StockCL: Selective Contrastive Learning for Stock Trend Forecasting via Learnable Concepts

Stock trend forecasting is crucial for quantitative investment and various deep learning models have been proposed to obtain superior performance. Due to the limited stock data, existing deep learning models suffer from overfitting. It has been revealed that supervised contrastive learning can provide additional supervision signals by pulling closer samples with the same label and pushing apart samples with different labels. However, stock data has continuous labels and it is nontrivial to identify appropriate contrastive pairs based on label information. In this paper, we develop a novel selective contrastive learning framework named StockCL for stock trend forecasting, which is applicable to any stock trend forecasting models. Our key insight is to identify latent concepts that drive the stock trends and select reliable contrastive pairs according to the samples’ belonging concepts and their label similarity. Experiments on two datasets with four stock trend forecasting models demonstrate that StockCL consistently improves the forecasting performance with a significant margin by 5%–18%. Code is available at .

Database Systems for Advanced Applications: 29th International Conference (DASFAA'24)