_TODO

• Standardize numeric columns in categorized data set. This will fix the condition number even though the VIF analysis confirmed the matrix was fine
• Test categorized data set with XGBoost
• Further investigation of XGBoost parameters, refine search space for tuning
• add lagged features e.g. wait_time_30m_ago
• add weather data e.g. temp, is_raining (does this data exist somewhere?)
• add lagged PCA scores from first exercise