kindred.LogisticRegressionWithThreshold

class kindred.LogisticRegressionWithThreshold(threshold=0.5)[source]

A modified Logistic Regression classifier that will filter calls by a custom threshold, instead of the default 0.5. This allows for control of the precision-recall tradeoff, e.g. false positives versus false negatives.

Variables:
  • clf – The underlying LogisticRegression classifier
  • threshold – Threshold to use, should be between 0 and 1

Methods

__init__(threshold=0.5)[source]

Set up a Logistic Regression classifier that can use a different threshold for predictions and thereby be more lenient (lower threshold, false positives increase, false negatives decrease) or more conservative (higher threshold, false positives decrease, false negative increase).

Parameters:threshold (float) – Threshold to use, should be between 0 and 1
fit(X, Y)[source]

Train the classifier using the associated matrix X and classes Y. Class zero should represent no associated class.

Parameters:
  • X (sparse matrix) – Training vector
  • Y (matrix) – Associated class for each row of X
predict(X)[source]

Make predictions for the class of each row in X. Class zero should represent no prediction.

Parameters:X (sparse matrix) – Testing vector
Returns:Predictions of classes for each row in X
Return type:matrix
predict_proba(X)[source]

Calculate probabilities for the class of each row in X. Class zero should represent no prediction. Returns a matrix of probabilities

Parameters:X (sparse matrix) – Testing vector
Returns:Probabilities of classes for each row in X
Return type:matrix