Trade the Event: Corporate Events Detection for News-based Event-driven Trading

Zhihan Zhou, Liqian Ma, Han Liu

Abstract: We introduce an event-driven trading strategy that detects corporate events from news articles and trades the related securities accordingly. Existing news-based stock prediction models usually suffer from poor applicability and low signal-to-noise ratios since they either ignore the timeliness of news or directly use news sentiments and other features (e.g., bag-of-words) to predict the stock movements. In contrast, we consider corporate events as the driving force behind stock movements and aim to profit from the temporary stock mispricing that may occur when corporate events take place. The core of the proposed strategy is a bi-level event detection model. The low-level event detector identifies events' existences from each token, while the high-level event detector incorporates the entire article's representation and the low-level detected results to make the final decisions. We also develop a dataset EDT for corporate event detection and news-based intraday/interday trading benchmark. EDT includes 9406 news articles with token-level event labels and 107930 news articles with minute-level timestamps. Experiments on EDT and historical stock price data indicate that the proposed strategy significantly outperforms all the baselines in winning rate, excess returns over the market, and the average return on each transaction.