Forecasting the Natural Gas Price Trend - Evaluation of a Sentiment Analysis

Abstract

Nowadays, text messages are of interest, because they quickly convey information about events. For this reason, we analyze, whether a prevailing sentiment in a text-based financial news has impact on the price trend of natural gas at an energy exchange. This prediction method supports utility companies, because it allows faster trading decisions on the natural gas market and thus reduce associated business risks. It is also transferable into other business domains. We initially applied text mining methods to gain first results and moved over to sentiment analysis (SAN) to be able to evaluate their capability to support trading decisions. The calculated performance metrics of SAN made obvious that the consideration of the sentiment in the text is suitable for identifying no price influences, but is weak for identifying the impact of text news on the price trend itself. This results demands further research on applying different approaches on text analysis.