2790 papers
This paper introduces "Training as Treatment," a causal framework that models data-role assignments and pipeline choices as interventions to rigorously analyze machine-learning evaluation metrics, accounting for model-induced interference and enabling more robust diagnostics for fairness, leakage, and generalization.
This study proposes and evaluates a speaker-independent Tigrigna automatic speech recognition system based on a Hidden Markov Model framework for Ethiopian court proceedings, achieving a 73.84% word accuracy using a context-dependent acoustic model trained on a custom corpus of 24 native speakers.
This paper demonstrates that Universal Differential Equations, which integrate known mechanistic SIR structures with neural networks for uncertain transmission dynamics, outperform fully data-driven Neural ODEs in forecasting accuracy, robustness, and data efficiency, particularly under conditions of limited data and high noise.
This systematic synthesis of review evidence from 2015 to 2025 identifies key determinants and barriers to Virtual Reality adoption across diverse domains, highlighting the need for an integrated framework that addresses methodological gaps and combines cognitive, experiential, physiological, and organizational factors to ensure long-term user engagement.
This paper presents a computer-assisted multimodal feedback evaluation model for junior high school art education that integrates visual and textual analysis with evidence tracking and teacher review to significantly improve assessment reliability and reduce error rates.
This paper presents a framework that operationalizes data lineage as infrastructure for trusted financial data pipelines, demonstrating through deployment that it significantly reduces audit response times and enables machine-verifiable tracking of data sources and transformations.
This paper audits educational AI systems using character-level language models and finds that while they can successfully predict held-out curriculum transitions in synthetic and real-world datasets, this capability stems from exploiting explicit metadata factorization rather than semantic understanding of lesson content.
This paper proposes a non-invasive Parkinson's disease severity classification system using triaxial IMU sensor data and ensemble learning, demonstrating that the LightGBM model outperforms other machine learning algorithms with approximately 97% accuracy.
This study evaluates five supervised machine learning models for detecting zero-day attacks on imbalanced data, ultimately selecting XG Boost as the optimal solution due to its superior balance of high accuracy and fast processing time compared to the more accurate but slower Random Forest.
This study introduces a three-stage machine learning pipeline, validated on over 670,000 Kenyan magistrate court cases, which utilizes LightGBM and CatBoost models to predict case processing durations and classify same-day disposals, thereby enabling proactive backlog management through an empirically derived risk scheme and a deployed web application.