News clustering based on semantic analysis can yield information on how actively different topics are discussed and if any kind of judgment takes place. Thus a valid representation of the “world views” can be achieved. Thin in turn fosters situation awareness among the public that is proving to be critical for crisis management such as Covid-19,
On top of standard industry metrics like on-chain activities and network hash rates, one can assess the network usage (via the gas price), statistics related to the projects built on top of it (number of dApps, their on-chain transactions), smart contracts executed and more. In addition, there are several twitter accounts with large followership that might affect the public sentiment. In order to examine the predictive power of these data points and arrive at a price movement signalling tool, 3 models need to be built as it follows:
- A web scraper extracting relevant information from various sources and updating the quantitative models in real time
- A sentiment model assessing the historical predictive power of a number of twitter accounts
- A multi-factor regression model, predicting the ETH/USD price action in the next 5 to 15 minutes
Multivariate regression model taking into account quantitative and qualitative data in real-time to signal potential price movements and expected volatility over a period of 15 minutes into the future.