art_rfc.py 文件源码

def train_classifier(x_train, y_train, x_cv, y_cv):
    clf = RandomForestClassifier(n_estimators=100)

    print 'starting fit'
    # excluding the patient_id column from the fit and prediction (patient_id?)
    clf.fit(x_train[::5], y_train[::5])
    print 'starting pred'

    y_pred = np.zeros(x_cv.shape[0])
    for i in xrange(4):
        y_pred[i::4] = clf.predict_proba(x_cv[i::4])[:, 1]

    if y_cv is not None:
        print roc_auc_score(y_cv, y_pred)

    return y_pred, clf