Interpretability Of The Perception Of Security Based On Tweets Content

The study of the perception of security (PoS) can improve the understanding of the feeling of citizens about the risk and the magnitude of its consequences, to recognize and identify the fear of crime. Traditionally, the measurement of this type of subjective perception was carried out through surveys with low performance-effective, which means, a small sample of the population in specific periods of time. To solve these problems, a social network such as Twitter has been used to extract and quantify the perception of security. This quantification is done through a supervised learning classification method. To identify which of these methods is the most suitable for quantifying perception, the local interpretability of the results from these models is presented. This work shows a case study in Bogotá city, where is the first time that the social networks are used as a measure of the PoS.