Minimizing cost of churn with 2 machine learning models (Logistic Regression and XGBoost) for a telecommunications company. The project includes data preprocessing, model training, and evaluation. Cost reduction of >75% achieved vs. baseline.
Benchmark of three transformer models (SieBERT, RoBERTa and XLNet) for sentiment analysis on a dataset of customer-agent dialogues. The project includes data cleaning, a comparison of sequential and parallel NLP processing, and model evaluation. Best accuracy of >90% achieved with SieBERT.
