Kebinnuil/suicidal_detection_model

Suicidal Detection System

This is a fine-tuned model based on a transformer architecture distilBERT for detecting suicidal intent or ideation in text. This model purpose is for text-classification in suicidal detection system.

Example output

Text Input	Label	Score
"I want to jump off this bridge"	Suicidal	0.89

Example

from transformers import pipeline, DistilBertTokenizer, DistilBertForSequenceClassification

tokenizer = DistilBertTokenizer.from_pretrained("Kebinnuil/suicidal_detection_model")
model = pipeline("text-classification", model="Kebinnuil/suicidal_detection_model")

result = model("I want to jump off the bridge")
print(result)

Training Metrics

The dataset was split into 80/10/10 for train/validation/test set. Table below shows the result of the model's training metrics.

Epoch	Training Loss	Validation Loss	Accuracy	AUC
1	0.442800	0.348061	0.838000	0.925000
2	0.304100	0.331631	0.850000	0.935000
3	0.261600	0.329701	0.851000	0.936000

Classification Report

Class	Precision	Recall	F1-score	Support
0	0.87	0.84	0.85	1211
1	0.84	0.87	0.86	1189

Accuracy: 0.86
Macro avg: Precision 0.86, Recall 0.86, F1-score 0.86
Weighted avg: Precision 0.86, Recall 0.86, F1-score 0.86
Total samples: 2400

Mapping Config

Please follow the config.json