Modelevaluatie: ongebalanceerde classificatiemodellen

Machine Learning-sollicitatievragen oefenen in Python

Lisa Stuart

Data Scientist

Klasonbalans

• Categorische targetvariabele
  • Ongeveer evenveel observaties per klasse
  • Grote verschillen -> misleidende resultaten

Confusion matrix

¹ https://scaryscientist.blogspot.com/2016/03/confusion-matrix.html

Prestatiemetrics

¹ https://scaryscientist.blogspot.com/2016/03/confusion-matrix.html

Metrics uit de matrix

¹ https://scaryscientist.blogspot.com/2016/03/confusion-matrix.html

Resamplingtechnieken

• Oversample minderheidsklasse
• Undersample meerderheidsklasse
• LET OP: eerst train/test splitsen vóór resampling!

¹ https://www.svds.com/learning-imbalanced-classes/

Functies

Functie	retourneert
sklearn.linear_model.LogisticRegression	logistische regressie
sklearn.metrics.confusion_matrix(y_test,y_pred)	confusion matrix
sklearn.metrics.precision_score(y_test,y_pred)	precisie
sklearn.metrics.recall_score(y_test,y_pred)	recall
sklearn.metrics.f1_score(y_test,y_pred)	f1-score
sklearn.utils.resample(deny, n_samples=len(approve))	resampled

Laten we oefenen!

Machine Learning-sollicitatievragen oefenen in Python