One Step Ahead: A Framework for Detecting Unexpected Incidents and Predicting the Stock Markets

Unexpected incidents can be destructive or even disastrous, affecting the financial markets. Incidents such as 9/11 attacks (2001), Fukushima nuclear disaster (2011), and COVID-19 outbreaks (2019, 2020) severely shocked both local global For investors, it is crucial to quantify key facts that affect incidents' impacts, estimate reactions of markets accurately efficiently for event-driven investment strategies. Though Web data other alternative allow a possibility, still very challenging mine noisy often biased heterogeneous sources, construct unified framework modeling across time regions. As first attempt, we build extracts incident globally based on deep neural network, feeds them into models built event database complemented with novel socioeconomic datasets (e.g. nightlight from satellites), predicts stock market directions in simulated real-world setting interpretable results outperform various baselines. Specifically, study terrorist three countries over 20 years average, effort systematically impact at large scale using indicators.