2009-01-14 Bled meeting at lunchtime
Participants
- Cefriel: Emanuele and Irene
- Siemens: Volker and Yi
- external "auditors": Armando from Cefriel and Josè from Max Plank
Discussion
final goal: get the traffic data from Milano Municipality
- desired goal: get the most part of the available data , potentially all of them
- short objective: prepare a data request for Milano municipality to explain what data we need and why, so that they can be interested in the results of our experimentations
- dimensions of data: time, number of sensors, geo-positioning
priority should be the time: the longest the time interval the better --> 2 years should be very nice even with a subset of sensors (that however are stable over time)
- how to select a subset of sensors? The more sensors the better, if we have to select a subset of them, it's better to select an area in which sensors are "correlated" (i.e. a meaningful area in which cars probably move from one point of that area to another point of the same area)
- minimal request: a meaningful area of Milano with sensors stable over time with 2 years of measurement
- problem of positioning of sensors: we definitely need the mapping of the sensors position to the map data we already have (with "links" and "nodes")
- if AMA has this information, fine
- otherwise, we have to "calculate" this mapping to make sense of the data; in case Cefriel will take this task, but it will not be so easy
- other additional data would be very useful, but to start experimenting there is no need for them, so if we manage to get them at some point, we can add them later on
- what can be useful and how difficult is to get data:
- recurring events: those impact quite a lot on the data analysis; so far, we got info about markets, we can try to search for more data of this kind, for example football matches at San Siro stadium
- holidays: relevance as before; maybe easier to find them for the past (one option is to take the closing days of Cefriel for the last years)
- weather information: this info can be very useful for the prediction as well; for the "future", it can be easy to keep track of weather information (we can just record the result of live services); on the contrary, we don't know if it is possible of easy to get historical data about weather but we will investigate it
- car accidents: they can be quite relevant; however, from what we understood from AMA, it is very unlikely that we will ever get them
- it would be very nice if AMA gives us some requirements from their point of view: what kind of traffic prediction do they envision? What could be the further exploitation of our experimentations?
ACTION ITEMS
- Emanuele to integrate Siemens document and circulate it
- Irene to circulate info about original GIS format of map data and possible free tool to open and visualize it