DoubleAdapt: A Meta-learning Approach to Incremental Learning for Stock Trend Forecasting

Lifan Zhao,Shuming Kong,Yanyan Shen

2024-04-07

Abstract:Stock trend forecasting is a fundamental task of quantitative investment where precise predictions of price trends are indispensable. As an online service, stock data continuously arrive over time. It is practical and efficient to incrementally update the forecast model with the latest data which may reveal some new patterns recurring in the future stock market. However, incremental learning for stock trend forecasting still remains under-explored due to the challenge of distribution shifts (a.k.a. concept drifts). With the stock market dynamically evolving, the distribution of future data can slightly or significantly differ from incremental data, hindering the effectiveness of incremental updates. To address this challenge, we propose DoubleAdapt, an end-to-end framework with two adapters, which can effectively adapt the data and the model to mitigate the effects of distribution shifts. Our key insight is to automatically learn how to adapt stock data into a locally stationary distribution in favor of profitable updates. Complemented by data adaptation, we can confidently adapt the model parameters under mitigated distribution shifts. We cast each incremental learning task as a meta-learning task and automatically optimize the adapters for desirable data adaptation and parameter initialization. Experiments on real-world stock datasets demonstrate that DoubleAdapt achieves state-of-the-art predictive performance and shows considerable efficiency.

Statistical Finance,Artificial Intelligence,Computational Engineering, Finance, and Science,Machine Learning,Computational Finance

What problem does this paper attempt to address?

The paper primarily addresses the issue of incremental learning in stock trend prediction and proposes a new method called DoubleAdapt. Specifically, the study aims to solve the following key problems:

1. **Distribution Shift Challenge in Incremental Learning**: In stock market prediction, due to the dynamic nature of the market, the distribution between new data and old data may change (i.e., "concept drift" or "distribution shift"), making it difficult for the model to effectively utilize the new data for updates.

2. **Enhancing Model Adaptability and Robustness**: The paper proposes a DoubleAdapt framework to address the distribution shift problem in incremental learning through two approaches: data adaptation and model adaptation. Specifically:
- **Data Adaptation**: The goal is to reduce the distribution difference between incremental data and test data, allowing the model to train in a distribution closer to the actual test data.
- **Model Adaptation**: Focuses on finding a good initialization point for the model parameters, enabling the model to quickly adapt to new information in the incremental data while maintaining robustness to potential distribution shifts.

3. **Achieving Efficient and Accurate Predictions**: By using meta-learning techniques, the proposed method can achieve efficient model updates when handling real-world datasets and demonstrates higher predictive performance compared to traditional Rolling Retraining (RR) methods.

In summary, the core objective of this study is to develop an incremental learning method that can effectively address distribution shifts in non-stationary environments, thereby improving the accuracy and efficiency of stock trend prediction. DoubleAdapt overcomes the challenges faced by traditional incremental learning methods through adaptive mechanisms at both the data and model levels, providing a new solution for stock trend prediction.