A new denoising approach based on mode decomposition applied to the stock market time series: 2LE-CEEMDAN

PeerJ Computer Science, Journal Year: 2024, Volume and Issue: 10, P. e1852 - e1852

Published: Feb. 20, 2024

Time series, including noise, non-linearity, and non-stationary properties, are frequently used in prediction problems. Due to these inherent characteristics of time series data, forecasting based on this data type is a highly challenging problem. In many studies within the literature, high-frequency components commonly excluded from data. However, can contain valuable information, their removal may adversely impact performance models. study, novel method called Two-Level Entropy Ratio-Based Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (2LE-CEEMDAN) proposed for first effectively denoise Financial high noise levels utilized validate effectiveness method. The 2LE-CEEMDAN-LSTM-SVR model introduced predict next day’s closing value stock market indices scope financial series. This comprises two main components: denoising forecasting. section, 2LE-CEEMDAN eliminates resulting denoised intrinsic mode functions (IMFs). part, next-day estimated by training IMFs obtained. Two different artificial intelligence methods, Long Short-Term Memory (LSTM) Support Vector Regression (SVR), during process. IMF, characterized more linear than IMFs, trained using SVR, while others LSTM final result obtained integrating results each IMF. Experimental demonstrate that positively influences model’s performance, outperforms other models existing literature.

Language: Английский

---

Enhanced Multi-variate Time Series Prediction Through Statistical-Deep Learning Integration: The VAR-Stacked LSTM Model

SN Computer Science, Journal Year: 2024, Volume and Issue: 5(5)

Published: May 23, 2024

Language: Английский

---

Multi level perspectives in stock price forecasting: ICE2DE-MDL

PeerJ Computer Science, Journal Year: 2024, Volume and Issue: 10, P. e2125 - e2125

Published: June 24, 2024

This study proposes a novel hybrid model, called ICE2DE-MDL, integrating secondary decomposition, entropy, machine and deep learning methods to predict stock closing price. In this context, first of all, the noise contained in financial time series was eliminated. A denoising method, which utilizes entropy two-level ICEEMDAN methodology, is suggested achieve this. Subsequently, we applied many methods, including long-short term memory (LSTM), LSTM-BN, gated recurrent unit (GRU), SVR, IMFs obtained from classifying them as noiseless. Afterward, best training method determined for each IMF. Finally, proposed model's forecast by hierarchically combining prediction results The ICE2DE-MDL model eight market indices three data sets, next day's price these items predicted. indicate that RMSE values ranged 0.031 0.244, MAE 0.026 0.144, MAPE 0.128 0.594, R-squared 0.905 0.998 forecasts. Furthermore, comparisons were made with various models within scope forecasting evaluate performance model. Upon comparison, demonstrated superior relative existing literature both individual stocks. Additionally, our knowledge, effectively eliminate item using concepts ICEEMDAN. It also second apply problem.

Language: Английский

---

Predictive multi-period multi-objective portfolio optimization based on higher order moments: Deep learning approach

Computers & Industrial Engineering, Journal Year: 2023, Volume and Issue: 183, P. 109450 - 109450

Published: July 20, 2023

Language: Английский

---

Enhancing Stock Price Prediction with Deep Cross-Modal Information Fusion Network

Fluctuation and Noise Letters, Journal Year: 2023, Volume and Issue: 23(02)

Published: Nov. 23, 2023

Stock price prediction is considered a classic and challenging task, with the potential to aid traders in making more profitable trading decisions. Significant improvements stock methods based on deep learning have been observed recent years. However, most existing are reliant solely historical data for predictions, resulting inability capture market dynamics beyond indicators, thus limiting their performance some extent. Therefore, combining social media text information has proposed novel method, known as Deep Cross-Modal Information Fusion Network (DCIFNet). The process initiated by DCIFNet, which employs temporal convolution processes encode prices Twitter content. This ensures that each element sufficient about its surrounding components. Following this, outcomes inputted into cross-modal fusion structure transformers enhance integration of crucial from Lastly, multi-graph attention network introduced depict relationships between different stocks diverse perspectives. facilitates effective capturing industry affiliations, Wikipedia references, associated among linked stocks, ultimately leading an enhancement accuracy. Trend simulated experiments conducted high-frequency datasets spanning nine industries. Comparative assessments Multi-Attention Prediction (MANGSF) well ablation experiments, confirm effectiveness DCIFNet approach, accuracy rate 0.6309, marked improvement compared representative field.

Language: Английский

---

Enhanced stock price forecasting through a regularized ensemble framework with graph convolutional networks

Expert Systems with Applications, Journal Year: 2024, Volume and Issue: 250, P. 123948 - 123948

Published: April 11, 2024

Language: Английский

---

Analyzing the critical steps in deep learning-based stock forecasting: a literature review

PeerJ Computer Science, Journal Year: 2024, Volume and Issue: 10, P. e2312 - e2312

Published: Sept. 23, 2024

Stock market or individual stock forecasting poses a significant challenge due to the influence of uncertainty and dynamic conditions in financial markets. Traditional methods, such as fundamental technical analysis, have been limited coping with uncertainty. In recent years, this has led growing interest using deep learning-based models for prediction. However, accuracy reliability these depend on correctly implementing series critical steps. These steps include data collection feature extraction selection, noise elimination, model selection architecture determination, choice training-test approach, performance evaluation. This study systematically examined literature, investigating effects model’s performance. review focused studies between 2020–2024, identifying influential by conducting systematic literature search across three different databases. The identified regarding seven essential creating successful reliable prediction were thoroughly examined. findings from examinations summarized tables, gaps detailed. not only provides comprehensive understanding current but also serves guide future research.

Language: Английский

---

Bankacılık Sektörüne Derin Öğrenme Yöntemiyle Bakış: BİST Banka Endeksi Hareket Yönlerinin Tahmini

İnsan ve Toplum Bilimleri Araştırmaları Dergisi, Journal Year: 2024, Volume and Issue: unknown

Published: Aug. 8, 2024

Finansal sistemdeki temel oyuncular olan bankalar, ekonominin sağlıklı işlemesinde kritik bir rol oynamaktadırlar. Banka endeksleri ise, genellikle ülkenin finansal sektöründeki performansı yansıtarak ekonomik sağlığın göstergesi olarak kabul edilmektedir. BIST Endeksi, Türkiye'nin önde gelen banka hisselerini içeren endeks olup, bankacılık sektörünün performansını temsil etmektedir. Diğer yandan, hisse senedi fiyatlarının tahmin edilebilirliği, karmaşık ve değişken faktörlerle etkilenen konudur. piyasalarda amacıyla kullanılan analiz teknik gibi geleneksel yöntemlere ek olarak, son dönemde çok sayıda makine öğrenimi yöntemi geliştirilmiştir. Makine yöntemleri, serilerin doğrusal durağan olmayan özelliklerini ele alarak doğru tahminler yapabilmektedir. Tahmin uygulamalarındaki başarısı ile ön plana çıkan derin öğrenme büyük veri setlerini etkili şekilde işleyerek ilişkileri belirlemekte yüksek doğrulukla çıkarım Bu çalışmanın amacı, Endeksi’nin hareket yönlerinin edilmesidir. Analizde, Endeksi'nin 01.01.2013-31.12.2023 dönemindeki haftalık kapanış değerleriyle birlikte, yine bazda elde edilen mevduat kredi faiz oranları, gecelik hacimleri, sektörü aktif toplamı, döviz kurları (Dolar Euro) 100 endeksi değerleri girdi verisi kullanılmıştır. Her değişkeni için 574 edilmiş olup toplam 5.740 adet analizde Gerçekleştirilen sonucunda, yönleri %88,70 doğrulukta edilmiştir. Elde bulgular, kullanılarak belirli seviyede edilebileceğini göstermektedir.

---

Multi-scale contrast approach for stock index prediction with adaptive stock fusion

Expert Systems with Applications, Journal Year: 2024, Volume and Issue: unknown, P. 125590 - 125590

Published: Oct. 1, 2024

Language: Английский

---

Performance modeling of flame-assisted fuel cells based on a swirl burner

AIP Advances, Journal Year: 2024, Volume and Issue: 14(1)

Published: Jan. 1, 2024

Aiming at the problems of a narrow operating range and complex modeling Flame-assisted Fuel Cells (FFCs), an FFC system based on swirl burner is proposed, neural network algorithms are used to construct prediction model for polarization curve system. First, output voltage power values measured under different working conditions, various experimental parameters collected form dataset; second, correlation analysis method screen out that highly correlated with as input variables network; finally, constructed, back propagation (BP), long short term memory, 1D-CNN chosen examine applicability networks The characteristic results show can obtain maximum 10.6 V 7.71 W. average relative errors three 5.23%, 4.08%, 6.19%, respectively, BP algorithm showing best generalization ability. study provides support application in aerospace other fields.

Language: Английский

---