This project focuses on predicting employee burnout using machine learning techniques.
I explored and prepared workplace and mental fatigue data, engineered relevant features, and trained multiple models to identify burnout risk patterns.
The final model was evaluated and deployed as an interactive Streamlit web application, enabling proactive insights to support employee well-being and data-driven HR decisions.
This project showcases five SQL projects demonstrating database design, advanced querying, performance optimization, and automation. I analyzed customer transactions, sales performance, and employee data, implemented views and stored procedures, and normalized databases to ensure consistency and scalability.
This project analyzes sales, profitability, and customer trends for a retail superstore using Power BI. I built interactive dashboards with KPIs, sales trend analysis, and regional/product breakdowns, enabling executives to track performance and make data-driven business decisions.
This project classifies 24 types of boats in Venice using deep learning. I explored a labeled image dataset, trained a CNN from scratch, and applied transfer learning with MobileNetV2. The final fine-tuned MobileNetV2 model achieved ~82% training accuracy and ~78% validation accuracy, providing a stable and reliable multi-class boat classifier.
This project demonstrates the use of computer vision and transfer learning to automate boat recognition for tourism, safety, research, and urban planning, while serving as a benchmark for real-world image classification tasks.
This project classifies news articles as fake or real using natural language processing. I cleaned and tokenized text data, extracted TF-IDF and sentiment features, and trained multiple models including Logistic Regression, Random Forest, Naive Bayes, SVM, and HistGradientBoosting. The models achieved reasonable accuracy, identifying linguistic patterns that distinguish fake from real news.
This project predicts insurance premiums using a dataset of over 200,000 records. I performed EDA, handled missing values, transformed skewed targets, engineered features, and trained multiple regression models including Linear Regression, Random Forest, HistGradientBoosting, LightGBM, and XGBoost. The best model was a tuned HistGradientBoostingRegressor on a log-transformed target.
This project classifies deliveries as early, on-time, or late using logistic classification. I performed EDA, feature engineering, encoding, scaling, outlier capping, and handled class imbalance using SMOTE. Ensemble voting and hyperparameter tuning improved classification performance.
This project analyzes the impact of marketing campaigns using A/B testing to optimize conversions. I compared conversion rates between users exposed to ads and a control group, applied statistical tests, and analyzed engagement patterns to identify the best times and strategies for ad delivery. The analysis confirmed that the ad campaign significantly increased conversions and provided actionable recommendations for future campaigns.
This interactive web app simulates the impact of inflation on economic indicators. Users can adjust inputs and immediately see effects on prices, purchasing power, and economic trends. The app is deployed with Streamlit, providing an engaging tool for exploring real-world inflation scenarios.
