Train_39_Node_Net.py 文件源码

def evalData(z,test_set_y):
    " z- prediction test_set_y is the truth "
    diff=z-test_set_y
    fpr, tpr, thresholds = metrics.roc_curve(test_set_y.ravel(), z.ravel(), pos_label=1)
    auc=metrics.auc(fpr, tpr)
    ap=metrics.average_precision_score(test_set_y.ravel(), z.ravel())

    Q=test_set_y.shape[0]
    Pk10=0
    Pk20=0
    Pk30=0
    Pk50=0
    Pk37=0
    for i in range(Q):
        Pk10+=ranking_precision_score(test_set_y[i], z[i], k=10)
        Pk20+=ranking_precision_score(test_set_y[i], z[i], k=20)
        Pk30+=ranking_precision_score(test_set_y[i], z[i], k=30)
        Pk37+=ranking_precision_score(test_set_y[i], z[i], k=37)
        Pk50+=ranking_precision_score(test_set_y[i], z[i], k=30)
    Pk10=Pk10/Q
    Pk20=Pk20/Q
    Pk30=Pk30/Q
    Pk50=Pk50/Q
    Pk37=Pk37/Q
    cross=metrics.log_loss(test_set_y,z)
    print '\n'
    print 'AUC',auc,'MSE',np.mean((diff)**2),'Cross-entropy:',cross
    print 'Precision at k=10: ',Pk10,' k=20: ',Pk20,' k=30: ',Pk30,' k=50: ',Pk50, ' k=37: ',Pk37
    return Pk37