Article
Engineering, Electrical & Electronic
Xia Zhao, Zhao Li, Chen Zhao, Chang Wang
Summary: Accurately predicting a driver's head pose is crucial for distraction warning and vehicle takeover rule-making in human-machine codriving. This study proposes an innovative scheme for multistep time-series prediction, integrating machine learning and geometric methods. The proposed hybrid CNN-BiLSTM-attention model outperforms traditional machine-learning-based models and provides reliable driver's head pose data.
IEEE SENSORS JOURNAL
(2022)
Article
Engineering, Aerospace
Weijie Ding, Jin Huang, Guanyu Shang, Xuexuan Wang, Baoqiang Li, Yunfei Li, Hourong Liu
Summary: This paper proposes a trajectory prediction model based on CNN-BiLSTM network, dual attention, and genetic algorithm optimization, aiming to address the problems in existing trajectory prediction. Experimental results show that this model has significant advantages in prediction accuracy and applicability.
Article
Computer Science, Information Systems
Wanting Qin, Jun Tang, Songyang Lao
Summary: Mining key information from trajectory data can effectively help people in their life. This paper proposes a trajectory prediction model based on deep feature representation, which extracts various features to achieve more accurate and stable trajectory prediction.
INFORMATION SCIENCES
(2022)
Article
Electrochemistry
Yulin Peng, Tao Chen, Fei Xiao, Shaojie Zhang
Summary: In this study, a method based on convolutional neural network and long short-term memory was proposed to predict the remaining useful lifetime of proton exchange membrane fuel cells. The experiments demonstrated that this method can quickly and accurately predict both long-term and short-term remaining useful lifetime.
Article
Computer Science, Information Systems
Yajing Guo, Xiujuan Lei, Lian Liu, Yi Pan
Summary: In this study, a method called circ2CBA is proposed to predict the binding sites between circRNAs and RBPs using only sequence information. The results show that circ2CBA is an effective method with an AUC value of 0.8987.
FRONTIERS OF COMPUTER SCIENCE
(2023)
Article
Biochemistry & Molecular Biology
Jing Liu, Pu Chen, Hongdong Song, Pengxiao Zhang, Man Wang, Zhenliang Sun, Xiao Guan
Summary: This study combines transfer learning and SMILES enumeration data augmentation strategy to predict new CCK-secretory peptides. The fusion model of hierarchical attention network and bidirectional long short-term memory efficiently predicts CCK-secretory peptides.
Article
Mathematics
Jilin Zhang, Lishi Ye, Yongzeng Lai
Summary: Accurate prediction of stock prices is crucial in stock investment, but it is challenging due to the characteristics of high frequency, nonlinearity, and long memory in stock price data. This paper proposes a CNN-BiLSTM-Attention-based model, which extracts temporal features using CNN and BiLSTM, incorporates attention mechanism to assign weights automatically, and outputs final prediction results through dense layer. Experimental results show that the proposed model outperforms LSTM, CNN-LSTM, and CNN-LSTM-Attention models in predicting Chinese stock index (CSI300) price, and it also demonstrates robust effectiveness in predicting other Chinese and international stock indices.
Article
Energy & Fuels
Liqun Shan, Yanchang Liu, Min Tang, Ming Yang, Xueyuan Bai
Summary: This study proposes a model using deep learning methods to predict missing well logging data, achieving higher prediction accuracy by combining bidirectional long short-term memory networks, attention mechanism, and convolutional neural networks.
JOURNAL OF PETROLEUM SCIENCE AND ENGINEERING
(2021)
Article
Multidisciplinary Sciences
Lei Nie, Lvfan Zhang, Shiyi Xu, Wentao Cai, Haoming Yang
Summary: This study proposes a tool wear evolution and RUL prediction method by combining CNN-BiLSTM and attention mechanism. By processing and extracting features from sensor data, the proposed method outperforms traditional methods in terms of prediction accuracy.
Article
Computer Science, Artificial Intelligence
Jing Bi, Zexian Chen, Haitao Yuan, Jia Zhang
Summary: This study proposes a hybrid prediction method called VBAED, combining VMD, bidirectional attention mechanism, BiLSTM, and bidirectional temporal attention mechanism, to predict water quality time series. Experimental results demonstrate that VBAED achieves the best prediction performance compared to other methods.
EXPERT SYSTEMS WITH APPLICATIONS
(2024)
Article
Computer Science, Information Systems
Tripti Singh, Mohan Mohadikar, Shilpa Gite, Shruti Patil, Biswajeet Pradhan, Abdullah Alamri
Summary: Automated human pose estimation is a growing research area, requiring reliable techniques for accurate pose estimation. This paper proposes a method for head pose estimation using deep neural network regression and elastic net regression, achieving high precision.
Article
Energy & Fuels
Shuxin Liu, Yankai Li, Shuyu Gao, Chaojian Xing, Jing Li, Yundong Cao
Summary: In this paper, the authors propose a novel approach that combines CNNs and BiLSTM to predict the residual electrical life of railway relays. By collecting voltage and current signals from relay tests and applying feature selection, they achieve an effective prediction accuracy of 96.3%. The proposed CNN BiLSTM model offers higher accuracy, better stability, and greater practicality compared to other prediction models.
Article
Energy & Fuels
Mingyue Zhang, Yang Han, Amr S. Zalhaf, Chaoyang Wang, Ping Yang, Congling Wang, Siyu Zhou, Tianlong Xiong
Summary: This paper proposes a combined ultra-short-term load forecasting model based on improved ICEEMDAN, CNN, and BiLSTM neural networks. The model effectively captures the non-linear characteristics of load changes and achieves high prediction accuracy.
SUSTAINABLE ENERGY GRIDS & NETWORKS
(2023)
Article
Engineering, Marine
Zhenguo Ji, Huibing Gan, Ben Liu
Summary: This paper proposes a hybrid neural network (HNN) prediction model based on deep learning (DL) for predicting the exhaust gas temperature (EGT) of marine diesel engines. The model uses CNN to extract features from time-series data, BiLSTM to predict the time series through modeling, and Attention to improve the accuracy and robustness of fault prediction. Comparison experiments with other neural network prediction models have shown that the CNN-BiLSTM-Attention method is more accurate.
JOURNAL OF MARINE SCIENCE AND ENGINEERING
(2023)
Article
Environmental Sciences
Zheng Zhao, Honggang Nan, Zihan Liu, Yuebo Yu
Summary: This paper proposes an ultra-short-term wind power multi-step interval prediction method based on CEEMDAN-FIG and CNN-BiLSTM. It decomposes the wind power time series and combines different components with other meteorological data to predict wind power intervals. The experimental results validate the effectiveness of the proposed model.
ENVIRONMENTAL SCIENCE AND POLLUTION RESEARCH
(2022)
