Skip to content

fedormlevin/Fedor-Portfolio

Repository files navigation

Fedor Levin's Data Science Portfolio

  • Extracted macroeconomic indicators from FRED
  • Selected features and trained Logistic Regression model
  • Achieved Area Under Curve: 0.94

image info

  • Created industry indexes based on S&P500 prises and market capitalization
  • Applied CPPI model using STIP ETF as a "safe" asset
  • Compared price returns of the combined portfolio to the "risky" (equity only) portfolio
  • Achieved significant max drawdown reduction (38% for "Energy Minerals")

image info

  • Created Logistic Regression models predicting Positive/Negative labels for Amazon reviews of acrylic markers
  • Engineered features using N-grams and Tfidf
  • Optimized using GridsearchCV to reach the best model
  • Achieved Area Under Curve: 0.90

image info