Forecasting Stock Market Indices Using the Recurrent Neural Network Based Hybrid Models: CNN-LSTM, GRU-CNN, and Ensemble Models

Applied Sciences, Journal Year: 2023, Volume and Issue: 13(7), P. 4644 - 4644

Published: April 6, 2023

Various deep learning techniques have recently been developed in many fields due to the rapid advancement of technology and computing power. These widely applied finance for stock market prediction, portfolio optimization, risk management, trading strategies. Forecasting indices with noisy data is a complex challenging task, but it plays an important role appropriate timing buying or selling stocks, which one most popular valuable areas finance. In this work, we propose novel hybrid models forecasting one-time-step multi-time-step close prices DAX, DOW, S&P500 by utilizing recurrent neural network (RNN)–based models; convolutional network-long short-term memory (CNN-LSTM), gated unit (GRU)-CNN, ensemble models. We averaging high low as feature. The experimental results confirmed that our outperformed traditional machine-learning 48.1% 40.7% cases terms mean squared error (MSE) absolute (MAE), respectively, case 81.5% MSE MAE forecasting.

Language: Английский

---

Stock Market Prediction With Transductive Long Short-Term Memory and Social Media Sentiment Analysis

IEEE Access, Journal Year: 2024, Volume and Issue: 12, P. 87110 - 87130

Published: Jan. 1, 2024

In an era dominated by digital communication, the vast amounts of data generated from social media and financial markets present unique opportunities challenges for forecasting stock market prices. This paper proposes innovative approach that harnesses power sentiment analysis combined with to predict prices, directly addressing critical in this domain. A major challenge is uneven distribution across different categories. Traditional models struggle accurately identify fewer common sentiments (minority class) due overwhelming presence more (majority class). To tackle this, we introduce Off-policy Proximal Policy Optimization (PPO) algorithm, specifically designed handle class imbalance adjusting reward mechanism training phase, thus favoring correct classification minority instances. Another effectively integrating temporal dynamics prices results. Our solution implementing a Transductive Long Short-Term Memory (TLSTM) model incorporates findings historical data. excels at recognizing patterns gives precedence points are temporally closer prediction point, enhancing accuracy. Ablation studies confirm effectiveness PPO TLSTM components on overall performance. The proposed advances field analytics providing nuanced understanding but also offers actionable insights investors policymakers seeking navigate complexities greater precision confidence.

Language: Английский

---

Data vs. information: Using clustering techniques to enhance stock returns forecasting

International Review of Financial Analysis, Journal Year: 2023, Volume and Issue: 88, P. 102657 - 102657

Published: April 23, 2023

This paper explores the use of clustering models stocks to improve both (a) prediction stock prices and (b) returns trading algorithms. We cluster using k-means several alternative distance metrics, as features quarterly financial ratios, daily returns. Then, for each cluster, we train ARIMA LSTM forecasting predict price in cluster. Finally, employ clustering-empowered analyze different obtain three key results: (i) outperform benchmark models, obtaining positive investment scenarios; (ii) is improved by additional information provided methods, therefore selecting relevant data an important preprocessing task process; (iii) from whole sample deteriorates ability models. These results have been validated 240 companies Russell 3000 index spanning 2017 2022, training testing with subperiods.

Language: Английский

---

Deep Learning in Finance: A Survey of Applications and Techniques

AI, Journal Year: 2024, Volume and Issue: 5(4), P. 2066 - 2091

Published: Oct. 28, 2024

Machine learning (ML) has transformed the financial industry by enabling advanced applications such as credit scoring, fraud detection, and market forecasting. At core of this transformation is deep (DL), a subset ML that robust in processing analyzing complex large datasets. This paper provides comprehensive overview key models, including Convolutional Neural Networks (CNNs), Long Short-Term Memory networks (LSTMs), Deep Belief (DBNs), Transformers, Generative Adversarial (GANs), Reinforcement Learning (Deep RL). Beyond summarizing their mathematical foundations processes, study offers new insights into how these models are applied real-world contexts, highlighting specific advantages limitations tasks algorithmic trading, risk management, portfolio optimization. It also examines recent advances emerging trends alongside critical challenges data quality, model interpretability, computational complexity. These can guide future research directions toward developing more efficient, robust, explainable address evolving needs sector.

Language: Английский

---

Methods of Data Preparation for Multilingual Sentiment Analysis Using Neural Networks

Studies in big data, Journal Year: 2025, Volume and Issue: unknown, P. 437 - 444

Published: Jan. 1, 2025

Language: Английский

---

Artificial Intelligence Based Hybrid Models for Prediction of Stock Prices

Published: March 3, 2023

Stock price forecasting has recently become an important practical component of the economic arena. An intriguing task, stock is regarded to be related volatility and noise market activity. To address these issues accurately predict prices, this paper proposes a hybrid framework based on learning model such as stacked Long Short Term Memory (LSTM) Convolutional network. Experiments with several possible outcomes are run assess proposed using data set. The was trained ADANI from last roughly fourteen years LSTM network evaluated assessment criteria Root Mean Square Error (RMSE). proven competitive against other models in prediction various scenarios.

Language: Английский

---

A Time Series Analysis-Based Stock Price Prediction Framework Using Artificial Intelligence

Communications in computer and information science, Journal Year: 2023, Volume and Issue: unknown, P. 280 - 289

Published: Dec. 2, 2023

Language: Английский

---

Deep Learning in Finance: A Survey of Applications and Techniques

Published: Aug. 20, 2024

Machine learning (ML) has transformed the financial industry by enabling advanced applications such as credit scoring, fraud detection, and market forecasting. At core of this transformation is deep (DL), a subset ML that robust at processing analyzing complex large datasets. This paper provides concise overview key models, including Convolutional Neural Networks (CNNs), Long Short-Term Memory networks (LSTMs), Deep Belief (DBNs), Transformers, Generative Adversarial (GANs), Reinforcement Learning (Deep RL). The study examines their processes, mathematical foundations, practical in finance. It also explores recent advances emerging trends alongside critical challenges data quality, model interpretability, computational complexity, offering insights into future research directions can guide development more explainable models.

Language: Английский

---

Machine Learning Models-Based Forecasting Moroccan Stock Market

Lecture notes in networks and systems, Journal Year: 2024, Volume and Issue: unknown, P. 56 - 66

Published: Jan. 1, 2024

Language: Английский

---

A Deep Learning Approach with Extensive Sentiment Analysis for Quantitative Investment

Electronics, Journal Year: 2023, Volume and Issue: 12(18), P. 3960 - 3960

Published: Sept. 20, 2023

Recently, deep-learning-based quantitative investment is playing an increasingly important role in the field of finance. However, due to complexity stock market, establishing effective methods facing challenges from various aspects because market. Existing research has inadequately utilized news information, overlooking significant details within content. By constructing a deep hybrid model for comprehensive analysis historical trading data and complemented by momentum strategies, this paper introduces novel approach. For first time, we fully consider two dimensions news, including headlines contents, further explore their combined impact on modeling price. Our approach initially employs fundamental screen valuable stocks. Subsequently, built technical factors based data. We then integrated content summarized through language models extract semantic information representations. Lastly, constructed neural capture global features combining with representations, enabling prediction decisions. Empirical results conducted over 4000 stocks Chinese market demonstrated that incorporating enriched enhanced objectivity sentiment analysis. proposed method achieved annualized return rate 32.06% maximum drawdown 5.14%. It significantly outperformed CSI 300 index, indicating its applicability guiding investors making more strategies realizing considerable returns.

Language: Английский

---