| Popis: |
The problem of forecasting protests on digital footprints of social media users is still unsolved, and there are no scientific articles on the issue. This paper is devoted to estimating the possibility of such forecasting and to pointing out potential difficulties and limitations associated with this study. We use a comparative analysis of this problem and that of election outcome forecasting, which is well studied. We used six criteria to compare the cases: the presence of digital footprints in social media, the representativeness of digital footprint authors, openness of public discussion on topics, predictability, mathematical statement of the problem, and data availability. The last criterion separates two problems dramatically, since elections are 'regular events', but protests are not, so there is no obvious way to collect or store data for learning or testing protest prediction models. |
