Questions for project

[JBaldys]

1.) consolidate explanatory variables into sub-categories (i.e., all rate data is in one variable) - is this possible to do? Or just use 100+ explanatory variables? Which variables can we remove, if any? (perform linear regression and remove non-significant variables). Or.. fit tree model and retain only features with high variable importance. Also re-label returns / predictive variables to better, less confusing names. Also, apply uniformization

2.) separate data into training, validation, and test sets

3.) build model- how to incorporate rebalancing of portfolio / portfolio weights? re-allocate portfolio according to maximum Sharpe or IC? use classification for ^... if underperform, sell, if overperform, buy potentially transform target variables (if outperform, over threshold, then +1 = buy, if underperform over threshold, then -1 = sell, if within threshold, 0 = hold)

4.) measure performance of individual factors on model; compute financial performance stats

5.) calculate new weights for portfolio

[JBaldys]

Note: we predict under or over-perform, but we still need return data in order to compute annualized returns/standard deviation of the portfolio

[JBaldys]

5.) Add to EDA section- compute autocorrelation for all features (pg. 37 in factor investing for R book)