Projects
Customer Churn Predictor
Engineered ML churn prediction models for a telecom provider, achieving 0.93–0.94 ROC-AUC and 0.90 accuracy via CatBoost. Key feature analysis revealed customer tenure as the primary driver, providing data-backed evidence to prioritize early-stage customer loyalty programs.
Automated Sentiment Analysis
Developed an intelligent system to automatically filter and classify reviews. I trained machine learning models with lemmatization using NLTK, spaCy, and BERT for sentiment analysis. As a result, I suggested using Logistic Regression with NLTK or spaCy lemmatization, due to its high F1 score (>0.85) and low prediction times, balancing accuracy and efficiency in production.
Taxi Demand
Prediction
Developed a predictive model to forecast airport taxi demand using hourly historical data (March–August 2018). The LGBMRegressor model achieved the best performance with an MSE of 40.06, enabling accurate prediction of demand spikes. These insights optimized driver allocation during peak hours and improved fleet availability.
Automated Used Car Price Prediction
Predicted the market value of used cars using machine learning, utilizing a dataset of 354,369 records. XGBoost and LightGBM were the standout models, both offering a strong balance between accuracy (RMSE 0.0790 and 0.0784) and execution speed (~7 seconds). XGBoost was selected for its superior performance in optimizing both quality and efficiency.
Customer Profiling & Identification
Developed an algorithm to identify target customers for marketing strategies at an insurance company. The analysis revealed a young, gender-balanced demographic with lower-middle incomes and small households. The model achieved an F1 score of 0.97, demonstrating high precision and effectiveness in customer segmentation.
Video Game Sales Analysis
Analyzed key patterns in video game sales for a store using historical data to forecast sales for the upcoming year and optimize marketing campaigns. Statistical analysis was used to reveal a moderate correlation (0.41) between sales and critical reviews, highlighting the dominance of Action games and Mature ratings as success factors.
