Keep the Useful Columns
Most of the columns of X are noise, and only some help predict y. Score every column by its F statistic for a linear relationship with y, keep the k best, and return their column numbers in increasing order.
Example 1
X [[0,0,0,0],[7,1,3,5],[4,2,6,1],[1,3,2,6],[8,4,5,2],[5,5,1,7],[2,6,4,3],[9,7,0,8],[6,8,3,4],[3,9,6,0],[0,10,2,5],[7,11,5,1],[4,12,1,6],[1,13,4,2],[8,14,0,7],[5,15,3,3],[2,16,6,8],[9,17,2,4],[6,18,5,0],[3,19,1,5],[0,20,4,1],[7,21,0,6],[4,22,3,2],[1,23,6,7],[8,24,2,3],[5,25,5,8],[2,26,1,4],[9,27,4,0],[6,28,0,5],[3,29,3,1]]
y [0,5,6,7,8,13,14,15,16,21,22,23,24,29,30,31,32,37,38,39,40,45,46,47,48,53,54,55,56,61]
k 1
returns [1]
y is twice column 1 plus a little noise. No other column has anything to do with it.
Constraints
30 <= len(X) <= 60- 4 to 8 columns,
1 <= k <= 3 - The
kbest columns stand clearly apart
Write it in your browser, against the clock. The model solution, stepped through line by line, comes after.
More scikit-learn problems
- Leak-Free ScoreHard
- Mixed ColumnsHard
- Pick the PenaltyHard
- Right Amount of BendHard
- Scored in Time OrderHard
- Spam or NotHard
- Tune Then TestHard
- Hold Some BackBeginner