02 / PROJECTS
Selected work
Machine Learning
MATCH 01
Football Match Outcome Analysis & Prediction
Analyzed 1,941 matches across 6 European competitions to test whether home advantage still holds, and whether recent team form can predict match outcomes. Engineered rolling 5-match form features with strict leakage prevention, compared Logistic Regression against Random Forest, and chose a class-balanced model after finding raw accuracy masked poor performance on draws.
Accuracy 46.2%
Macro F1 0.43
Home Win Rate 42.9%
Web Scraping
MATCH 02
Book Store Analysis - Scraping, EDA, SQL & Dual Dashboards
Built a 1,000-row dataset entirely from scratch - scraping 50 pages with Python, then a second pass per book to extract categories. Caught and documented a real data quality issue rather than ignoring it, and found no meaningful link between price and rating, independently confirmed across Python, Power BI, and Tableau.
Books 1,000
Correlation r=0.028
BI Tools 2