Forecasting Stock Market Indices Using Integration of Encoder, Decoder, and Attention Mechanism DOI Creative Commons
Tien Thanh Thach

Entropy, Journal Year: 2025, Volume and Issue: 27(1), P. 82 - 82

Published: Jan. 17, 2025

Accurate forecasting of stock market indices is crucial for investors, financial analysts, and policymakers. The integration encoder decoder architectures, coupled with an attention mechanism, has emerged as a powerful approach to enhance prediction accuracy. This paper presents novel framework that leverages these components capture complex temporal dependencies patterns within price data. effectively transforms input sequence into dense representation, which the then uses reconstruct future values. mechanism provides additional layer sophistication, allowing model selectively focus on relevant parts making predictions. Furthermore, Bayesian optimization employed fine-tune hyperparameters, further improving forecast precision. Our results demonstrate significant improvement in precision over traditional recurrent neural networks. indicates potential our integrated handle

Language: Английский

Improving predictions of shale wettability using advanced machine learning techniques and nature-inspired methods: Implications for carbon capture utilization and storage DOI

Hemeng Zhang,

Hung Vo Thanh, Mohammad Rahimi

et al.

The Science of The Total Environment, Journal Year: 2023, Volume and Issue: 877, P. 162944 - 162944

Published: March 20, 2023

Language: Английский

Citations

48
Harmonizing Macro-Financial Factors and Twitter Sentiment Analysis in Forecasting Stock Market Trends DOI Creative Commons
Md Shahedul Amin,

Eftekhar Hossain Ayon,

Bishnu Padh Ghosh

et al.

Journal of Computer Science and Technology Studies, Journal Year: 2024, Volume and Issue: 6(1), P. 58 - 67

Published: Jan. 7, 2024

The surge in generative artificial intelligence technologies, exemplified by systems such as ChatGPT, has sparked widespread interest and discourse prominently observed on social media platforms like Twitter. This paper delves into the inquiry of whether sentiment expressed tweets discussing advancements AI can forecast day-to-day fluctuations stock prices associated companies. Our investigation involves analysis containing hashtags related to ChatGPT within timeframe December 2022 March 2023. Leveraging natural language processing techniques, we extract features, including positive/negative scores, from collected tweets. A range classifier machine learning models, encompassing gradient boosting, decision trees random forests, are employed train tweet sentiments features for prediction price movements among key companies, Microsoft OpenAI. These models undergo training testing phases utilizing an empirical dataset gathered during stipulated timeframe. preliminary findings reveal intriguing indications suggesting a plausible correlation between public reflected Twitter discussions surrounding subsequent impact market valuation trading activities concerning pertinent gauged through prices. study aims bullish or bearish trends leveraging derived extensive comprising 500,000 In conjunction with this Twitter, incorporate control variables macroeconomic indicators, uncertainty index data several prominent

Language: Английский

Citations

17
Multivariate gated recurrent unit for battery remaining useful life prediction: A deep learning approach DOI Open Access
Reza Rouhi Ardeshiri, Chengbin Ma

International Journal of Energy Research, Journal Year: 2021, Volume and Issue: 45(11), P. 16633 - 16648

Published: June 5, 2021

This paper proposes the gated recurrent unit (GRU)-recurrent neural network (RNN), a deep learning approach to predict remaining useful life (RUL) of lithium-ion batteries (LIBs), accurately. The GRU-RNN structure can self-learn parameters utilizing adaptive gradient descent algorithms, leading reduced computational cost. Unlike long short-term memory (LSTM) model, allows time-series dependencies be tracked between degraded capacities without using any cell. enables method non-linear capacity degradations and build an explicitly capacity-oriented RUL predictor. Additionally, feature selection based on random forest technique was used enhance prediction precision. analyses were conducted four separate cycling testing datasets battery. experimental results indicate that average percentage root mean square error for proposed is about 2% which respectively 1.34 times 8.32 superior LSTM support vector machine methods. outcome this work managing Li-ion battery's improvement optimization.

Language: Английский

Citations

66
Enhancing carbon sequestration: Innovative models for wettability dynamics in CO2-brine-mineral systems DOI
Hung Vo Thanh,

Hemeng Zhang,

Mohammad Rahimi

et al.

Journal of environmental chemical engineering, Journal Year: 2024, Volume and Issue: 12(5), P. 113435 - 113435

Published: June 26, 2024

Language: Английский

Citations

13
Mapping stock market dynamics: A tripartite neural network approach using modified grid search for stock market prediction DOI
Sachin Singh, Mohinder Singh, Shradha Attri

et al.

Expert Systems with Applications, Journal Year: 2025, Volume and Issue: unknown, P. 127243 - 127243

Published: March 1, 2025

Language: Английский

Citations

1
Stock values predictions using deep learning based hybrid models DOI Creative Commons

Konark Yadav,

Milind Yadav,

Sandeep Saini

et al.

CAAI Transactions on Intelligence Technology, Journal Year: 2021, Volume and Issue: 7(1), P. 107 - 116

Published: June 24, 2021

Predicting the correct values of stock prices in fast fluctuating high-frequency financial data is always a challenging task. A deep learning-based model for live predictions aimed to be developed here. The authors' have proposed two models different applications. first one based on Fast Recurrent Neural Networks (Fast RNNs). This used price time this work. second hybrid learning by utilising best features FastRNNs, Convolutional Networks, and Bi-Directional Long Short Term Memory predict abrupt changes company. 1-min interval four companies period three days considered. Along with lower Root Mean Squared Error (RMSE), low computational complexity as well, so that they can also predictions. models' performance measured RMSE along computation time. outperforms Auto Regressive Integrated Moving Average, FBProphet, LSTM, other both values.

Language: Английский

Citations

51
Groundwater-Potential Mapping Using a Self-Learning Bayesian Network Model: A Comparison among Metaheuristic Algorithms DOI Open Access
Sadegh Karimi-Rizvandi,

Hamid Valipoori Goodarzi,

Javad Hatami Afkoueieh

et al.

Water, Journal Year: 2021, Volume and Issue: 13(5), P. 658 - 658

Published: Feb. 28, 2021

Owing to the reduction of surface-water resources and frequent droughts, exploitation groundwater has faced critical challenges. For optimal management these valuable resources, careful studies potential status are essential. The main goal this study was determine network structure a Bayesian (BayesNet) machine-learning model using three metaheuristic optimization algorithms—a genetic algorithm (GA), simulated annealing (SA) algorithm, Tabu search (TS) algorithm—to prepare groundwater-potential maps. methodology applied town Baghmalek in Khuzestan province Iran. modeling, location 187 springs area 13 parameters (altitude, slope angle, aspect, plan curvature, profile topography wetness index (TWI), distance river, fault, drainage density, rainfall, land use/cover, lithology, soil) affecting were provided. In addition, statistical method certainty factor (CF) utilized input weight hybrid models. results OneR technique showed that altitude, density more important for compared other parameters. mapping (GPM) employing receiver operating characteristic (ROC) under curve (AUC) an estimation accuracy 0.830, 0.818, 0.810, 0.792, BayesNet-GA, BayesNet-SA, BayesNet-TS, BayesNet models, respectively. BayesNet-GA improved GPM BayesNet-SA (4.6% 7.5%) BayesNet-TS (21.8% 17.5%) models with respect root mean square error (RMSE) absolute (MAE), Based on metric indices, GA provides higher capability than SA TS algorithms optimizing determining GPM.

Language: Английский

Citations

41
A Comprehensive Survey on Higher Order Neural Networks and Evolutionary Optimization Learning Algorithms in Financial Time Series Forecasting DOI
Sudersan Behera, Sarat Chandra Nayak, A. V. S. Pavan Kumar

et al.

Archives of Computational Methods in Engineering, Journal Year: 2023, Volume and Issue: 30(7), P. 4401 - 4448

Published: May 23, 2023

Language: Английский

Citations

18
Optimizing agricultural management in China for soil greenhouse gas emissions and yield balance: A regional heterogeneity perspective DOI
Hanbing Li, Xiaobin Jin, Wei Shan

et al.

Journal of Cleaner Production, Journal Year: 2024, Volume and Issue: 452, P. 142255 - 142255

Published: April 15, 2024

Language: Английский

Citations

8
Enhanced multi-layer perceptron for CO2 emission prediction with worst moth disrupted moth fly optimization (WMFO) DOI Creative Commons
Oluwatayomi Rereloluwa Adegboye, Ezgi Deniz Ülker, Afi Kekeli Feda

et al.

Heliyon, Journal Year: 2024, Volume and Issue: 10(11), P. e31850 - e31850

Published: May 27, 2024

This study introduces the Worst Moth Disruption Strategy (WMFO) to enhance Fly Optimization (MFO) algorithm, specifically addressing challenges related population stagnation and low diversity. The WMFO aims prevent local trapping of moths, fostering improved global search capabilities. Demonstrating a remarkable efficiency 66.6 %, outperforms MFO on CEC15 benchmark test functions. Friedman Wilcoxon tests further confirm WMFO's superiority over state-of-the-art algorithms. Introducing hybrid model, WMFO-MLP, combining with Multi-Layer Perceptron (MLP), facilitates effective parameter tuning for carbon emission prediction, achieving an outstanding total accuracy 97.8 %. Comparative analysis indicates that MLP-WMFO model surpasses alternative techniques in precision, reliability, efficiency. Feature importance reveals variables such as Oil Efficiency Economic Growth significantly impact MLP-WMFO's predictive power, contributing up 40 Additionally, Gas Efficiency, Renewable Energy, Financial Risk, Political Risk explain 26.5 13.6 8 6.5 respectively. Finally, WMFO-MLP performance offers advancements optimization modeling practical applications prediction.

Language: Английский

Citations

7