Article
Chemistry, Multidisciplinary
Xue Zhang, Helmut Kuehnelt, Wim De Roeck
Summary: This paper investigates the modeling of traffic noise using deep learning, aiming to predict traffic noise from real-life traffic data. By studying recurrent neural networks (RNN) and different architectures, such as many-to-one, many-to-many, and encoder-decoder, the research reveals that a multivariate bi-directional GRU model with many-to-many architecture achieves the best performance in terms of accuracy and computation efficiency. The trained model could potentially be used for real-time traffic noise predictions in smart cities, helping regulation and policy makers in mitigating noise levels.
APPLIED SCIENCES-BASEL
(2021)
Article
Chemistry, Physical
Songyang Li, Weiling Luan, Chang Wang, Ying Chen, Zixian Zhuang
Summary: A fusion prognostic framework based on Bi-LSTM, Bi-GRU, and ESN is proposed to enhance the reliability and durability of PEMFC. With a smaller amount of training data, short-term degradation prediction and remaining useful life estimation of PEMFC can be achieved. Experimental results show that the proposed fusion prognostic framework outperforms traditional machine learning methods and has significant implications for online testing and health management of PEMFC.
INTERNATIONAL JOURNAL OF HYDROGEN ENERGY
(2022)
Article
Engineering, Marine
Yuchao Wang, Hui Wang, Bin Zhou, Huixuan Fu
Summary: Research on ship roll prediction methods based on deep learning, proposing single input single output and multiple input single output methods, analyzing and testing with real data to verify the accuracy and effectiveness of the models.
Article
Computer Science, Theory & Methods
Abdellatif Bekkar, Badr Hssina, Samira Douzi, Khadija Douzi
Summary: The development of global human activities, industrialization, and urbanization has made air pollution a life-threatening factor, with PM2.5 being a serious pollutant that requires accurate prediction of its concentration to prevent harm. In this study, a deep learning algorithm CNN-LSTM was used to predict hourly PM2.5 concentration in Beijing based on historical and meteorological data, demonstrating better accuracy and performance compared to traditional models.
JOURNAL OF BIG DATA
(2021)
Article
Environmental Sciences
Beom-Jin Kim, You-Tae Lee, Byung-Hyun Kim
Summary: This study showed the process and method of selecting the optimal deep learning model for predicting dam inflow using hydrologic data over the past 20 years. The optimal models for Andong Dam and Imha Dam were selected for drought and typhoon conditions, showing closer prediction to the observed inflow than the SFM. The deep learning models were more accurate under various typhoon conditions, but the SFM showed better results under certain conditions. Comparing the inflow predictions of the SFM and deep learning models is necessary for efficient dam operation and management.
Article
Multidisciplinary Sciences
H. Canli, S. Toklu
Summary: This study proposed a new approach using deep learning-based GRU model to predict parking lot occupancy rate, gathering data from various car parks and weather conditions in Istanbul. Experimental results demonstrated the effectiveness of the GRU model in forecasting parking spot availability with high accuracy.
ARABIAN JOURNAL FOR SCIENCE AND ENGINEERING
(2022)
Article
Mathematics
Mohamed Hassan Essai Ali, Ali R. Abdellah, Hany A. Atallah, Gehad Safwat Ahmed, Ammar Muthanna, Andrey Koucheryavy
Summary: This study proposes a new method using deep learning with peephole LSTM for pilot-based channel estimation in OFDM. The proposed estimator outperforms conventional estimators in low SNR regions and with few pilots, without requiring prior knowledge of channel statistics. The DL-based peephole LSTM model shows outstanding learning and generalization properties, enabling robust recovery of transmitted data in an OFDM communication system.
Article
Environmental Sciences
Peng Hao, Shuang Li, Yu Gao
Summary: In this study, the predictive performance of significant wave height (SWH) using recurrent neural network (RNN), long short-term memory network (LSTM), and gated recurrent unit network (GRU) is comprehensively analyzed by considering different input lengths, prediction lengths, and model complexity. The results show that the input length and prediction length have an impact on the SWH prediction, but longer input length and longer prediction length do not necessarily lead to better prediction performance. Moreover, the number of layers in the model does not always determine the prediction performance.
FRONTIERS IN MARINE SCIENCE
(2023)
Article
Environmental Sciences
Muhammad Waqas Saif-ul-Allah, Muhammad Abdul Qyyum, Noaman Ul-Haq, Chaudhary Awais Salman, Faisal Ahmed
Summary: This paper introduces how to use recurrent neural networks and the missing value imputation method to predict the PM2.5 concentration of Guangzhou City, China. It demonstrates the superiority of this approach compared to other machine learning techniques.
FRONTIERS IN ENVIRONMENTAL SCIENCE
(2022)
Article
Computer Science, Information Systems
Eslam Amer, Kyung-Sup Kwak, Shaker El-Sappagh
Summary: With the widespread use of the internet, people now rely on social media as a convenient and low-cost source of news. However, the dissemination of fake news has become a pressing issue, distorting people's views and knowledge. This paper presents experimental findings that demonstrate the effectiveness of deep learning models in accurately identifying fake news.
Article
Biology
Gelany Aly Abdelkader, Soualihou Ngnamsie Njimbouom, Tae-Jin Oh, Jeong-Dong Kim
Summary: Protein-ligand interaction is crucial in drug discovery. This study presents ResBiGAAT, a novel deep learning model that combines a deep Residual Bidirectional Gated Recurrent Unit with two-sided self-attention mechanisms, to efficiently predict protein-ligand binding affinity. The model exhibits competitive performance and generalizability on both internal and external datasets.
COMPUTATIONAL BIOLOGY AND CHEMISTRY
(2023)
Article
Engineering, Marine
Jinwan Park, Jungsik Jeong, Youngsoo Park
Summary: According to maritime accident statistics, most collisions are caused by human factors. This paper proposes a method for predicting ship trajectories for intelligent collision avoidance at sea, utilizing spectral clustering and the Bi-LSTM model to improve prediction accuracy.
JOURNAL OF MARINE SCIENCE AND ENGINEERING
(2021)
Article
Green & Sustainable Science & Technology
Ling Qing
Summary: This paper proposes a method based on deep learning to predict PM2.5 concentration, which comprehensively considers various meteorological elements and analyzes the correlation between them and PM2.5 concentration. It uses the GRU model as the core network structure of the recurrent neural network, which is simpler and has better convergence time compared to the traditional LSTM model. Experimental results show that the method can achieve a root mean square error of 18.32 ug.m-3 and an average absolute error of 13.54 ug.m-3 in the range of 0-80 hours.
Article
Computer Science, Information Systems
Deep Kothadiya, Chintan Bhatt, Krenil Sapariya, Kevin Patel, Ana-Belen Gil-Gonzalez, Juan M. Corchado
Summary: This paper proposes a deep learning-based model to detect and recognize words from gestures, aiming to reduce communication barriers for individuals with impaired speech or hearing. Experimental results demonstrate that the proposed model achieves high accuracy in recognizing different gestures.
Article
Computer Science, Artificial Intelligence
Ahmed Sedik, Amr A. Abohany, Karam M. Sallam, Kumudu Munasinghe, T. Medhat
Summary: With the rise of social media and online forums, fake news has become a global source of news. This study proposes a deep learning-based method for detecting fake news and validates its effectiveness through experiments on multiple datasets.
EXPERT SYSTEMS WITH APPLICATIONS
(2022)