Overview

In the reddit-cnn-2.py file, we put the comments of Reddit users into a dataframe and "cleaned" the comments for our use. This dataframe contains the comment text and the depression status of the commenter which was gathered based off of the DSM-5 criteria from our survey. With this data, we use Tensorflow software and Natural Language Processing techniques to train a Convulational Neural Network to detect depression. This model is stored in the reddit-cnn-2.zip file.

DD-1

Reddit Models

Transformer v1.0

-31.47%

Accuracy: 61.64%
Loss: 0.3713

CNN v2.0 (Best Model)

-0.65%

Accuracy: 92.46%
Loss: 0.2267

CNN v2.0 (Best Model)

+0.00%

Accuracy: 93.11%
Loss: 0.2510

Emotion Models

Sadness CNN v1.0

+0.00%

Accuracy: 95.50% Loss: 0.1547