Analysis of data obtained from the social network Twitter for the early identification of users’ suicidal tendencies

Abstract

Although not everyone is aware of it, data available on the Internet are very useful and have a great potential to help our society. The digital platform Twitter is a social network where people sometimes express their feelings and emotions. And this paper arises from the idea of doing an analysis of these data through a Machine Learning tool, to find a psychiatric picture of depression, and if it is possible, the associated suicidal tendency. Twitter data extraction tool has been Tweepy, and with the profile data users, it has been made, an excel database that collects the information. Next, with the Machine Learning tool called UMAP, an unsupervised analysis of the database has been carried out, thanks to which it has been possible to differentiate three groups, with a very low inter cluster distance, which suggest that each observation looks a lot like its neighbors. From these three groups, we find one which behavior or use of the platform would be associated with a normal or standard way. The two other two group of meet part of the characteristics associated with depression.