Abstract:
People post frequent updates on the social media platforms regarding their
activities, likes and dislikes that adds the element of uniqueness and per-
sonalization to the content. This study explores how data-driven methods
can leverage the information available on social media platforms to predict
Postpartum Depression (PPD). Early screening of mental disorders plays a
crucial role in diagnosis and treatment. A generalized approach is proposed
where linguistic features are extracted from user generated textual posts and
categorized as general, depressive and PPD representative using multiple ma-
chine learning techniques. We use Linguistic Inquiry Word Count (LIWC) to
extract a standard set of features and combine it with an additional feature
based on Absolutist dictionary before identifying a list of most important
features for the task of prediction. We nd that the techniques used in our
study exhibit strong predictive capabilities for PPD content. Multi-Layer
Perceptron outperformed other techniques like SVM & Logistic Regression
with 91.7% accuracy for depressive content identi cation and up to 86.9%
accuracy for PPD content prediction. Our proposed methodology will help
the government and humanitarian organizations to improve the systems and
utilize available professional resources e ciently in order to deal with the
situation of increasing occurrence of mental disorders.
