Year of Graduation
Text Features in Forecasting Stock Market Prices Volatility
The paper proposes an improvement for GARCH-type volatility models using modern natural language processing techniques. The method improves volatility forecasts for Magnit company stocks, because it incorporates information extracted from economic news, which is important for decision making in the stock market. In addition to the model improvement, performance estimates for various methods of representing news texts such as human pre-processed news, named entities and subject-verb-object triples are made.