1st Place at DEU Statistics & Data Science Datathon (Datathon'25)
First place winner at the datathon competition organized by Dokuz Eylul University Statistics and Data Science Community (DEU IVBT), in collaboration with the Computer Science & AI Society and Econometrics Society.
- Date: April 25-26, 2025
- Format: 20-hour intensive hackathon
- Sponsors: Miuul, I-Akademi, Kent Lokantasi, Red Bull, Bono Coffee, DEPARK Bambu Girisim Ofisi, Unides
- Coaching: 2-week coaching program by Ilhami Demirci (Microsoft MCT, Academy Lead, Data Analyst, Power BI & MS Fabric)
Developed a location-based disease prediction model using U.S. time-series health data (CDC weekly mortality statistics, 2014-2023). The model predicts disease-specific mortality rates across U.S. states using geographic and temporal features.
- Data Processing: Cleaned and merged CDC mortality datasets covering 27,000+ weekly records across all U.S. states
- Feature Engineering: State-level geographic coordinates, temporal features (MMWR week/year), and disease-specific mortality ratios for missing data imputation
- Models: KNeighborsRegressor, MLPRegressor, RandomForestRegressor — trained per state and disease category
- Inference: PyTorch-based prediction pipeline for real-time health risk assessment
- Diseases Covered: Septicemia, Cancer, Diabetes, Alzheimer's, Influenza & Pneumonia, Chronic Respiratory Diseases, Heart Disease, Cerebrovascular Diseases, COVID-19, and more
Python, Pandas, NumPy, Scikit-learn, PyTorch, Matplotlib