mentBERT

This model is a hybrid fine-tuned version of distilbert-base-uncased on Reddit dataset contains text related to mental health reports of users. it predicts mental health disorders from textual content.

It achieves the following results on the validation set:

• Loss: 0.1873
• F1: 0.6797
• AUC: 0.7942
• Precision: 0.7731

Description

This model is a finetuned BERT (bert-base-uncased) model that predict different mental disorders.

• It is trained on a costume dataset of texts or posts (from Reddit) about general experiences of users with mental health problems.
• Dataset was cleaned and all direct mentions of the disorder names in the texts were removed.

It includes the following classes:

• Borderline
• Anxiety
• Depression
• Bipolar
• OCD
• ADHD
• Schizophrenia
• Asperger
• PTSD

Training

Train size: 90%
Val size: 10%

Training set class counts (text samples) after balancing:
Borderline: 10398
Anxiety: 10393
Depression: 10400
Bipolar: 10359
OCD: 10413
ADHD: 10412
Schizophrenia: 10447
Asperger: 10470
PTSD: 10489

Validation set class counts after balancing:
Borderline: 1180
Anxiety: 1185
Depression: 1178
Bipolar: 1219
OCD: 1165
ADHD: 1166
Schizophrenia: 1131
Asperger: 1108
PTSD: 1089

model-finetuning: bert-base-uncased

The following hyperparameters were used during training:

learning_rate: 5e-05
train_batch_size: 32
val_batch_size: 32
optimizer: AdamW
num_epochs: 2-3

Training results

F1 Score: 0.6797
AUC Score: 0.7942

Classification Report

Borderline:
Precision: 0.6682
Recall: 0.5923
F1-score: 0.6280

Anxiety:
Precision: 0.6620
Recall: 0.6497
F1-score: 0.6558

Depression:
Precision: 0.7261
Recall: 0.5424
F1-score: 0.6209

Bipolar:
Precision: 0.8055
Recall: 0.5233
F1-score: 0.6345

OCD:
Precision: 0.8200
Recall: 0.6532
F1-score: 0.7271

ADHD:
Precision: 0.8740
Recall: 0.6603
F1-score: 0.7523

Schizophrenia:
Precision: 0.8017
Recall: 0.6472
F1-score: 0.7162

Asperger:
Precision: 0.7368
Recall: 0.6570
F1-score: 0.6946

PTSD:
Precision: 0.8612
Recall: 0.5812
F1-score: 0.6940