Bringing it all together
One of the engineers in your arrhythmia detection startup rushes into your office to let you know that there is a problem with the ECG sensor for overweight users. You decide to reduce the influence of examples with weight over 80 by 50%. You are also told that since your startup is targeting the fitness market and makes no medical claims, scaring an athlete unnecessarily is costlier than missing a possible case of arrhythmia. You decide to create a custom loss that makes each "false alarm" ten times costlier than missing a case of arrhythmia. Does down-weighting overweight subjects improve this custom loss? Your training data X_train
, y_train
and test data X_test
, y_test
are preloaded, as are confusion_matrix()
, numpy
as np
, and DecisionTreeClassifier()
.
This exercise is part of the course
Designing Machine Learning Workflows in Python
Hands-on interactive exercise
Have a go at this exercise by completing this sample code.
# Create a scorer assigning more cost to false positives
def my_scorer(y_test, y_est, cost_fp=10.0, cost_fn=1.0):
tn, fp, fn, tp = ____
return ____