Abstract:The price movement prediction of stock market has been a classical yet challenging problem, with the attention of both economists and computer scientists. In recent years, graph neural network has significantly improved the prediction performance by employing deep learning on company relations. However, existing relation graphs are usually constructed by handcraft human labeling or nature language processing, which are suffering from heavy resource requirement and low accuracy. Besides, they cannot effectively response to the dynamic changes in relation graphs. Therefore, in this paper, we propose a temporal and heterogeneous graph neural network-based (THGNN) approach to learn the dynamic relations among price movements in financial time series. In particular, we first generate the company relation graph for each trading day according to their historic price. Then we leverage a transformer encoder to encode the price movement information into temporal representations. Afterward, we propose a heterogeneous graph attention network to jointly optimize the embeddings of the financial time series data by transformer encoder and infer the probability of target movements. Finally, we conduct extensive experiments on the stock market in the United States and China. The results demonstrate the effectiveness and superior performance of our proposed methods compared with state-of-the-art baselines. Moreover, we also deploy the proposed THGNN in a real-world quantitative algorithm trading system, the accumulated portfolio return obtained by our method significantly outperforms other baselines.

What problem does this paper attempt to address?

The paper primarily addresses the problem of stock price movement prediction in financial time series forecasting, specifically how to leverage Graph Neural Network (GNN) technology to improve prediction accuracy. Specifically, the paper proposes a new method—Temporal and Heterogeneous Graph Neural Network (THGNN)—to tackle the following challenges: 1. **Dynamic Relationship Modeling**: Existing methods typically construct relationship graphs that are static, generated through manual annotation or natural language processing techniques, which perform poorly when relationships change frequently. 2. **Heterogeneous Relationship Handling**: Relationships between companies are diverse, such as suppliers and customers, shareholders and invested companies, etc. These types of relationships are not fully utilized in existing prediction models. 3. **Resource Demand and Accuracy Issues**: Relationship graphs constructed through manual annotation or natural language processing techniques often require a large amount of human resources and may have low accuracy. To address the above issues, the main contributions of the paper include: - **Proposing a Graph Learning Framework**: This framework can effectively model the internal relationships between entities, is suitable for financial time series forecasting, and can adapt to market dynamics, aligning with actual price movement trends. - **Designing a Temporal and Heterogeneous Graph Neural Network Model**: Through a two-stage attention mechanism to learn dynamic relationships, this model can efficiently and effectively learn jointly from historical price sequences and internal relationships. - **Empirical Study**: Extensive experimental validation was conducted on real-world datasets, showing that the proposed THGNN method significantly outperforms current state-of-the-art baseline methods. Additionally, the method was tested on actual trading platforms, demonstrating its effectiveness in industrial-grade systems. In summary, this paper aims to improve financial time series forecasting tasks by introducing a novel Temporal and Heterogeneous Graph Neural Network method, achieving significant results particularly in stock price movement prediction.

Temporal and Heterogeneous Graph Neural Network for Financial Time Series Prediction

Price graphs: Utilizing the structural information of financial time series for stock prediction

MDGNN: Multi-Relational Dynamic Graph Neural Network for Comprehensive and Dynamic Stock Investment Prediction

Integrated GCN-LSTM stock prices movement prediction based on knowledge-incorporated graphs construction

Temporal-Relational hypergraph tri-Attention networks for stock trend prediction

Dynamic Graph Representation with Contrastive Learning for Financial Market Prediction: Integrating Temporal Evolution and Static Relations

A Dynamic Attributes-driven Graph Attention Network Modeling on Behavioral Finance for Stock Prediction

Financial time series forecasting with multi-modality graph neural network

Futures Quantitative Investment with Heterogeneous Continual Graph Neural Network

Stock Price Prediction Using Temporal Graph Model with Value Chain Data

Corporate Investment Prediction Using a Weighted Temporal Graph Neural Network.

Heterogeneous Dual-Dynamic Attention Network for Modeling Mutual Interplay of Stocks

Conceptual-temporal graph convolutional neural network model for stock price movement prediction and application

Multi-Graph Convolutional Network for Relationship-Driven Stock Movement Prediction

ST-Trader: A Spatial-Temporal Deep Neural Network for Modeling Stock Market Movement

A graph neural network-based stock forecasting method utilizing multi-source heterogeneous data fusion

Dynamic Graph Construction Via Motif Detection for Stock Prediction

Hierarchical Adaptive Temporal-Relational Modeling for Stock Trend Prediction

Quantitative Stock Selection Model Using Graph Learning and a Spatial–Temporal Encoder

Relational Temporal Graph Convolutional Networks for Ranking-Based Stock Prediction.

Multi-Relational Graph Convolution Network for Stock Movement Prediction