Practical Fama-French 5 Factors analysis: I combine PCA and MDS to reveal latent risk dimensions, then validate them on 25 size-value portfolios. The project shows when factor compression preserves model quality and how data geometry can inform economically meaningful investment decisions.

Time series analysis of U.S. housing starts using monthly data from the U.S. Census Bureau (1995-2025). The project explores seasonal patterns, decomposes the series, and evaluates Holt–Winters additive and multiplicative models. Forecast accuracy is assessed using MAE, MSE, MAPE, and AMAPE on an out-of-sample test set.

Time-series analysis and forecasting of zinc prices (1990-2025) using ARIMA models in R. The project includes data preprocessing, stationarity diagnostics (ADF, Box-Cox), autocorrelation analysis, model selection using AIC/BIC, and out-of-sample forecast evaluation with comparison to Holt exponential smoothing benchmarks.