Concise project decks with clear takeaways—preview instantly or download a polished PDF to dive deeper.
Cleaned and feature-engineered a dataset of 1M+ U.S. flights (Jan–Feb 2024), conducted EDA across airports and days of the week, then trained a logistic regression model to predict delay probability. Found origin airport is the dominant predictor — JFK, LGA, and EWR exceeded 80% delay rates — while distance and departure time had minimal influence. Model achieved 65.2% accuracy on a held-out test set.
Analyzed 13,050 wine records to understand how type and geography affect price and quality. Sparkling wines performed best, Rosé worst. Used t-tests to validate regional differences and guide sourcing decisions.
Built regression models linking price to rating, vintage, and popularity. Found sparkling wines command premium prices. Created interactive Tableau dashboard to support pricing strategy.
Tested gender and seasonal payment patterns. Men carried $125 higher balances on average. September payments exceeded April by $15. Results shaped marketing and retention strategies.
Compared 26k shoe prices across retailers. ANOVA confirmed Walmart consistently undercuts competitors. Recommended targeted price adjustments and aggressive price-match messaging.
Modeled earnings per share using EBT, operating margin, and pre-tax margin. EBT proved the strongest predictor. Identified unexpected negative margin relationships worth investigating.
Predicted client debt levels using demographics and financial behavior. Identified key risk factors and multicollinearity issues. Enabled better targeting for consolidation offers.
Tidied and reshaped 50-state average tuition data (2004–2016) into long format, joined with U.S. Census regions, and produced five base visualizations revealing the Northeast's consistent cost premium over other regions.
Refined five tuition visualizations with titles, axis labels, reference lines, direct data labels, and on-plot annotations to transform raw charts into publication-ready figures that communicate regional cost trends clearly.
Joined 271k Olympic athlete records with NOC region data to analyze medal counts across regions. Built multi-series line plots and faceted medal-type breakdowns for the top 6 regions over 26 years.
Analyzed U.S. milk production (1970–2017) using regional scatter plots and a choropleth map by state. Measured a moderate positive correlation (r = 0.63) between state population and milk output, with noted data limitations.
Tested four nonlinear regression models (Quadratic, Lin-Log, Log-Lin, Log-Log) across 140 state universities to predict mid-career salary from starting salary. The Log-Log model outperformed all others with an adjusted R² of 0.7649 and the lowest standard error.
Built two regression models on ~24,000 credit card customers to examine how education and demographics influence credit limits and default probability. Found repayment behavior and age as key risk signals, and recommended collecting additional financial variables for stronger predictive power.