フリー問題

問題文

A table has 140 columns. A data scientist plans to remove irrelevant fields before modeling. Which criterion for removal is sound?

選択肢

1. Remove the columns with any missing values, because a column that is not complete cannot be relied on.
2. Remove the columns that were added most recently, because recent columns have the least history and therefore the least value.
3. Remove columns that are constant, that are identifiers with no predictive meaning, or that become knowable only after the outcome has already occurred.
4. Remove the columns that have the weakest correlation with the target, because a weak correlation means the column cannot contribute anything the stronger columns do not already carry.