1. # -*- coding:utf-8 -*-
    2. # @auth ivan
    3. # @time 20190131
    4. # @goal Test_F1-micro&F1-macro
    5. """
    6. https://www.cnblogs.com/techengin/p/8962024.html
    8. F1-score = 2*(P*R)/(P+R)
    10. 准确率(P) = TP/(TP+FP)
    11. 召回率(R) = TP/(TP+FN)
    13. 对于数据测试结果有下面4种情况:
    14. 真阳性(TP): 预测为正, 实际为正;
    15. 假阳性(FP): 预测为正, 实际为负;
    16. 假阴性(FN): 预测为负, 实际为正;
    17. 真阴性(TN): 预测为负, 实际为负;
    18. """
    19. """
    20. from sklearn.metrics import f1_score
    22.     average : string, [None, 'binary' (default), 'micro', 'macro', 'samples', \
    23.                        'weighted']
    24.         This parameter is required for multiclass/multilabel targets.
    25.         If ``None``, the scores for each class are returned. Otherwise, this
    26.         determines the type of averaging performed on the data:
    28.         ``'binary'``:
    29.             Only report results for the class specified by ``pos_label``.
    30.             This is applicable only if targets (``y_{true,pred}``) are binary.
    31.         ``'micro'``:
    32.             Calculate metrics globally by counting the total true positives,
    33.             false negatives and false positives.
    34.         ``'macro'``:
    35.             Calculate metrics for each label, and find their unweighted
    36.             mean.  This does not take label imbalance into account.
    37.         ``'weighted'``:
    38.             Calculate metrics for each label, and find their average, weighted
    39.             by support (the number of true instances for each label). This
    40.             alters 'macro' to account for label imbalance; it can result in an
    41.             F-score that is not between precision and recall.
    42.         ``'samples'``:
    43.             Calculate metrics for each instance, and find their average (only
    44.             meaningful for multilabel classification where this differs from
    45.             :func:`accuracy_score`).
    46. """
    48. from sklearn.metrics import f1_score, precision_recall_fscore_support
    49. y_true = [1, 1, 1, 1, 1, 2, 2, 2, 2, 3, 3, 3, 4, 4]
    50. y_pred = [1, 1, 1, 0, 0, 2, 2, 3, 3, 3, 4, 3, 4, 3]
    52. print(f1_score(y_true, y_pred, labels=[1, 2, 3, 4], average='micro'))
    53. # 0.6153846153846153
    55. p_class, r_class, f_class, support_micro = precision_recall_fscore_support(
    56.     y_true=y_true, y_pred=y_pred, labels=[1, 2, 3, 4], average=None)
    57. print(f_class.mean(), f_class)
    58. # 0.6041666666666666 [0.75       0.66666667 0.5        0.5       ]
    60. print(f1_score(y_true, y_pred, labels=[1, 2, 3, 4], average='macro'))
    61. # 0.6041666666666666