What is it about?

We present temporal generative filter that implement a hybrid attention-based LSTM architecture to capture the stocks’ individual behavioural patterns. These patterns are then fed to hypergraph convolution layers to obtain spatio-temporal embeddings that are optimized with respect to the potential of the stocks for short-term profit. We propose a mechanism that combines the temporal patterns of stocks with spatial convolutions through hypergraph attention, thereby integrating the internal dynamics and the multi-order dynamics. Our convolution process uses the wavelet basis, which is efficient and also effective in terms of maintaining the localization.

Featured Image

Why is it important?

Advances in deep neural network (DNN) architectures have enabled new prediction techniques for stock market data. Unlike other multivariate time-series data, stock markets show two unique characteristics: (i) multi-order dynamics, as stock prices are affected by strong non-pairwise correlations (e.g., within the same industry); and (ii) internal dynamics, as each individual stock shows some particular behaviour. Recent DNN-based methods capture multi-order dynamics using hypergraphs, but rely on the Fourier basis in the convolution, which is both inefficient and ineffective. In addition, they largely ignore internal dynamics by adopting the same model for each stock, which implies a severe information loss.

Perspectives

We propose ESTIMATE, a stock recommendation framework that supports learning of the multi-order correlation of the stocks (i) and their individual temporal patterns (ii), which are then encoded in node embeddings derived from hypergraph representations. Extensive experiments on real-world data illustrate the effectiveness of our techniques and highlight its applicability in trading recommendation. In future work, we plan to tackle this issue by exploring time-evolving hypergraphs with the ability to memorize distinct periods of past data and by incorporating external data sources such as earning calls, fundamental indicators, news data [29, 35, 36], social networks [31, 32, 45, 46], and crowd signals [15, 16, 28].

Thanh Tam Nguyen
Griffith University

Read the Original

This page is a summary of: Efficient Integration of Multi-Order Dynamics and Internal Dynamics in Stock Movement Prediction, February 2023, ACM (Association for Computing Machinery),
DOI: 10.1145/3539597.3570427.
You can read the full text:

Read

Resources

Contributors

The following have contributed to this page