Microstructure-Empowered Stock Factor Extraction and Utilization

Xianfeng Jiao jiaoxianfeng@stu.pku.edu.cn Peking UniversityChina , Zizhong Li zzoli@ucdavis.edu University of California, DavisUnited States , Chang Xu chanx@microsoft.com MicrosoftChina , Yang Liu yangliu2@microsoft.com MicrosoftChina , Weiqing Liu weiqing.liu@microsoft.com Microsoft ResearchChina and Jiang Bian jiang.bian@microsoft.com Microsoft ResearchChina

Abstract.

the relevant factors extracted from these signals are utilized in downstream tasks. In empirical studies, we verify our method on an entire year of stock order flow data across diverse scenarios, thus providing a wider range of potential applications in comparison to existing tick-level approaches that have only been evaluated on small datasets spanning a few days of stock data. We demonstrate that our method extracts superior factors from order flow data, enabling significant improvement for stock trend prediction and order execution tasks at the second and minute level.