This thesis proposes an approach for the early detection of Anorexia Nervosa (AN)
on social media. Our method is based on machine learning techniques using the
processed texts written by social media users. This method relies on a set of features
based on domain-specific vocabulary, topic modelling, psychological processes and
linguistic information extracted from the users’ writings. Moreover, other features
are studied in order to exploit all the dataset resources. Additionally, we have
compared ...
This thesis proposes an approach for the early detection of Anorexia Nervosa (AN)
on social media. Our method is based on machine learning techniques using the
processed texts written by social media users. This method relies on a set of features
based on domain-specific vocabulary, topic modelling, psychological processes and
linguistic information extracted from the users’ writings. Moreover, other features
are studied in order to exploit all the dataset resources. Additionally, we have
compared some of the most known learning algorithms and we have introduced a
minimum amount of information threshold to avoid some false positive predictions.
This approach penalises the delay in the detection of positive cases in order to classify
the users at risk as early as possible. By the early identification of anorexia, along
with an appropriate treatment, the speed of recovery and the likelihood of staying
free of the illness improves. The results of this thesis showed that our proposal is
suitable for the early detection of AN symptoms in social media. Further research
in this topic is needed to solve the problems stated in this project.
+
