Air Quality Class Prediction Using Machine Learning Methods Based on Monitoring Data and Secondary Modeling

Atmosphere, Journal Year: 2024, Volume and Issue: 15(5), P. 553 - 553

Published: April 30, 2024

Addressing the constraints inherent in traditional primary Air Quality Index (AQI) forecasting models and shortcomings exploitation of meteorological data, this research introduces a novel air quality prediction methodology leveraging machine learning enhanced modeling secondary data. The dataset employed encompasses forecast data on pollutant concentrations conditions, alongside actual observations concentration measurements, spanning from 23 July 2020 to 13 2021, sourced long-term projections at various monitoring stations within Jinan, China. Initially, through rigorous correlation analysis, ten factors were selected, comprising both measured forecasted across five categories each. Subsequently, significance these was assessed ranked based their impact different concentrations, utilizing combination univariate multivariate analyses random forest approach. Seasonal characteristic analysis highlighted distinct seasonal impacts temperature, humidity, pressure, general atmospheric conditions six key pollutants. performance evaluation learning-based classification revealed Light Gradient Boosting Machine (LightGBM) classifier as most effective, achieving an accuracy rate 97.5% F1 score 93.3%. Furthermore, experimental results for AQI indicated Long Short-Term Memory (LSTM) model superior, demonstrating goodness-of-fit 91.37% predictions, 90.46% O3 perfect fit test set. Collectively, findings affirm reliability efficacy forecasting.

Language: Английский

---

Transforming Dermatopathology With AI: Addressing Bias, Enhancing Interpretability, and Shaping Future Diagnostics

Dermatological Reviews, Journal Year: 2025, Volume and Issue: 6(1)

Published: Jan. 17, 2025

ABSTRACT Background Artificial intelligence (AI) is transforming dermatopathology by enhancing diagnostic accuracy, efficiency, and precision medicine. Despite its promise, challenges such as dataset biases, underrepresentation of diverse populations, limited transparency hinder widespread adoption. Addressing these gaps can set a new standard for equitable patient‐centered care. To evaluate how AI mitigates improves interpretability, promotes inclusivity in while highlighting novel technologies like multimodal models explainable (XAI). Results AI‐driven tools demonstrate significant improvements precision, particularly through that integrate histological, genetic, clinical data. Inclusive frameworks, the Monk scale, advanced segmentation methods effectively address biases. However, “black box” nature AI, ethical concerns about data privacy, access to low‐resource settings remain. Conclusion offers transformative potential dermatopathology, enabling equitable, innovative diagnostics. Overcoming persistent will require collaboration among dermatopathologists, developers, policymakers. By prioritizing inclusivity, transparency, interdisciplinary efforts, redefine global standards foster

Language: Английский

---

Skin-lesion segmentation using boundary-aware segmentation network and classification based on a mixture of convolutional and transformer neural networks

Frontiers in Medicine, Journal Year: 2025, Volume and Issue: 12

Published: March 10, 2025

Background Skin cancer is one of the most prevalent cancers worldwide. In clinical domain, skin lesions such as melanoma detection are still a challenge due to occlusions, poor contrast, image quality, and similarities between lesions. Deep-/machine-learning methods used for early, accurate, efficient Therefore, we propose boundary-aware segmentation network (BASNet) model comprising prediction residual refinement modules. Materials The module works like U-Net densely supervised by an encoder decoder. A hybrid loss function used, which has potential help in domain dermatology. BASNet handles these challenges providing robust outcomes, even suboptimal imaging environments. This leads accurate early diagnosis, improved treatment workflows. We further compact convolutional transformer (CCTM) based on convolution transformers classification. was designed selected number layers hyperparameters having two convolutions, transformers, 64 projection dimensions, tokenizer, position embedding, sequence pooling, MLP, batch size, heads, 0.1 stochastic depth, 0.001 learning rate, 0.0001 weight decay, 100 epochs. Results CCTM evaluated six skin-lesion datasets, namely MED-NODE, PH2, ISIC-2019, ISIC-2020, HAM10000, DermNet achieving over 98% accuracy. Conclusion proposed holds significant domain. Its ability combine local feature extraction global context understanding makes it ideal tasks medical analysis disease diagnosis.

Language: Английский

---

Influence of feature-to-feature interactions on chloride migration in type-I cement concrete: A robust modeling approach using extra random forest

Materials Today Communications, Journal Year: 2024, Volume and Issue: 40, P. 109419 - 109419

Published: May 31, 2024

Language: Английский

---

Predicting Paediatric Brain Disorders from MRI Images Using Advanced Deep Learning Techniques

Neuroinformatics, Journal Year: 2025, Volume and Issue: 23(2)

Published: Jan. 16, 2025

Language: Английский

---

Optimized deep transfer learning techniques for spine fracture detection using CT scan images

Multimedia Tools and Applications, Journal Year: 2025, Volume and Issue: unknown

Published: Feb. 8, 2025

Language: Английский

---

Early Detection of Skin Diseases Across Diverse Skin Tones Using Hybrid Machine Learning and Deep Learning Models

Information, Journal Year: 2025, Volume and Issue: 16(2), P. 152 - 152

Published: Feb. 19, 2025

Skin diseases in melanin-rich skin often present diagnostic challenges due to the unique characteristics of darker tones, which can lead misdiagnosis or delayed treatment. This disparity impacts millions within diverse communities, highlighting need for accurate, AI-based tools. In this paper, we investigated performance three machine learning methods -Support Vector Machines (SVMs), Random Forest (RF), and Decision Trees (DTs)-combined with state-of-the-art (SOTA) deep models, EfficientNet, MobileNetV2, DenseNet121, predicting conditions using dermoscopic images from HAM10000 dataset. The features were extracted labels encoded numerically. To address data imbalance, SMOTE resampling techniques applied. Additionally, Principal Component Analysis (PCA) was used feature reduction, fine-tuning performed optimize models. results demonstrated that RF DenseNet121 achieved a superior accuracy 98.32%, followed by SVM MobileNetV2 at 98.08%, Tree 85.39%. proposed overcome SOTA EfficientNet model, validating robustness approaches. Evaluation metrics such as accuracy, precision, recall, F1-score benchmark performance, showcasing potential these advancing disease diagnostics populations.

Language: Английский

---

MFENet: Multi-scale and Local Frequency Enhancement Network for Skin Lesion Classification

Lecture notes in computer science, Journal Year: 2025, Volume and Issue: unknown, P. 192 - 203

Published: Jan. 1, 2025

Language: Английский

---

MOAT: MobileNet‐Optimized Attention Transfer for Robust and Scalable Dermatology Image Classification

Microscopy Research and Technique, Journal Year: 2025, Volume and Issue: unknown

Published: March 8, 2025

In recent times, dermatological disease is a common health issue around the world. Timely and accurate detection mandatory for proper treatment planning improving patient outcomes. Prior to this, various classification methodologies were developed early prediction of skin using dermatology images. However, they have struggled with poor accuracy require high computational time. This research proposes novel MobileNet-Optimized Attention Transfer framework disease. this study, MobileNet model deployed feature extraction, which integrates self-attention cross-attention mechanisms. The attention mechanisms prioritize essential features within images allow identify subtle patterns associated conditions. For hyperparameter tuning, an Optical Microscope Algorithm initial search strategy applied. algorithm iteratively fine-tunes parameters balance global local prevent from converging on suboptimal configurations. performance proposed method validated Skin Cancer ISIC dataset MNIST: HAM10000 compared existing in terms some assessing metrics. experimental results demonstrate that effectively classified achieves 98.89%, lower mean squared error 0.186, low time 1.53 s methodologies. result indicates scalable, adaptable solution clinical applications, aiding dermatologists reliable diagnostic support enabling intervention.

Language: Английский

---

Generalized Matrix Learning Vector Quantization Computational Method for Intelligent Decision Making: A Systematic Literature Review

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

Published: March 10, 2025

Language: Английский

---