Building up on the concept of the “Wisdom of Crowds”, the usage of social media data for stock market prediction is not a new idea. However, the ongoing improvement of sentiment analysis tools opens up new possibilities for the deeper utilization of such data.
We, a team of students at the Ludwig Maximilian University of Munich, used the Semantria Excel Add-In for a research project, which aimed to prove that sentiment analysis has the potential to become an integral component of future prediction systems.
The study focused on seven US companies from the area of Hotels, Resorts and Cruise Lines in the timeframe of Q1, 2014. In a multi-step analysis it was first shown, that stock market indicators are highly correlated with tweet sentiment metrics and also with the number of daily $Cashtag tweets. Building up on these findings, a learning prediction model was developed.
The exemplary application of the derived stock trading strategy showed that this model leads to significantly higher returns on investment, than a model that is based on stock market indicators only.
Of course, these results are not generalizable per se and should be rather seen as a proposition for further research. Also we are aware that the characteristic language style of stock market related text differs from natural language and might pose a challenge to sentiment analysis.
Nevertheless we evaluate the use of sentiment analysis in prediction models, especially stock market prediction, to be promising and worth of further attention, for both, researchers and practitioners.
We experienced Semantria to be fast to implement and set-up, its functionalities are easy to learn and well documented. The steps of analysis are transparent and can be reproduced by the user.
Even for non-technical experts like us, the control was intuitive and self-explanatory. We would like to thank the Semantria Team for their support and recommend the software for related research.