2890 papers
This study proposes the Dynamic Adaptive Fusion Model (DAFM), a real-time intelligent forecasting system that dynamically integrates Decision Trees, Random Forests, XGBoost, and BiLSTM networks via adaptive gating mechanisms to overcome the limitations of static models, achieving superior accuracy (R² = 0.97) and rapid inference for oil production prediction.
The paper introduces ModernBERT-TR, a 150M-parameter Turkish encoder pretrained from scratch on 144.4 billion tokens using the ModernBERT architecture and a custom tokenizer, which significantly outperforms existing Turkish baselines and matches larger concurrent models despite being trained on substantially fewer resources.
DriftSense++ is a dynamic framework that enhances spatial crowdsourcing task allocation by employing a closed-loop, region-aware approach to detect and adaptively manage both data and concept drift, significantly outperforming existing methods in accuracy, false alarm reduction, and recovery speed across diverse real-world datasets.
This paper focuses on evaluating various machine learning techniques to effectively detect and predict SQL injection attacks, which allow attackers to manipulate database queries and compromise server security.
This study evaluates machine learning models for predicting ICU mortality and length of stay in Ethiopian hospitals using real and synthetic data, finding that hybrid training with CTGAN-generated data significantly outperforms other methods despite its high computational cost, while highlighting oxygen requirement and SpO2 as the strongest mortality predictors.
This paper proposes an orientation-aware mesh learning method that utilizes a novel signed half-dihedral scalar to encode local face orientation, significantly improving robustness and classification accuracy for 3D shapes under structural perturbations like cutting and open boundaries compared to conventional edge-centric approaches.
This retrospective study developed and evaluated a multimodal AI-based clinical decision support system integrating demographic, clinical, genomic, and imaging data to enhance personalized prostate cancer treatment prediction in Brunei, demonstrating that a hybrid LSTM–Random Forest architecture with late-stage fusion achieved robust performance (AUC 0.964) despite limitations related to the small sample size.
This study demonstrates that machine learning algorithms, particularly Support Vector Classifiers, achieve high accuracy (up to 99%) in detecting forest degradation in Nigeria's Omo Forest Reserve using remote sensing data, thereby advocating for their integration into national conservation policies for sustainable forest management.
This study employs a Design of Experiments methodology to analyze and optimize critical hyperparameters—specifically chunking strategy, context volume, and LLM selection—within a Retrieval-Augmented Generation system tailored for the Public Prosecutor's Office of Espírito Santo, revealing that these factors significantly influence response quality through statistical analysis using the Aligned Rank Transform.
This paper introduces THEA, a prompt-defined revision policy that treats executable world models as falsifiable theories to orchestrate parent agents and specialist subagents, achieving state-of-the-art performance on 22 of 25 ARC-AGI-3 games with significantly reduced costs while ensuring hypothesis revision remains explicit, auditable, and cost-effective.