Project Details
Description
We plan to model influence within social media networks using methods employed in engineering and network science. In doing so, we can potentially reduce the impact of noise interactions which have typically affected studies testing for causal relationships between social media and external factors. For example, this sophisticated approach to influence detection can provide a new method for weighting online sentiment, collected from Twitter, from which relationships with financial market activity, political events, and healthcare policy may be tested.
In particular, we use financial time series to test the predictive power of our influence-weighted social media sentiment. Although previous studies in academic finance have investigated the extent to which online interactions between investors can affect financial trading activity, the findings of such studies are often accompanied by high error terms and thus no consensus has been reached. Our measure may provide a solution to this issue, while also offering a methodological contribution that is of interest to academics in computer science.
In particular, we use financial time series to test the predictive power of our influence-weighted social media sentiment. Although previous studies in academic finance have investigated the extent to which online interactions between investors can affect financial trading activity, the findings of such studies are often accompanied by high error terms and thus no consensus has been reached. Our measure may provide a solution to this issue, while also offering a methodological contribution that is of interest to academics in computer science.
Layman's description
We aim to analyse social media networks to identify the influence of individuals on financial trading.
Status | Finished |
---|---|
Effective start/end date | 11/06/19 → 31/07/20 |
Fingerprint
Explore the research topics touched on by this project. These labels are generated based on the underlying awards/grants. Together they form a unique fingerprint.