An explainable NLP-based customer support intelligence framework built using TF-IDF and calibrated Support Vector Machines (SVM) on a large-scale real-world Twitter customer support dataset containing over 2.8 million interactions.
The system performs:
- Customer intent classification
- Churn-risk estimation
- Escalation analysis
- Retention recommendation generation
- Explainable customer support analytics
Accuracy Score : 0.9118
Weighted F1 : 0.9109
Customer Support on Twitter Dataset:
https://www.kaggle.com/datasets/thoughtvector/customer-support-on-twitter
pip install kaggle
kaggle datasets download -d thoughtvector/customer-support-on-twitter- Logistic Regression
- Linear SVM
- Calibrated SVM
- Python
- Scikit-learn
- Pandas
- NLP
- TF-IDF
Customer-Support-Intelligence-System/
│
├── customer_intelligence_system.ipynb
├── customer_intelligence_svm.pkl
└── tfidf_vectorizer.pkl
Deepam Mhatre
Sardar Patel Institute of Technology