Explaining deep learning models for ozone pollution prediction via embedded feature selection

Applied Soft Computing, Journal Year: 2024, Volume and Issue: 157, P. 111504 - 111504

Published: March 22, 2024

Ambient air pollution is a pervasive global issue that poses significant health risks. Among pollutants, ozone (O3) responsible for an estimated 1 to 1.2 million premature deaths yearly. Furthermore, O3 adversely affects climate warming, crop productivity, and more. Its formation occurs when nitrogen oxides volatile organic compounds react with short-wavelength solar radiation. Consequently, urban areas high traffic volume elevated temperatures are particularly prone levels, which pose risk their inhabitants. In response this problem, many countries have developed web mobile applications provide real-time information using sensor data. However, while these offer valuable insight into current predicting future pollutant behavior crucial effective planning mitigation strategies. Therefore, our main objectives develop accurate efficient prediction models identify the key factors influence levels. We adopt time series forecasting approach address objectives, allows us analyze predict behavior. Additionally, we tackle feature selection problem most relevant features periods contribute accuracy by introducing novel method called Time Selection Layer in Deep Learning models, significantly improves model performance, reduces complexity, enhances interpretability. Our study focuses on data collected from five representative Seville, Cordova, Jaen provinces Spain, multiple sensors capture comprehensive compare performance of three models: Lasso, Decision Tree, without incorporating Layer. results demonstrate including effectiveness interpretability achieving average improvement 9% across all monitored areas.

Language: Английский

---

Fault diagnosis of uncertain photovoltaic systems using deep recurrent neural networks based Lissajous curves

Computers & Electrical Engineering, Journal Year: 2025, Volume and Issue: 123, P. 110191 - 110191

Published: Feb. 27, 2025

Language: Английский

---

Hour-Ahead Photovoltaic Power Prediction Combining BiLSTM and Bayesian Optimization Algorithm, with Bootstrap Resampling for Interval Predictions

Sensors, Journal Year: 2024, Volume and Issue: 24(3), P. 882 - 882

Published: Jan. 29, 2024

Photovoltaic (PV) power prediction plays a critical role amid the accelerating adoption of renewable energy sources. This paper introduces bidirectional long short-term memory (BiLSTM) deep learning (DL) model designed for forecasting photovoltaic one hour ahead. The dataset under examination originates from small PV installation located at Polytechnic School University Alcala. To improve quality historical data and optimize performance, robust preprocessing algorithm is implemented. BiLSTM synergistically combined with Bayesian optimization (BOA) to fine-tune its primary hyperparameters, thereby enhancing predictive efficacy. performance proposed evaluated across diverse meteorological seasonal conditions. In deterministic forecasting, findings indicate superiority over alternative models employed in this research domain, specifically multilayer perceptron (MLP) neural network random forest (RF) ensemble model. Compared MLP RF reference models, achieves reductions normalized mean absolute error (nMAE) 75.03% 77.01%, respectively, demonstrating effectiveness type prediction. Moreover, interval utilizing bootstrap resampling method conducted, acquired intervals carefully adjusted meet desired confidence levels, robustness flexibility predictions.

Language: Английский

---

Explainable deep learning on multi-target time series forecasting: an air pollution use case

Results in Engineering, Journal Year: 2024, Volume and Issue: 24, P. 103290 - 103290

Published: Nov. 7, 2024

Language: Английский

---

A hybrid Bayesian optimization-based deep learning model for modeling the condition of saltwater pipes in Hong Kong

International Journal of Construction Management, Journal Year: 2024, Volume and Issue: unknown, P. 1 - 17

Published: Feb. 5, 2024

The water distribution network is one of the most pivotal services for communities, and that's why its effective secure operation crucial to growth national global economies. To this end, paramount objective study devise an automated self-adaptive model deterioration prediction saltwater pipes. developed envisioned on coupling deep learning with Bayesian optimization (HBO-DL) forecasting condition different material categories pipes stepping their pip-related, soil-related, operational-related, environmental-related features. In regard, leveraged amplify training mechanism neural through iterative hyper parameters. validated several folds validation that encompass performance evaluation, statistical analysis, graphical comparison, unified ranking. conducted comparative analysis evinced HBO-DL managed significantly perform better than feed forward network, support vector machines Gaussian process regression by 76.85%, 73.31% 79.08%, respectively. can stand as a practical useful tool forecast networks which aids municipalities in designing optimum intervention plans evading socioeconomic losses elicited from pipe failures.

Language: Английский

---

Handover algorithm based on Bayesian-optimized LSTM and multi-attribute decision making for heterogeneous networks

Ad Hoc Networks, Journal Year: 2024, Volume and Issue: 157, P. 103454 - 103454

Published: Feb. 17, 2024

Language: Английский

---

Runoff simulation in the upper Han River Basin using physics-informed machine learningmodel

Journal of Lake Sciences, Journal Year: 2025, Volume and Issue: 37(1), P. 279 - 292

Published: Jan. 1, 2025

Language: Английский

---

A Deep Learning-Based Cyberattack Detection Method for Line Differential Relays

Internet of Things, Journal Year: 2025, Volume and Issue: unknown, P. 101574 - 101574

Published: March 1, 2025

Language: Английский

---

A novel approach based on clustering and optimized ensemble deep learning for energy consumption forecasting in Ethiopia

Neurocomputing, Journal Year: 2025, Volume and Issue: unknown, P. 130027 - 130027

Published: March 1, 2025

Language: Английский

---

Hyperparameter optimization of machine learning models for predicting actual evapotranspiration

Machine Learning with Applications, Journal Year: 2025, Volume and Issue: 20, P. 100661 - 100661

Published: May 6, 2025

Language: Английский

---