Classification of Traffic Events Notified in Social Networks' Texts

Abstract

It is interesting to exploit the user-generated content (UGC) and to use it with a view to infer new data; volunteered geographic information (VGI) is a concept derived from UGC, whose main importance lies in its continuously updated data. The present approach tries to explode the use of VGI by collecting data from a social network and a RSS service; the short texts collected from the social network are written in Spanish language; text mining and a recovery information processes are applied over the data in order to remove special characters on text and to extract relevant information about the traffic events on the study area; then data are geocoded. The texts are classified by using a machine learning algorithm into five classes, each of them represents a specific traffic event or situation.

Geospatial Modeling of Road Traffic Using a Semi-Supervised Regression Algorithm

“…Before to process the sensed data, it is necessary to identify their features. The corpus used in the proposed method belongs to [28], which consists of a set of tweets whose main content are descriptions of the traffic situation that citizens of Mexico City have posted when trying to prevent other people from being stopped at a traffic jam. It is important to mention that all tweets that composed of the corpus were written in Spanish.…”