MultiLabel Classifier Evaluation Metrics
This toolkit focuses on different evaluation metrics that can be used for evaluating the performance of a multilabel classifier.
Intro
The evaluation metrics for multi-label classification can be broadly classified into two categories:
- Example-Based Evaluation Metrics
- Label Based Evaluation Metrics
Metrics
- Exact Match Ratio (EMR)
- 1/0 Loss
- Hamming Loss
- Example-Based Accuracy
- Example-Based Precision
- Label Based Metrics
- Macro Averaged Accuracy
- Macro Averaged Precision
- Macro Averaged Recall
- Micro Averaged Accuracy
- Micro Averaged Precision
- Micro Averaged Recall
- α- Evaluation Score
Examples
from multilabel_eval_metrics import *
import numpy as np
if __name__=="__main__":
y_true = np.array([[0, 1], [1, 1], [1, 1], [0, 1], [1, 0]])
y_pred = np.array([[1, 1], [1, 0], [1, 1], [0, 1], [1, 0]])
print(y_true)
print(y_pred)
result=MultiLabelMetrics(y_true,y_pred).get_metric_summary(show=True)
License
The multilabel-eval-metrics
toolkit is provided by Donghua Chen with MIT License.
Reference
Evaluation Metrics for Multi-Label Classification