karolzak/support-tickets-classification

Demonstrates three distinct ML approaches—Azure ML Studio's visual interface, CNTK deep learning, and scikit-learn with Azure ML Service—for multi-label ticket classification on a 50k+ real-world dataset. Implements complete preprocessing pipelines using NLTK and pandas for text cleaning and anonymization, then deploys trained models as Azure web services. Addresses practical challenges like severe class imbalance across seven classification targets (ticket_type, urgency, impact, category, etc.) with model selection via GridSearchCV hyperparameter tuning.

169 stars. No commits in the last 6 months.