Grievance Classification Model
This model classifies grievances into five distinct buckets:
- Label 0: Account number is not Correct
- Label 1: Installment not received
- Label 2: Others
- Label 3: convo_starter
- Label 4: convo_ender
Description of the Buckets
Account number is not Correct: This pertains to feedback or grievances related to the bank account number not being displayed correctly on the portal for farmers.
Installment not received: This pertains to feedback where the farmer has not received the installment or money for a given month.
Others: This covers a broad range of other grievances including, but not limited to:
- Gender being displayed incorrectly
- Online application pending approval
- Other payment-related issues
- Problem with Aadhaar correction
- Issue with biometric-based e-KYC
- Issue with OTP-based e-KYC
- Transaction failures
Convo_starter: Any phrase that could be a conversation starter like 'Hi','Hello','Namasthe'
Convo_ender: Any phrase that could signal the end of a conversation like 'Thank you','Goodbye'
Training code:
The code for training the model can be found here
Training Metrics
The following training metrics were observed over 15 epochs:
| Epoch | Loss | Accuracy | F1 Score |
|---|---|---|---|
| 1 | 1.0574 | 0.4242 | 0.4202 |
| 2 | 0.7780 | 0.7500 | 0.6952 |
| 3 | 0.5212 | 0.8939 | 0.8825 |
| 4 | 0.3214 | 0.9773 | 0.9767 |
| 5 | 0.1612 | 1.0000 | 1.0000 |
| 6 | 0.0883 | 1.0000 | 1.0000 |
| 7 | 0.0592 | 1.0000 | 1.0000 |
| 8 | 0.0418 | 1.0000 | 1.0000 |
| 9 | 0.0321 | 1.0000 | 1.0000 |
- Downloads last month
- 15,789
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social
visibility and check back later, or deploy to Inference Endpoints (dedicated)
instead.