What Would Flip the Answer? — medium Machine Learning in Code problem

What Would Flip the Answer?

The most useful explanation for a refused applicant is what would have had to change. A lender's model reads [income_thousands, debt_thousands] and predicts 1 (approve) or 0 (refuse).

Task: write income_to_flip(rows, decisions, applicant, incomes).

Fit make_pipeline(StandardScaler(), LogisticRegression(max_iter=1000)) on rows and decisions.
applicant is one [income, debt] row. incomes is a list of candidate incomes, sorted from smallest to largest.
Try the candidates in order. For each one, keep the applicant's debt, swap in the candidate income, and predict.
Return the first candidate income whose prediction differs from the applicant's own prediction.
If none of them changes it, return -1.

Examples

rows = [[85.7, 9.2], [31.7, 23.6], [75.4, 22.0], [51.9, 27.3], [26.5, 22.5], [89.4, 6.3], [85.7, 17.4], [60.4, 6.9], [22.1, 10.8], [27.7, 14.6], [57.2, 10.3], [63.2, 25.6], [71.3, 11.2], [23.4, 16.2], [90.0, 9.3], [54.5, 26.8], [19.2, 13.2], [38.0, 0.9], [21.1, 16.6], [50.3, 9.3], [70.5, 4.1], [41.7, 24.1], [46.5, 12.6], [88.1, 19.1], [61.4, 30.0], [22.0, 7.2], [50.8, 12.8], [76.9, 15.7], [58.0, 29.6], [43.6, 25.4], [55.0, 5.6], [27.2, 17.1], [21.0, 23.0], [20.9, 24.9], [85.4, 9.8], [42.6, 8.5], [57.5, 2.5], [25.4, 18.6], [21.4, 0.5], [39.8, 24.7]]

decisions = [1, 0, 1, 0, 0, 1, 1, 1, 0, 0, 1, 0, 1, 0, 1, 0, 0, 1, 0, 1, 1, 0, 0, 1, 0, 0, 0, 1, 0, 0, 1, 0, 0, 0, 1, 1, 1, 0, 0, 0]

applicant = [30, 10]

incomes = [30, 40, 50, 60, 70, 80]

returns 50

decisions = [1, 0, 1, 0, 0, 1, 1, 1, 0, 0, 1, 0, 1, 0, 1, 0, 0, 1, 0, 1, 1, 0, 0, 1, 0, 0, 0, 1, 0, 0, 1, 0, 0, 0, 1, 1, 1, 0, 0, 0]

applicant = [25, 2]

incomes = [26, 30, 34, 38, 42, 46]

returns 30

Hints

1from sklearn.linear_model import LogisticRegression

2from sklearn.pipeline import make_pipeline

3from sklearn.preprocessing import StandardScaler

6def income_to_flip(rows, decisions, applicant, incomes):

7 pass