Unsupervised Stock Clustering Based on Adversarial Learning and Its Application in Quantitative Investment

Quantitative investment is one of the primary trading methods in stock markets. Quantitative models based on artificial intelligence technologies have overcome the limitations and drawbacks of traditional quantitative models, providing richer sources of revenue. Supervised learning-based quantitative models rely on various forms of stock labels, which can generate good returns but cannot reveal common features among groups of stocks with potentially high-yield trends. This paper proposes unsupervised learning on stock datasets using clustering to explore potential features within high-dimensional stock time-series datasets. This approach aims to obtain useful information from limited data and optimize existing trading. Finally, we used backtesting results to validate the effectiveness of the model in the real stock market.