An efficient loss function and deep learning approach for ranking stock returns in the absence of prior knowledge

Article Computer Science, Artificial Intelligence

Deep LOB trading: Half a second please!

Jie Yin et al.

Summary: This article introduces an expert deep-learning system for limit order book trading and validates it with real data from the Chinese A-share market.

EXPERT SYSTEMS WITH APPLICATIONS (2023)

Add to Collection

Article Computer Science, Artificial Intelligence

Enhancing stock movement prediction with market index and curriculum learning

Jiahao Yang et al.

Summary: Stock movement prediction is a complex task due to the sophisticated and noisy nature of the stock market system. In this study, we propose a method that enhances stock movement prediction through the use of market index and curriculum learning.

EXPERT SYSTEMS WITH APPLICATIONS (2023)

Add to Collection

Article Computer Science, Information Systems

Forecasting movements of stock time series based on hidden state guided deep learning approach

Junji Jiang et al.

Summary: Stock movement forecasting is often treated as a sequence prediction task using time series data. While deep learning models have been increasingly employed for fitting dynamic stock time series, few of them have focused on understanding the internal dynamics of the market system. To address this, the proposed HMM-ALSTM framework integrates the Hidden Markov Model (HMM) into the deep learning process, allowing for the discovery of hidden states and patterns that contribute to the stock time series data.

INFORMATION PROCESSING & MANAGEMENT (2023)

Add to Collection

Article Computer Science, Information Systems

Self-attention based deep direct recurrent reinforcement learning with hybrid loss for trading signal generation

Dongkyu Kwak et al.

Summary: This paper proposes a novel algorithmic trading model based on recurrent reinforcement learning, which maximizes the expected rewards by extracting temporal features from complex observations. The model incorporates hybrid learning loss and self-attention mechanism to enhance the decision-making process. The proposed SA-DDR-HL model demonstrates superior performance compared to baseline benchmark models.

INFORMATION SCIENCES (2023)

Add to Collection

Article Computer Science, Information Systems

Residual long short-term memory network with multi-source and multi-frequency information fusion: An application to China?s stock market

Songsong Li et al.

Summary: This paper introduces a feature fusion residual LSTM (FFRL) model to address the limitations of LSTM and handle multi-source and multi-frequency information. FFRL consists of three modules, namely the feature selection module, feature extraction module, and residual module, to improve the limitations of LSTM. Significant performance improvements of FFRL over comparison models, ablation networks, and visualization methods are demonstrated on a variety of Chinese stocks.

INFORMATION SCIENCES (2023)

Add to Collection

Article Mathematics

Stock Market Analysis Using Time Series Relational Models for Stock Price Prediction

Cheng Zhao et al.

Summary: The ability to predict stock prices is essential for investment decisions, but the complexity of factors influencing stock prices has been extensively studied. Traditional methods that focus on time-series information for a single stock lack a holistic perspective. A time series relational model (TSRM) is proposed in this paper to integrate time and relationship information. The TSRM utilizes transaction data, K-means model, LSTM, and GCN to predict stock prices, yielding significant improvements in cumulative returns and maximum drawdown in the Chinese stock markets.

MATHEMATICS (2023)

Add to Collection

Article Automation & Control Systems

A representation-learning-based approach to predict stock price trend via dynamic spatiotemporal feature embedding

Bowen Pang et al.

Summary: Stock price trend prediction is a challenging research topic. Recently, GNN-based models have been proposed to improve prediction accuracy by considering information about stocks themselves and the relationships between stocks. However, the static and unstructured nature of graph data limits their ability to capture dynamic relationships. In this study, we propose a novel model called PE-Net, which effectively utilizes temporal and cross-sectional information in stock prices to predict trends. Experimental results on real-world S&P 500 constituents show that PE-Net outperforms state-of-the-art models in terms of accuracy and AUC.

ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE (2023)

Add to Collection

Article Computer Science, Information Systems

A stock rank prediction method combining industry attributes and price data of stocks

Huajin Liu et al.

INFORMATION PROCESSING & MANAGEMENT (2023)

Add to Collection

Article Computer Science, Information Systems

Stock ranking prediction using a graph aggregation network based on stock price and stock relationship information

Guowei Song et al.

Summary: The volatility of stock prices makes it difficult to predict stock price trends correctly. Graph neural networks are used to learn how stocks interact and provide more information. However, current methods struggle to capture the dynamic interplay between stocks over time and the complex relationships that affect stock prices.

INFORMATION SCIENCES (2023)

Add to Collection

Article Computer Science, Artificial Intelligence

China?s commercial bank stock price prediction using a novel K-means-LSTM hybrid approach

Yufeng Chen et al.

Summary: This study proposes a novel hybrid deep learning approach that utilizes an improved clustering algorithm and LSTM neural network model to predict stock prices more accurately. Experimental results demonstrate that this method outperforms traditional statistical models in terms of prediction ability and accuracy.

EXPERT SYSTEMS WITH APPLICATIONS (2022)

Add to Collection

Article Business, Finance

Detecting the lead-lag effect in stock markets: definition, patterns, and investment strategies

Yongli Li et al.

Summary: This study validates the power-law distribution in stock trading activity and identifies the lead-lag effect based on 10 years of trading data. It designs investment strategies utilizing this effect, which significantly improve the performance of basic alpha-factor strategies.

FINANCIAL INNOVATION (2022)

Add to Collection

Article Computer Science, Artificial Intelligence

Prediction of stock price direction using a hybrid GA-XGBoost algorithm with a three-stage feature engineering process

Kyung Keun Yun et al.

Summary: This study introduces a hybrid GA-XGBoost prediction system with enhanced feature engineering process, demonstrating the importance of feature engineering in stock price prediction. By comparing obtained feature sets to the original dataset and improving prediction performance, the study empirically proves that successful prediction largely depends on a deliberate combination of feature engineering processes.

EXPERT SYSTEMS WITH APPLICATIONS (2021)

Add to Collection

Article Computer Science, Artificial Intelligence

A Multimodal Event-Driven LSTM Model for Stock Prediction Using Online News

Qing Li et al.

Summary: In financial markets, it is believed that market information affects stock movements, creating a multimodal challenge; handling this challenge involves addressing data mode interactions and sampling time heterogeneity; previous research assumed news affects specific stocks, but in reality impacts related stocks.

IEEE TRANSACTIONS ON KNOWLEDGE AND DATA ENGINEERING (2021)

Add to Collection

Article Computer Science, Information Systems

A novel graph convolutional feature based convolutional neural network for stock trend prediction

Wei Chen et al.

Summary: The paper introduces a novel method for stock trend prediction using GC-CNN model, which considers both stock market information and individual stock information. Experimental analysis demonstrates that the proposed method outperforms several stock trend prediction methods and stock trading strategies.

INFORMATION SCIENCES (2021)

Add to Collection

Article Economics

Temporal Fusion Transformers for interpretable multi-horizon time series forecasting

Bryan Lim et al.

Summary: This paper introduces the Temporal Fusion Transformer (TFT), a novel attention-based architecture that combines high-performance multi-horizon forecasting with interpretable insights into temporal dynamics. TFT utilizes recurrent layers for local processing and interpretable self-attention layers for long-term dependencies, achieving high performance in a wide range of scenarios. By selecting relevant features and suppressing unnecessary components, TFT demonstrates significant performance improvements over existing benchmarks on various real-world datasets.

INTERNATIONAL JOURNAL OF FORECASTING (2021)

Add to Collection

Proceedings Paper Computer Science, Artificial Intelligence

Accurate Multivariate Stock Movement Prediction via Data-Axis Transformer with Multi-Level Contexts

Jaemin Yoo et al.

Summary: This study introduces a new approach named DTML (Data-axis Transformer with Multi-Level contexts) for learning correlations between multiple stocks for stock movement prediction. By learning temporal correlations within each stock, generating multi-level contexts based on a global market context, and utilizing a transformer encoder for learning inter-stock correlations, DTML achieves state-of-the-art accuracy and higher profits compared to competitors on datasets collected from US, China, Japan, and UK stock markets.

KDD '21: PROCEEDINGS OF THE 27TH ACM SIGKDD CONFERENCE ON KNOWLEDGE DISCOVERY & DATA MINING (2021)

Add to Collection

Proceedings Paper Computer Science, Artificial Intelligence

Learning Multiple Stock Trading Patterns with Temporal Routing Adaptor and Optimal Transport

Hengxu Lin et al.

Summary: The paper introduces a novel architecture called Temporal Routing Adaptor (TRA) to empower existing stock prediction models with the ability to model multiple stock trading patterns. By using a learning algorithm based on Optimal Transport, the router can be effectively optimized to improve the accuracy of stock prediction.

KDD '21: PROCEEDINGS OF THE 27TH ACM SIGKDD CONFERENCE ON KNOWLEDGE DISCOVERY & DATA MINING (2021)

Add to Collection

Article Computer Science, Artificial Intelligence