Câu 21: RAI: Risk and Artificial Intelligence

A quantitative risk team is building a regression model to forecast operational losses. They observed that the features “equipment age” and “years since last replacement” have a correlation of 0.97. In addition, there is one data point with extremely high losses due to a rare catastrophic breakdown. Which of the following would be the best solution for addressing these two issues?

1. A. Apply regularization to eliminate the outlier and keep both correlated variables, since regularization automatically solves multicollinearity.
2. B. Keep both highly correlated variables to improve predictive accuracy and remove the extreme-loss observation since it increases model error.
3. C. Transform both correlated variables into categorical bins and treat the extreme-loss observation as missing data.
4. D. Remove one of the highly correlated variables to address multicollinearity and examine the extreme-loss observation. — đáp án hiện tại