Event-driven data mining methods for large-scale market prediction: a case study of an agricultural products company

Abstract

Stock market is often affected by events, especially emergencies, such as natural disasters. Stock price prediction is significant to traders in this market as the references for the future to better invest and for market supervision. In this paper, the forecasting model combing topic models with data mining tools, namely event-driven prediction, is aimed to seek for more accurate predicting price results through extracting topics from news articles related to the stock as well as the historical price data. Our experiment is carried out in an famous agricultural products company in China and the empirical results show that the proper information extracted from news in popular portal website in previous day can be beneficial for the current price prediction.