News-driven financial warning based on label information attention

Zhilin Ren, Long Chen, Xiaohua Huang, Wenjing Wang, Cai Xu, Wei Zhao, Ziyu Guan

Abstract: Most existing news-driven stock market prediction methods ignore the potential relationship between financial news and stocks. The complex relationship can help us to improve the accuracy of algorithmic trading systems. Therefore, we propose a deep learning method for financial warning by fusing Stock Label Information (SLI). We extract events from news texts and fuse stock information together as feature vectors, using neural networks to model the underlying relationship between news and stocks. Experimental results show that our method outperforms other baseline methods in experiments on TPX500 and TPX100 datasets. CCS CONCEPTS • Computing methodologies • Artificial intelligence • Natural language processing