LSTM based Model for Real-time Stock Market Prediction on Unexpected Incidents

Sudden and unlooked events can be adverse to the world financial community, especially stock exchange markets. Events like the 9/11 terrorist attacks can be destructive for global and local markets. Investors need to assess the sudden events' effects on a minute-by-minute basis to evaluate market reactions precisely for a better investment strategy. Despite the web crawler's availability, it is challenging to mine biased data; however, we build an artificial intelligence (AI) based model to make a better investment strategy that uses unexpected incidents like terrorist attacks and other natural disasters like floods and earthquakes. For this, we used Pakistan's real-time terrorist attacks, floods, and earthquakes data and Pakistan stock market KSE100 index data to feed the model which predicts swift market changes. We study unexpected real-world incident data and stock market data from 2001 to 2020. First of all, a web crawler crawls the data from different websites using AI techniques. Then this data is fed into the information extractor module, which extracts the meaningful information based on AI. This unexpected incident data and stock market data are then used to train the AI-based LSTM to predict the market direction. Results prove that our model outperforms in measuring the influence of these shocking incidents and consequently detecting abrupt alterations in stock market directions.