bedroc_score()

Complexity

Conditions

3

Size

Total Lines

38

Duplication

Lines	0
Ratio	0 %

#! /usr/bin/env python
def bedroc_score(y_true, y_pred, decreasing=True, alpha=20.0):

    """ BEDROC metric implemented according to Truchon and Bayley
     (10.1021/ci600426e).

     @param y_true      class labels, 1 for positive class, 0 otherwise
     @param y_pred      prediction values
     @param decreasing  :boolean: if high metric correlates to positive class
     @param alpha       early recognition parameter

     @returns float between 0 and 1, indicating degree to which the predictive
     technique employed detects (early) the positive class.
     """

    assert len(y_true) == len(y_pred), \
     'The number of scores must be equal to the number of labels'

    N = len(y_true)

The name N does not conform to the variable naming conventions ([a-z_][a-z0-9_]{2,30}$).

    n = sum(y_true == 1)

The name n does not conform to the variable naming conventions ([a-z_][a-z0-9_]{2,30}$).

    if decreasing:
        order = np.argsort(-y_pred)

Undefined variable 'np'

    else:
        order = np.argsort(y_pred)

Undefined variable 'np'

    m_rank = (y_true[order] == 1).nonzero()[0]

    s = np.sum(np.exp(-alpha * m_rank / N))

The name s does not conform to the variable naming conventions ([a-z_][a-z0-9_]{2,30}$).

Undefined variable 'np'

    r_a = n / N

    rand_sum = r_a * (1 - np.exp(-alpha))/(np.exp(alpha/N) - 1)

Undefined variable 'np'

    fac = r_a * np.sinh(alpha / 2) / (np.cosh(alpha / 2) - np.cosh(alpha/2 - alpha * r_a))

Undefined variable 'np'

    cte = 1 / (1 - np.exp(alpha * (1 - r_a)))

Undefined variable 'np'

    return s * fac / rand_sum + cte