The number of depression cases has grown worldwide. The World
Health Organization estimates that 5.8% of the Brazilian population
already present depression symptoms. In the world, 4.8% of
the entire population has presented some symptoms. These data
are alarming because they represent about 12 million people only
in Brazil and 368 million worldwide. Therefore, it is essential to
build applications that adequately identify the population’s feelings
about depression to drive public health policies. Appropriate policies
can save money on public health and keep people active. Thus,
this work investigates how to apply machine learning in classifying
depression posts on Tweeter. The data were extracted from the
social media network, reaching a total of 31.177 tweets classified as
depressive and non-depressive. The application was implemented
in Python with Pandas and SciKit Learning. Results have shown
that SVM overcomes the Naive Bayes algorithm and can reach an
accuracy of 94%, precision of 91%, a recall of 91%, and an F1 Score
of 91%.