Research on stock prediction based on CED-PSO-StockNet time series model

Scientific Reports, Journal Year: 2024, Volume and Issue: 14(1)

Published: Nov. 10, 2024

To tackle the challenge of low accuracy in stock prediction within high-noise environments, this paper innovatively introduces CED-PSO-StockNet time series model. Initially, model decomposes raw data using Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) technique and reconstructs components by estimating their frequencies via extreme point method. This process enhances component stability mitigates noise interference. Subsequently, an Encoder-Decoder framework equipped attention mechanism is employed for precise reconstructed components, facilitating more effective extraction utilization features. Furthermore, utilizes Improved Particle Swarm Optimization (IPSO) algorithm to optimize parameters. On Pudong Bank dataset, through ablation experiments comparisons baseline models, various optimization strategies incorporated into proposed were effectively validated. Compared standalone LSTM model, achieved a remarkable 45.59% improvement R

Language: Английский

---

Integrative Analysis of Financial Market Sentiment Using CNN and GRU for Risk Prediction and Alert Systems

2022 International Conference on Electronics and Devices, Computational Science (ICEDCS), Journal Year: 2024, Volume and Issue: unknown, P. 410 - 415

Published: Sept. 23, 2024

Language: Английский

---

Forecasting Stock Market Indices Using Gated Recurrent Unit (GRU) Based Ensemble Models: LSTM-GRU

International Journal of Computer and Communication Technology, Journal Year: 2023, Volume and Issue: unknown, P. 85 - 90

Published: July 1, 2023

A "time sequence analysis" is a particular method for looking at group of data points gathered over long period time. Instead merely randomly or infrequently, time series analyzers gather information from predetermined length scheduled times. But this kind research requires more than just accumulating Data in may be analyzed to illustrate how variables change time, which makes them different other types data. To put it another way, crucial element since demonstrates the changes and outcomes. It offers architecture dependencies as well an extra source. Time Series forecasting field deep learning because many issues have temporal component. collection observations that are made sequentially across In study, we examine distinct machine learning, ensemble model algorithms predict Nike stock price. We going use price January 2006 2018 make predictions accordingly. The outcome hybrid LSTM-GRU outperformed models terms performance.

Language: Английский

---

Forex market directional trends forecasting with Bidirectional-LSTM and enhanced DeepSense network using all member-based optimizer

Intelligent Decision Technologies, Journal Year: 2023, Volume and Issue: 17(4), P. 1351 - 1382

Published: Sept. 26, 2023

This study focuses on successful Forex trading by emphasizing the importance of identifying market trends and utilizing trend analysis for informed decision-making. The authors collected low-correlated currency pair datasets to mitigate multicollinearity risk. Authors developed a two-stage predictive model that combines regression classification tasks, using predicted closing price determine entry exit points. incorporates Bi-directional long short-term memory (Bi-LSTM) improved forecasting higher highs lower lows (HHs-HLs LHs-LLs) identify changes. They proposed an enhanced DeepSense network (DSN) with all member-based optimization (AMBO-DSN) optimize decision variables DSN. performance models was compared various machine learning, deep statistical approaches including support vector regressor (SVR), artificial neural (ANN), auto-regressive integrated moving average (ARIMA), vanilla-LSTM (V-LSTM), recurrent (RNN). optimized form DSN genetic algorithm (GA), particle swarm (PSO), differential evolution (DE) AMBO-DSN, yielding satisfactory results demonstrated comparable quality observed original pairs. effectiveness reliability AMBO-DSN approach in USD/EUR, AUD/JPY, CHF/INR pairs were validated through while considering computational cost.

Language: Английский

---

Precipitation forecast using RNN variants by analyzing Optimizers and Hyperparameters for Time-series based Climatological Data

International journal of electrical and computer engineering systems, Journal Year: 2024, Volume and Issue: 15(3), P. 261 - 274

Published: Jan. 1, 2024

Flood is a significant problem in many regions of the world for catastrophic damage it causes to both property and human lives; excessive precipitation being major cause. The AI technologies, Deep Learning Neural Networks Machine algorithms attempt realistic solutions numerous disaster management challenges. This paper works on RNN- based rainfall/ forecasting models by investigating performances various Recurrent Network (RNN) architectures, Bidirectional RNN (BRNN), Long Short-Term Memory (LSTM), Gated Unit (GRU) ensemble such as BRNN-GRU, BRNN-LSTM, LSTM-GRU, BRNN-LSTM-GRU using NASAPOWER datasets Andhra Pradesh (AP) Tamil Nadu (TN) India. different stages workflow methodology are Data collection, pre-processing, splitting, Defining hyperparameters, Model building Performance evaluation. Experiments identifying improved optimizers hyperparameters time-series climatological data investigated accurate forecast. metrics: Mean Absolute Error (MAE), Squared (MSE), Root Square (RMSE) Logarithmic (RMSLE) values used compare predictions models. variants models, BRNN, LSTM, GRU, produce with RMSLE 2.448, 0.555, 0.255, 1.305, 1.383, 0.364, 1.740 AP 1.735, 0.663, 0.152, 0.889, 1.118, 0.379, 1.328 TN respectively. best performing model, GRU when ensembled existing statistical model SARIMA produces an value 0.754 1.677 respectively TN.

Language: Английский

---