimental study showed the effectiveness of our approach compared to the existing ones when applied on sixteen MOR datasets. The genetic operators employed in our BEMTCI ensure the variation of the population and guarantee a fair and a precise prediction due to the evaluation process. The obtained results prove the performance of our BEMTCI in solving MOR problems.

Abstract

Cloud resource brokerage is a fundamental challenge in cloud computing, requiring the efficient selection and allocation of services from multiple providers to optimize performance, sustainability, and cost-effectiveness. Traditional approaches often struggle with balancing conflicting objectives, such as minimizing the response time, reducing energy consumption, and maximizing broker profits. This paper presents NSGA-III-GKM++, an advanced multi-objective optimization model that integrates the NSGA-III evolutionary algorithm with an enhanced K-means++ clustering technique to improve the convergence speed, solution diversity, and computational efficiency. The proposed framework is extensively evaluated using Deb–Thiele–Laumanns–Zitzler (DTLZ) and Unconstrained Function (UF) benchmark problems and real-world cloud brokerage scenarios. Comparative analysis against NSGA-II, MOPSO, and NSGA-III-GKM demonstrates the superiority of NSGA-III-GKM++ in achieving high-quality tradeoffs between performance and cost. The results indicate a 20% reduction in the response time, 15% lower energy consumption, and a 25% increase in the broker’s profit, validating its effectiveness in real-world deployments. Statistical significance tests further confirm the robustness of the proposed model, particularly in terms of hypervolume and Inverted Generational Distance (IGD) metrics. By leveraging intelligent clustering and evolutionary computation, NSGA-III-GKM++ serves as a powerful decision support tool for cloud brokerage, facilitating optimal service selection while ensuring sustainability and economic feasibility.

Abstract

Financial markets exhibit high volatility due to various external factors, making stock price prediction a complex yet crucial task for investors and financial institutions. Accurate forecasting not only enhances decision making but also mitigates financial risks. This paper introduces SMAPF-HNNA, an advanced framework that integrates multiple neural network (NN) architectures for robust time-series analysis and stock price forecasting. The proposed approach leverages Convolutional Neural Networks (CNNs) for automatic feature extraction, followed by the application of diverse NN models, including Simple Recurrent Neural Networks (SRNNs), Long Short-Term Memory (LSTM), Bidirectional LSTM (BiLSTM), Gated Recurrent Units (GRU), Bidirectional GRU (BiGRU), and Multilayer Perceptron (MLP) for precise stock price prediction. The framework is rigorously evaluated on multiple benchmark datasets, including NYSE, S&P 500, NASDAQ, and SSE, through extensive training and testing phases. Experimental results demonstrate that the hybrid CNN-MLP model outperforms other architectures across all datasets, achieving exceptionally low error rates with five key regression metrics. The model yields mean squared error (MSE) values between 0.000031 and 0.000004, root mean squared error (RMSE) between 0.0020 and 0.0056, mean absolute error (MAE) between 0.0018 and 0.0042, mean absolute percentage error (MAPE) between 0.12% and 0.32%, and R-squared (R) values ranging from 0.9995 to 0.9999, while maintaining low computational complexity across datasets. These results highlight the potential of SMAPF-HNNA as a highly accurate and computationally efficient solution for stock market prediction, addressing the limitations of previous methods. The proposed framework offers valuable insights for researchers and practitioners, paving the way for more reliable financial market forecasting models.

Abstract

Effective portfolio management is crucial in today’s fast-moving and unpredictable financial landscape. This paper introduces a powerful and adaptive investment framework that fuses classical portfolio theory with cutting-edge artificial intelligence (AI) to optimize portfolio performance during volatile market conditions. Our methodology seamlessly integrates K-means clustering to identify asset groupings based on correlation structures of technical indicators, mean-variance optimization (MVO) to achieve an ideal risk-return trade-off, and advanced Machine Learning (ML) and reinforcement learning (RL) techniques to dynamically adjust asset allocations and simulate market behavior. The proposed framework is rigorously evaluated on historical stock data from 60 prominent stocks listed on NASDAQ, NYSE, and S&P 500 indices between 2021 and 2024, a period marked by significant economic shocks, global uncertainty, and structural market shifts. Our experimental results show that our framework consistently outperforms traditional strategies and recent state of the art models, achieving superior metrics including Sharpe ratio, Sortino ratio, annual return, maximum drawdown, and Calmar ratio. We also assess the computational efficiency of the approach, ensuring its feasibility for real-world deployment. This work demonstrates the transformative potential of AI-driven portfolio optimization in empowering investors to make smarter, faster, and more resilient financial decisions amid uncertainty.

Abstract

Emergency departments (EDs) face increasing
patient demand, leading to overcrowding and resource strain.
Accurate forecasting of ED visits is critical for optimizing
hospital operations and ensuring efficient resource allocation.
This paper proposes a hybrid model combining Variational
Autoencoder (VAE) and Gated Recurrent Unit (GRU) to enhance
patient flow predictions. The VAE extracts meaningful
latent features while handling missing data, whereas the GRU
captures complex temporal dependencies, improving forecasting
accuracy. Compared to traditional models such as LSTM,
GRU, and 1D CNN, our hybrid VAE-GRU model demonstrates
superior predictive performance. Experimental results, based
on real-world hospital data, highlight the model’s effectiveness
in reducing prediction errors and improving decision-making
in dynamic ED environments. Additionally, we compare the
proposed model with ARIMA-ML, emphasizing the tradeoffs
between computational efficiency and prediction accuracy.
The findings suggest that hybrid deep learning approaches
can significantly enhance healthcare resource management,
reducing patient waiting times and improving overall hospital
efficiency.

Abstract

Emergency departments (EDs) must quickly assess
and prioritize patients, especially during crises when demand

exceeds capacity. Traditional triage methods, such as the Jump-
START protocol for pediatric cases and the START (Simple

Triage and Rapid Treatment) method for adults, are commonly
used but may lack precision under high-pressure situations.
This paper proposes a hybrid approach combining ensemble
models—XGBoost, AdaBoost, and CatBoost—with synthetic data
augmentation using Generative Adversarial Networks (GANs) to
enhance triage accuracy for critically ill patients.
Models were trained on real-world ED data, including vital
signs, symptoms, medical history, and demographics. GANs
generated synthetic critical cases to address class imbalance,
improving model sensitivity to high-risk profiles.

Results show that GAN-augmented models outperform base-
line models, with CatBoost offering the best balance between

accuracy and computational efficiency. This approach improves
patient prioritization, reduces delays, and supports better clinical
decision-making in resource-limited environments.
Index Terms—Emergency Department (ED), Patient Triage,

Machine Learning (ML), AdaBoost, XGBoost, CatBoost, Genera-
tive Adversarial Networks (GANs), Urgency Classification, Crisis

Situations.

Abstract

this paper presents a novel ensemble
methodology that integrates deep learning models to enhance
the accuracy and robustness of financial volatility forecasts. By
combining Convolutional Neural Networks (CNNs) and GRU
networks, the proposed approach captures both spatial and
temporal patterns in financial time series data. Empirical results
demonstrate the superiority of this ensemble model over
traditional forecasting methods in various financial markets.
Keywords: Volatility Forecasting, Deep Learning, Ensemble
Modeling, CNN, GRU, Financial Time Series