Social networks provide information about activities of humans and social events. Thus, with the help of social networks, we can extract the traffic events that occur in a city. In the context of an urban area, this kind of data allows to obtaining contextual real-time information shared among citizens that will be useful to address social, environmental and economic issues. In this paper, the authors describe a methodology to obtain information related to traffic events such as accidents or congestion, from Twitter messages and RSS services. A text mining process is applied on the messages to acquire the relevant data, then data are classified by using a machine learning algorithm. The events are geocoded and transformed into geometric points to be represented on a map. The final repository lets data to be available for further works related to the traffic events on the study area. As a case of study we consider Mexico City.
TopIntroduction
On Internet we can found data from many sources, that can be analyzed, processed, or shared, and also we can infer information from these data, or try to integrate them (Neves, 2014). Advances in technology let people to be communicated and notify their activities and the places where they are all the time. Many of the social networks as Facebook, Twitter, Instagram, among others, let to the users show their location to others, furthermore some of this applications identify the place where the person is posting a message and show the coordinates of that place (Zhao, 2015).
Social networks are being used not just to share photos and comments, actually they are being used to ask governments for justice or information about their management, to share safety measures when natural catastrophes emerge, to demand for information, to communicate trends, to ask for help when an accident occurs, or to prevent others to do not use certain road (Adamko, 2015); this last usage of social networks is which interests more in this work. Since one purpose of citizens is to prevent others from being at transit problems and road congestions, people is publishing every day at social networks the vial situation that they regard during their trips, with the purpose of let others to move through the less congested road or trying to get help when a car accident occurs. In their publications, users can add a photo and the coordinates of the event they are watching; this elements make easier for others to understood the problem‘s magnitude and to formulate their own solutions, as new paths to move themselves through an specific area, or to look for another transport. People are not aware about how much data they have been producing during the last years (Li, 2015).
The big cities are trying to get an advantage of this kind of data and have been thinking about how to produce information with them, one of the solutions is to classify the messages and choose those that describe an event that modifies the urban dynamics, another is to resubmit the messages that have been published by trusty sources in order to let more people know them (Dou, 2013).
We are proposing the integration of social network and a web RSS service related to traffic events such as car accidents, protests march, traffic jams, among others that make difficult the movement of vehicles and people through the city; in order to apply them a text mining process to eliminate unnecessary words, a geocoding procedure to identify the location were the event has occurred and to classify them using a machine learning tool, with the purpose of create a complete data repository that merges information from heterogeneous data sources that could be used on many posterior researches.
By reading some other investigations and works interested on the traffic situation, we have noticed that most of the times, the predictions about the traffic are produced by using GPS data generated by sensors on taxi cabs and particular cars (Zhang, 2011), this aspect gives to the investigations more precision, but what about the big amount of data produced by people all around the cities? Few works consider it.
The rest of the paper is organized as follows; on section 2 we show some related works that have been recently developed in the topic of volunteered information; on section 3 we depict our methodology for the data integration and the methods that we have used to geocode the data; on section 4 we show some test and results obtained; and finally on section 5 we point out our conclusions.