Fewest Missed Cancers — hard Machine Learning in Code problem

Fewest Missed Cancers

In screening, the worst mistake is calling a malignant tumour benign. In this dataset the label is 1 for benign and 0 for malignant. A model "calls a tumour benign" when its probability of class 1 is greater than or equal to a threshold.

Write safest_model(names, threshold).

Load the data with load_breast_cancer(return_X_y=True, as_frame=True) and split it with train_test_split(X, y, test_size=0.25, random_state=0, stratify=y).
names lists which models to try, chosen from:
- "logistic": make_pipeline(StandardScaler(), LogisticRegression(max_iter=1000))
- "knn": make_pipeline(StandardScaler(), KNeighborsClassifier(n_neighbors=5))
- "bayes": GaussianNB()
- "forest": RandomForestClassifier(n_estimators=100, random_state=0)
For each named model, fit it on the training rows and take predict_proba(X_test)[:, 1].
Count its missed cancers: test rows it calls benign whose true label is 0.

Return a dict with each name in names mapped to its number of missed cancers (an int), plus one more entry "safest": the name with the fewest missed cancers. On a tie, pick the name that comes first in names.

Examples

names = ['logistic', 'knn']

threshold = 0.5

returns {'logistic': 3, 'knn': 6, 'safest': 'logistic'}

names = ['knn', 'bayes', 'forest']

threshold = 0.7

returns {'knn': 4, 'bayes': 5, 'forest': 2, 'safest': 'forest'}

Hints

1from sklearn.datasets import load_breast_cancer

2from sklearn.ensemble import RandomForestClassifier

3from sklearn.linear_model import LogisticRegression

4from sklearn.model_selection import train_test_split

5from sklearn.naive_bayes import GaussianNB

6from sklearn.neighbors import KNeighborsClassifier

7from sklearn.pipeline import make_pipeline

8from sklearn.preprocessing import StandardScaler

11def safest_model(names, threshold):

12 pass