Time series prediction of sea surface temperature based on BiLSTM model with attention mechanism

Journal of Sea Research, Journal Year: 2024, Volume and Issue: 198, P. 102472 - 102472

Published: Jan. 24, 2024

With the advancement of technology, ocean observation techniques have become increasingly prevalent in estimating marine variables such as Sea Surface Temperature (SST). This progress has led to a substantial surge volume data. Presently, abundance available data presents remarkable opportunity for training predictive models. The prediction SST poses challenge due its temporal-dependent structure and multi-level seasonality. In this study, we propose deep learning approach that combines Bidirectional Long Short-Term Memory (BiLSTM) model with attention mechanism forecast SST. By leveraging BiLSTM's ability effectively capture long-term dependencies through both forward backward LSTM processing, accentuates salient features, thereby enhancing model's evaluation accuracy. To evaluate effectiveness Attention-BiLSTM predicting SST, conducted case study Moroccan Sea, focusing on four distinct regions. We compared performance against alternative models LSTM, Attention-BiGRU, XGBoost, Random Forest (RF), Support Vector Regression (SVR), Transformers forecasting time series. experimental results unequivocally demonstrate achieves significantly superior outcomes is good candidate deployment field.

Language: Английский

---

Forecasting of fine particulate matter based on LSTM and optimization algorithm

Journal of Cleaner Production, Journal Year: 2023, Volume and Issue: 427, P. 139233 - 139233

Published: Oct. 10, 2023

Language: Английский

---

Improving the accuracy of multi-step prediction of building energy consumption based on EEMD-PSO-Informer and long-time series

Computers & Electrical Engineering, Journal Year: 2023, Volume and Issue: 110, P. 108845 - 108845

Published: July 18, 2023

Language: Английский

---

A multi-agent reinforcement learning framework for optimizing financial trading strategies based on TimesNet

Expert Systems with Applications, Journal Year: 2023, Volume and Issue: 237, P. 121502 - 121502

Published: Sept. 11, 2023

Language: Английский

---

Comprehensive Analysis of Air Quality Trends in India Using Machine Learning and Deep Learning Models

Published: Jan. 2, 2025

Language: Английский

---

Advancing harmful algal bloom predictions using chlorophyll-a as an Indicator: Combining deep learning and EnKF data assimilation method

Journal of Environmental Management, Journal Year: 2025, Volume and Issue: 382, P. 125441 - 125441

Published: April 19, 2025

Language: Английский

---

Transformer-Based Deep Learning Models for State of Charge and State of Health Estimation of Li-Ion Batteries: A Survey Study

Energies, Journal Year: 2024, Volume and Issue: 17(14), P. 3502 - 3502

Published: July 17, 2024

The global transportation system’s need for electrification is driving research efforts to overcome the drawbacks of battery electric vehicles (BEVs). accurate and reliable estimation states charge (SOC) health (SOH) Li-Ion batteries (LIBs) crucial widespread adoption BEVs. Transformers, cutting-edge deep learning (DL) models, are demonstrating promising capabilities in addressing various sequence-processing problems. This manuscript presents a thorough survey study previous papers that introduced modifications development Transformer-based architectures SOC SOH LIBs. also highlights approximately 15 different real-world datasets have been utilized training testing these models. A comparison made between architectures, each state using root mean square error (RMSE) absolute (MAE) metrics.

Language: Английский

---

Long Short-Term Memory Recurrent Neural Network and Extreme Gradient Boosting Algorithms Applied in a Greenhouse’s Internal Temperature Prediction

Applied Sciences, Journal Year: 2023, Volume and Issue: 13(22), P. 12341 - 12341

Published: Nov. 15, 2023

One of the main challenges agricultural greenhouses face is accurately predicting environmental conditions to ensure optimal crop growth. However, current prediction methods have limitations in handling large volumes dynamic and nonlinear temporal data, which makes it difficult make accurate early predictions. This paper aims forecast a greenhouse’s internal temperature up one hour advance using supervised learning tools like Extreme Gradient Boosting (XGBoost) Recurrent Neural Networks combined with Long-Short Term Memory (LSTM-RNN). The study uses many-to-one configuration, sequence three input elements output element. Significant improvements R2, RMSE, MAE, MAPE metrics are observed by considering various combinations. In addition, Bayesian optimization employed find best hyperparameters for each algorithm. research database data such as temperature, humidity, dew point external solar radiation, splitting into year’s four seasons performing eight experiments according two algorithms season. LSTM-RNN model produces results summer, achieving an R2 = 0.9994, RMSE 0.2698, MAE 0.1449, 0.0041, meeting acceptability criterion ±2 °C hysteresis.

Language: Английский

---

Novel MIA-LSTM Deep Learning Hybrid Model with Data Preprocessing for Forecasting of PM2.5

Algorithms, Journal Year: 2023, Volume and Issue: 16(1), P. 52 - 52

Published: Jan. 12, 2023

Day by day pollution in cities is increasing due to urbanization. One of the biggest challenges posed rapid migration inhabitants into increased air pollution. Sustainable Development Goal 11 indicates that 99 percent world’s urban population breathes polluted air. In such a trend urbanization, predicting concentrations pollutants advance very important. Predictions would help city administrations take timely measures for ensuring 11. data engineering, imputation and removal outliers are important steps prior forecasting concentration pollutants. For meteorological data, missing values critical problems need be addressed. This paper proposes novel method called multiple iterative using autoencoder-based long short-term memory (MIA-LSTM) which uses an extra tree regressor as estimator multivariate followed LSTM autoencoder detection present dataset. The preprocessed were given PM2.5 concentration. also presents effect removing from dataset well imputing process proposed provides better results with root mean square error (RMSE) value 9.8883. obtained compared traditional gated recurrent unit (GRU), 1D convolutional neural network (CNN), (LSTM) approaches Aotizhonhxin area Beijing China. Similar observed another two locations China one location India. show outlier/anomaly improve accuracy forecasting.

Language: Английский

---

S&P-500 vs. Nasdaq-100 price movement prediction with LSTM for different daily periods

Machine Learning with Applications, Journal Year: 2025, Volume and Issue: unknown, P. 100617 - 100617

Published: Jan. 1, 2025

Language: Английский

---