ticket_cat_bert

TicketCat: IT Support Ticket Classification Model

This is a fine-tuned DistilBERT model for classifying IT support tickets into 9 categories.

Model Description

This model was fine-tuned on customer support tickets to automatically categorize incoming IT support requests. It uses DistilBERT as the base model and was trained to classify tickets into 9 distinct categories.

Intended Use

This model is designed to automatically categorize IT support tickets to help route them to the appropriate support team or department.

Categories

The model classifies tickets into the following categories:

• Account Access / Login Issues
• Billing & Payments
• Bug / Defect Reports
• Feature Requests
• General Inquiries / Other
• How-To / Product Usage Questions
• Integration Issues
• Performance Problems
• Security & Compliance

Usage

from transformers import DistilBertForSequenceClassification, DistilBertTokenizer
import torch

# Load model and tokenizer
model_name = "YOUR_HF_USERNAME/ticketcat-distilbert"  # Replace with your actual model name
tokenizer = DistilBertTokenizer.from_pretrained(model_name)
model = DistilBertForSequenceClassification.from_pretrained(model_name)

# Classify a ticket
ticket_text = "I can't access my account, password reset link is not working"
inputs = tokenizer(ticket_text, return_tensors="pt", padding=True, truncation=True, max_length=128)

with torch.no_grad():
    outputs = model(**inputs)
    probs = torch.softmax(outputs.logits, dim=1)
    predicted_class = torch.argmax(probs, dim=1).item()
    confidence = probs.max().item()

# Map to category name
categories = {'0': 'Account Access / Login Issues', '1': 'Billing & Payments', '2': 'Bug / Defect Reports', '3': 'Feature Requests', '4': 'General Inquiries / Other', '5': 'How-To / Product Usage Questions', '6': 'Integration Issues', '7': 'Performance Problems', '8': 'Security & Compliance'}
predicted_category = categories[str(predicted_class)]

print(f"Category: {predicted_category}")
print(f"Confidence: {confidence:.4f}")

Training Details

• Base Model: distilbert-base-uncased
• Framework: Hugging Face Transformers
• Task: Multi-class text classification
• Number of Classes: 9
• Max Sequence Length: 128 tokens
• Training Approach: Fine-tuning with class weights for imbalanced data

Limitations

• The model was trained on IT/customer support tickets and may not perform well on other domains
• Performance may vary on tickets that don't fit clearly into one category
• Low confidence predictions (< 0.6) may need human review

Citation

If you use this model, please cite:

@misc{ticketcat2024,
  author = {TicketCat Team},
  title = {TicketCat: IT Support Ticket Classification},
  year = {2024},
  publisher = {Hugging Face},
  howpublished = {\url{https://huggingface.co/YOUR_USERNAME/ticketcat-distilbert}}
}