Aan de slagGa gratis aan de slag

Not all metrics agree

In the previous exercise you saw that not all metrics agree when it comes to identifying nearest neighbors. But does this mean they might disagree on outliers, too? You decide to put this to the test. You use the same data as before, but this time feed it into a local outlier factor outlier detector. The module LocalOutlierFactor has been made available to you as lof, and the data is available as features.

Deze oefening maakt deel uit van de cursus

Designing Machine Learning Workflows in Python

Cursus bekijken

Oefeninstructies

  • Detect outliers in features using the euclidean metric.
  • Detect outliers in features using the hamming metric.
  • Detect outliers in features using the jaccard metric.
  • Find if all three metrics agree on any one outlier.

Praktische interactieve oefening

Probeer deze oefening eens door deze voorbeeldcode in te vullen.

# Compute outliers according to the euclidean metric
out_eucl = ____(metric='euclidean').fit_predict(features)

# Compute outliers according to the hamming metric
out_hamm = ____(metric=____).fit_predict(features)

# Compute outliers according to the jaccard metric
out_jacc  = ____(____=____).____(features)

# Find if the metrics agree on any one datapoint
print(any(____ + ____ + ____ == ____))
Code bewerken en uitvoeren