The main goal of this project was to create an interactive and intuitive visualization tool showing multiple information gathered from geolocated tweets (focus on Switzerland).

The user can switch between three maps: the density map, the sentiments map and the events map.

• The first one shows the distribution of tweets per canton/municipality in Switzerland.
• The second shows the average sentiment associated with the tweets from each canton/municipality.
• The third one shows the distribution of detected events from the tweets.

The user can change the range of dates concerning the displayed data by adjusting the width of the orange bar and/or by horizontally dragging it.

This project was supervised by the lecturer Michele Catasta for the course of Applied Data Analysis (ADA) from the Master in Computer Science program from EPFL (Switzerland). It is an open source project so you can find the git repository by clicking here.

The dataset used consists of a compilation of tweets provided by Swisscom, from 2010 to 2016, geolocated in Switzerland.

The pre-processing steps can be found in the GitHub repository.

The main fields used for the project were:

• id – the integer representation of the unique identifier for the tweet
• createdAt – the UTC time when the tweet was created (timestamp in the format 0000-00-00 00:00:00)
• text – the actual UTF-8 text of the status update
• placeLongitude – the longitude of the place the tweet is associated to
• placeLatitude – the latitude of the place the tweet is associated to

The data used by the map was generated by another group which collaborated with us. They worked on developing a model to detect past events on the Twitter dataset. If you are interested to see their work, here is the link to their github repository.

Thank you again to Daniel Guggenheim, Sabrine Boumelala and Sergii Shynkaruk for their collaboration.