Deep Learning for Time Series Forecasting: A Survey

Big Data, Journal Year: 2020, Volume and Issue: 9(1), P. 3 - 21

Published: Dec. 4, 2020

Time series forecasting has become a very intensive field of research, which is even increasing in recent years. Deep neural networks have proved to be powerful and are achieving high accuracy many application fields. For these reasons, they one the most widely used methods machine learning solve problems dealing with big data nowadays. In this work, time problem initially formulated along its mathematical fundamentals. Then, common deep architectures that currently being successfully applied predict described, highlighting their advantages limitations. Particular attention given feed forward networks, recurrent (including Elman, long-short term memory, gated units, bidirectional networks), convolutional networks. Practical aspects, such as setting values for hyper-parameters choice suitable frameworks, successful also provided discussed. Several fruitful research fields analyzed obtained good performance reviewed. As result, gaps been identified literature several domains application, thus expecting inspire new better forms knowledge.

Language: Английский

---

Boosted mutation-based Harris hawks optimizer for parameters identification of single-diode solar cell models

Energy Conversion and Management, Journal Year: 2020, Volume and Issue: 209, P. 112660 - 112660

Published: March 9, 2020

Language: Английский

---

Network Intrusion Detection Based on PSO-Xgboost Model

IEEE Access, Journal Year: 2020, Volume and Issue: 8, P. 58392 - 58401

Published: Jan. 1, 2020

Network intrusion detection system (NIDS) is a commonly used tool to detect attacks and protect networks, while one of its general limitations the false positive issue. On basis our comparative experiments analysis for characteristics particle swarm optimization (PSO) Xgboost, this paper proposes PSO-Xgboost model given overall higher classification accuracy than other alternative models such like Random Forest, Bagging Adaboost. Firstly, based on Xgboost constructed, then PSO adaptively search optimal structure Xgboost. The benchmark NSL-KDD dataset evaluate proposed model. Our experimental results demonstrate that outperforms in precision, recall, macro-average (macro) mean average precision (mAP), especially when identifying minority groups U2R R2L. This work also provides arguments application intelligence NIDS.

Language: Английский

---

Frequency characteristics of FG-GPLRC viscoelastic thick annular plate with the aid of GDQM

Thin-Walled Structures, Journal Year: 2020, Volume and Issue: 150, P. 106683 - 106683

Published: Feb. 24, 2020

Language: Английский

---

An innovative coupled model in view of wavelet transform for predicting short-term PM10 concentration

Journal of Environmental Management, Journal Year: 2021, Volume and Issue: 289, P. 112438 - 112438

Published: April 16, 2021

Language: Английский

---

Electricity price forecasting with high penetration of renewable energy using attention-based LSTM network trained by crisscross optimization

Energy, Journal Year: 2022, Volume and Issue: 254, P. 124212 - 124212

Published: May 10, 2022

Language: Английский

---

Aggregating Prophet and Seasonal Trend Decomposition for Time Series Forecasting of Italian Electricity Spot Prices

Energies, Journal Year: 2023, Volume and Issue: 16(3), P. 1371 - 1371

Published: Jan. 29, 2023

The cost of electricity and gas has a direct influence on the everyday routines people who rely these resources to keep their businesses running. However, value is strongly related spot market prices, arrival winter increased energy use owing demand for heating can lead an increase in prices. Approaches forecasting costs have been used recent years; however, existing models are not yet robust enough due competition, seasonal changes, other variables. More effective modeling approaches required assist investors planning bidding strategies regulators ensuring security stability markets. In literature, there considerable interest building better pricing frameworks meet difficulties. this context, work proposes combining trend decomposition utilizing LOESS (locally estimated scatterplot smoothing) Facebook Prophet methodologies perform more accurate resilient time series analysis Italian This enhancing projections understanding variables driving data, while also including additional information such as holidays special events. combination improves forecast accuracy lowering mean absolute percentage error (MAPE) performance metric by 18% compared baseline model.

Language: Английский

---

Seasonal peak load prediction of underground gas storage using a novel two-stage model combining improved complete ensemble empirical mode decomposition and long short-term memory with a sparrow search algorithm

Energy, Journal Year: 2023, Volume and Issue: 274, P. 127376 - 127376

Published: March 30, 2023

Language: Английский

---

Forecasts of China Mainland New Energy Index Prices through Gaussian Process Regressions

Deleted Journal, Journal Year: 2024, Volume and Issue: 01

Published: Jan. 1, 2024

Energy index price forecasting has long been a crucial undertaking for investors and regulators. This study examines the daily predicting problem new energy on Chinese mainland market from January 4, 2016 to December 31, 2020 as insufficient attention paid in literature this financial metric. Gaussian process regressions facilitate our analysis, training procedures of models make use cross-validation Bayesian optimizations. From 2, 2020, was properly projected by created models, with an out-of-sample relative root mean square error 1.8837%. The developed may be utilized investors’ policymakers’ policy analysis decision-making processes. Because results provide reference information about patterns indicated they also useful building similar indices.

Language: Английский

---

Natural gas consumption forecasting using a novel two-stage model based on improved sparrow search algorithm

Journal of Pipeline Science and Engineering, Journal Year: 2024, Volume and Issue: 5(1), P. 100220 - 100220

Published: Aug. 23, 2024

The foundation of natural gas intelligent scheduling is the accurate prediction consumption (NGC). However, because its volatility, this brings difficulties and challenges in accurately predicting NGC. To address problem, an improved model developed combining sparrow search algorithm (ISSA), long short-term memory (LSTM), wavelet transform (WT). First, performance ISSA tested. Second, NGC divided into several high- low-frequency components applying different layers Coilfets', Fejer-Korovkins', Symletss', Haars', Discretes' orders. In addition, LSTM applied to forecast decomposed view one- multi-step, hyper-parameters are optimized by ISSA. At last, final results reconstructed. research indicate that: (1) Comparing other machine algorithms (e.g. fuzzy neural network), convergence speed stability stronger standard deviation mean; (2) better than that forecasting models; (3) single-step superior two-, three-, four- step; (4) computational load proposed highest compared models, accuracy still excellent on extended time series.

Language: Английский

---