Detecting possible cases of depression

Inspiration

During the COVID pandemic, mental health cases have spiked. The spike can be quantified by the comparison of 37% of the US population showing signs of mental health concerns before the lockdown. However, during the pandemic, this has risen to 49%. The number of suicides per day is increasing during the pandemic by threefold (WHO).

What it does

Our solution aims at detecting cases of depression or other mental health concerns through posts on social media made by users. Our Machine learning algorithm (Convolutional neural network) can detect cases of depression with upto 98.87% accuracy. Key point is, we are not diagnosing, but only detecting a possible case. Further, if a tweet has been detected to show signs of depression, an automated message will be sent to the user stating "Hello, we hope you are doing well. If you need help with regards to mental health concerns, please contact Smaritans at 116 123. We work 24/7 and are happy to listen."

How I built it

We built this using Machine Learning frameworks such as tensorflow. We coded this is Python. We built our automated messaging system using the Twilio framework in JavaScript.

Impact

Increased detection of possible cases indicating depression or other mental health concerns. Preventing lives by timely detection of possible cases of mental health. Providing right information at the right time.

Challenges I ran into

It was rather difficult for us to code in JavaScript but we managed to do it using various sources of help online.

Accomplishments that I'm proud of

Creating a full fledged Machine Learning model with an automated messaging system!

What's next for Detecting possible cases of depression

We realise that our model has been trained on a very biased dataset. So, we would like to gather a better dataset by scraping live data on broader keywords.