Towards Smarter E-Learning: Real-Time Analytics and Machine Learning for Personalized Education

International Journal of Computational and Experimental Science and Engineering, Journal Year: 2025, Volume and Issue: 11(1)

Published: Jan. 2, 2025

E-Learning platforms change fast, and real-time behavioural analytics with machine learning provides the most powerful means to enhance learner outcomes. The datasets undergo preprocessing techniques like Z-score outlier detection, Min-Max scaling for feature normalization, Ridge-RFE (Ridge regression Recursive Feature Elimination) selection in order improve accuracy reliability of predictions. Applying Gradient Boosting Machine, classification up a 94% level respect model about predictions on outcomes was achievable. Thus, applying this, feedback systems may offer timely recommendations or directions class that propel students toward better understanding how raise participation success percentages. However, this approach has some potential benefits but there are still various challenges such as managing data imbalance models generalize dynamic environment. Though hybrid methods mitigate problem, pipelines behaviour incorporation call significant computer-intensive resources infrastructure. This integration very high paybacks. It makes possible more responsive individual needs almost met manners, thus giving instantaneous feedback, content suggestions, interventions. Finally, convergence ML culminates adaptive environments which student engagement, retention, quality academic results.

Language: Английский

---

Ripple-Induced Whale Optimization Algorithm for Independent Tasks Scheduling on Fog Computing

IEEE Access, Journal Year: 2024, Volume and Issue: 12, P. 65736 - 65753

Published: Jan. 1, 2024

Due to the revolution of Internet Things (IoT), amount data generation has been redoubling, leading higher latency and network traffic. This resulted in delays services increased energy consumption cloud servers. Fog computing tackles issues associated with long geographical distance between end-users servers by extending service provision closer edge, reducing makespan, optimizing during workload execution. Instead offloading all tasks cloud, delay-sensitive are executed at fog nodes, while others offloaded cloud. However, resources layer limited, posing a challenge for task scheduling computing, particularly as multi-objective optimization problem. Meta-heuristic algorithms have potent find an optimal solution such problems within reasonable time. The Whale Optimization Algorithm (WOA) is relatively new meta-heuristic algorithm that received significant attention from researchers due its impressive characteristics. being exploitation-oriented technique, it falls into local optima lack generating solutions over Limited exploration capabilities also compromise diversity space prolong convergence Therefore, this study, enhanced Ripple-induced (RWOA) proposed, utilizing ripple effects schedule independent computing. It aims minimize makespan maximizing throughput fog-cloud infrastructure improving poor through substantial changes. Extensive simulations performed assess effectiveness proposed algorithm. RWOA outperformed TCaS, HFSGA, MGWO, WOAmM on two datasets: Random NASA Ames iPSC. statistical significance results validated Friedman test Wilcoxon Signed-rank test.

Language: Английский

---

Enhancing Fingerprint Localization Accuracy With Inverse Weight-Normalized Context Similarity Coefficient-Based Fingerprint Similarity Metric

IEEE Access, Journal Year: 2024, Volume and Issue: 12, P. 73642 - 73651

Published: Jan. 1, 2024

Distance-based metrics are the most common fingerprint similarity used in database clustering and localization processes a fingerprint-based system. In this paper, however, less but promising pattern-based metric is proposed as an alternative to distance-base metric. The based on inverse weight (IW) normalization of context coefficient (CSC)-based measure. performance system with IW-CSC-based determined compared square Euclidean, Manhattan, cosine distance-based metrics. k-means algorithm k-means++ cluster initialization process considered for clustering, while k-nearest neighbor (k-NN) localization. Based four databases considered, has slowest time moderate performance. However, it best performance, which at least 52% higher than performances three considered. recommended only when improved primary objective It also use small medium-sized

Language: Английский

---

Dealing with the Outlier Problem in Multivariate Linear Regression Analysis Using the Hampel Filter

Kurdistan Journal of Applied Research, Journal Year: 2025, Volume and Issue: 10(1)

Published: Feb. 9, 2025

Outliers in multivariate linear regression models can significantly distort parameter estimates, leading to biased results and reduced predictive accuracy. These outliers may occur the dependent variable or both independent variables, resulting large residual values that compromise model reliability. Addressing is essential for improving accuracy robustness of models. In this study, proposes a Hampel filter-modified algorithm dynamically detect mitigate extreme values, enhancing estimation performance. The optimizes window size threshold parameters minimize mean square errors, making it robust approach handling analysis. To assess its effectiveness, simulations real datasets were analyzed using MATLAB-based implementation. was compared with classical evaluate improvements outlier detection suppression. indicate proposed method effectively identifies removes improved accuracy, enhanced stability, greater performance Mean Squared Error (MSE). adaptive nature filter minimizes impact outliers, ensuring more reliable model. provides an effective solution By identifying mitigating enhances strengthens capabilities, ensures resilience against data variability. This offers valuable tool researchers practitioners working outlier-prone datasets, reliability

Language: Английский

---

Rejuvenation mechanisms in bituminous RAP mastics: insights from FTIR spectroscopy and multivariate discriminant analysis

Research Square (Research Square), Journal Year: 2025, Volume and Issue: unknown

Published: April 1, 2025

This study examines the rejuvenation mechanisms in bituminous RAP mastics using Fourier transform infrared (FTIR) spectroscopy and multivariate discriminant analysis. The chemical interactions between binders, virgin bitumen, warm mix additives, two commercial recycling agents (RJs) were evaluated to determine their effectiveness restoring aged binders. A hybrid approach combining partial least squares regression linear analysis (PLSR-LDA) was applied extract latent variables, classify samples, identify critical wavenumbers associated with rejuvenation. findings indicate that Sylvaroad mitigates oxidation effects, particularly around ~ 1758–1715 cm^{− 1}, while Storflux primarily influences methyl bending near 1370–1360 1. Despite these irreversible oxidative aging remains evident 1300–1160 1 range, linked oxygen-containing compounds, suggesting complete restoration of binder properties is unattainable. Additionally, mineral composition markers, 489–446 590–555 ranges, persist as key indicators for distinguishing varying contents after

Language: Английский

---

A Hybrid Framework for Robust Anomaly Detection: Integrating Unsupervised and Supervised Learning with Advanced Feature Engineering

International Journal of Computational and Experimental Science and Engineering, Journal Year: 2025, Volume and Issue: 11(2)

Published: April 2, 2025

Finding anomalous data is essential in various applications, from cyber security to healthcare industrial monitoring. Traditional methods- unsupervised or supervised—are far straightforward; methods are notoriously plagued by high false favorable rates and unclear distinction boundaries, while supervised tend rely on a great deal of labeled data, often limited supply highly imbalanced. Indeed, these problems call for unified approach that takes advantage the benefits both paradigms more robust anomaly detection. In this work, we develop hybrid outlier detection framework combining several scoring models (Isolation Forest, Local Outlier Factor, One-Class SVM) XGBoost Logistic Regression as classifier. Instead, combine proposed algorithm with advanced feature engineering techniques (e.g., topological space optimization) extract informative features our representation. Our empirical studies diverse benchmark datasets (Arrhythmia, Cardio, Letter, Mammography, MNIST, Satellite, Speech) indicate model consistently shows significant improvement over any single method. reduces positives negatives precise; recall, F1-score, ROC-AUC highest scores quantitative comparison. We demonstrate usefulness enabling it handle high-dimensional, imbalanced leading meaningful results real-world settings. Establishes new state-of-the-art performance also supplying an scalable versatile complex environments forming basis which build toward future integrated systems.

Language: Английский

---

Characterisation of bitumen grade and modification using a multilevel multivariate discriminant analysis approach and FTIR spectroscopy

Road Materials and Pavement Design, Journal Year: 2025, Volume and Issue: unknown, P. 1 - 21

Published: April 16, 2025

Language: Английский

---

Power supply quality prediction method based on LSTM and self-attention mechanism

Journal of Computational Methods in Sciences and Engineering, Journal Year: 2025, Volume and Issue: unknown

Published: April 26, 2025

Existing LSTM-based power quality (PQ) prediction models primarily rely on historical information, which limits their ability to fully capture contextual dependencies. Furthermore, these process inputs sequentially without accounting for the varying importance of different time steps, leading significant inaccuracies. To address limitations, this study proposes an enhanced PQ model that integrates Bidirectional Long Short-Term Memory (BiLSTM) with a Self-Attention (SA) mechanism. The BiLSTM module is introduced both forward and backward temporal dependencies, enabling more comprehensive long-term patterns in series data. SA mechanism dynamically adjusts steps through weighted summation, enhancing model’s focus critical features improving its capacity nonlinear relationships. from layer are then mapped connected generate final outputs. Experiments were conducted using data Nanchang as primary dataset, additional datasets Nanjing, Wuhan, Changsha, Beijing used generalization testing. results demonstrate BiLSTM-SA outperforms traditional LSTM across all metrics, achieving mean absolute error (MAE) 0.09 voltage deviation, 0.05 improvement over single-layer LSTM. Notably, maintains robust performance complex supply scenarios, generalized MAE only 0.2 Beijing. These findings highlight effectiveness combining reducing errors ensuring stability quality, offering advancement methodologies.

Language: Английский

---

SPT-AD: Self-Supervised Pyramidal Transformer Network-Based Anomaly Detection of Time Series Vibration Data

Applied Sciences, Journal Year: 2025, Volume and Issue: 15(9), P. 5185 - 5185

Published: May 7, 2025

Bearing fault diagnosis is a key factor in maintaining the stability and performance of mechanical systems, necessitating reliable methods for anomaly detection prediction. Unlike traditional conservative maintenance approaches, importance predictive where real-time condition monitoring enables proactive preventive measures has been growing steadily. In this study, we propose deep learning method that effectively discriminates between normal abnormal bearing conditions, while predicting potential faults advance. To achieve this, develop time series model based on supervised transformer architecture. Our proposed tackles data imbalance issue by generating four types synthetic anomalies from vibration incorporates pyramid-structured attention module to reduce computational costs enhance handling long-term dependencies. Experimental results real datasets demonstrate improved F1-scores over 6%p compared existing models significant reduction specific experimental environments. By reliably identifying at an early stage, research contributes reducing improving system stability. Furthermore, it expected have wide applicability state various rotating machinery systems.

Language: Английский

---

Artificial Intelligence Models for Bankruptcy Prediction in Agriculture: Comparing the Performance of Artificial Neural Networks and Decision Trees

Agriculture, Journal Year: 2025, Volume and Issue: 15(10), P. 1077 - 1077

Published: May 16, 2025

Debt levels are a crucial factor when assessing the financial stability of agricultural firms, and excessive indebtedness is usually most important indicator distress. As agriculture capital-intensive sector with high reliance on borrowed funds, firms in this more vulnerable to insolvency. This study examines performance artificial neural networks (ANNs) decision trees (DTs) predicting bankruptcy Slovak enterprises. In an attempt compare models’ performances, consequential ratios investigated through machine learning approaches. ANN DT models found perform significantly better than traditional forecast methods. achieved AUC 0.9500, accuracy 96.37%, precision 96.60%, recall 99.68%, F1-score 98.12%, determining its robust predictive ability. performed little (0.9550) 97.78%, 98.69%, 99.01%, 98.85%, ability interpretability. These findings confirm potential for applying AI-based enhance risk assessment. provides informative results analysts, policymakers, corporate managers support early intervention strategies. Additional research would be required explore state-of-the-art AI techniques further refine forecasting decision-making sectors like agriculture.

Language: Английский

---