Automated crypto trading system update: better XGBoost buy predictions and pipeline improvements 

Back gain for an update on the crypto trading system!
Another week flew by, and this one brought some improvements for the trading system:
1) refactored model training and scoring processes from a manual one-by-one approach to functions and pipelines (automation)
2) tuned some hyperparameters for the XGBoost model to strike a balance between being too conservative (not buying) and too trigger-happy (buying false dips on a downward swing)
3) improved Tableau visuals by showing the actual close price as well as the original line (smoothed version of the closing price for each time interval)
Things are coming together .