biasedclassifier 0.2.0

No project description provided

Biased Classifier

Biased Classifier

Current version: 0.2.0

Install

Directly from PyPi servers:

pip install biasedclassifier

Use

Example using Random Forests from scikit-learn.

Assume X, y is a training set with three classes and two heavily inbalanced classes. In this case, we'd like to bias two classifiers into these subsets. We've decided that 0.3 and 0.2 proportions are enough for the minority classes (from smaller up)

from biasedclassifier import BiasedClassifier
from sklearn.ensemble import RandomForestClassifier

clf = BiasedClassifier(
    k=5, 
    p=[0.3, 0.2], 
    unbiased_classifier=RandomForestClassifier(), 
    knn_jobs=1
)

# Train
clf.fit(X,y)

# Obtain probabilities for each class
prob = clf.predict_proba(X)

# Predicted values
y_pred = clf.predict(X)

# Average accuracy score
score = clf.score(X, y)