|
|
--- |
|
|
language: en |
|
|
license: apache-2.0 |
|
|
base_model: distilbert-base-uncased |
|
|
tags: |
|
|
- text-classification |
|
|
- fraud-detection |
|
|
- transformer |
|
|
- distilbert |
|
|
- huggingface |
|
|
pipeline_tag: text-classification |
|
|
widget: |
|
|
- text: "We require an urgent refund for the suspicious transaction on our account." |
|
|
--- |
|
|
|
|
|
# 🕵️♂️ Fraud Model Aura (KS-Vijay) |
|
|
|
|
|
This model uses **DistilBERT** to classify whether a given grievance or complaint text contains **fraudulent intent or behavior**. It is trained as part of an intelligent **Grievance Redressal Platform** to auto-detect fraud-related issues in startup complaints. |
|
|
|
|
|
## 🧠 Use Case |
|
|
|
|
|
Detects if the complaint relates to fraud: |
|
|
- `Fraud` |
|
|
- `Legitimate` |
|
|
|
|
|
This helps startups or service providers to **flag, escalate, or triage suspicious reports** quickly. |
|
|
|
|
|
## 🔍 Model Summary |
|
|
|
|
|
- **Model Type:** Text Classification |
|
|
- **Architecture:** DistilBERT (uncased) |
|
|
- **Output Labels:** `Fraud`, `Legitimate` |
|
|
- **Weights Format:** `safetensors` |
|
|
- **Dataset:** Custom (based on complaints.csv) |
|
|
- **Training Framework:** PyTorch using 🤗 `transformers` |
|
|
|
|
|
## 📥 Example Input |
|
|
|
|
|
```text |
|
|
"I think someone is misusing our company’s KYC information to open fake accounts." |
|
|
|
