Closed acmiyaguchi closed 2 years ago
Classes are very imbalanced, leading to predictions that are skewed in favor of the majority class. One way to resolve this is to implement a very naive undersampling routine.
https://github.com/acmiyaguchi/birdclef-2022/blob/main/notebooks/2022-05-18-acm-training-data-per-species.ipynb https://github.com/acmiyaguchi/birdclef-2022/blob/main/notebooks/2022-05-21-acm-classifier-prediction-dist.ipynb
https://machinelearningmastery.com/random-oversampling-and-undersampling-for-imbalanced-classification/
Classes are very imbalanced, leading to predictions that are skewed in favor of the majority class. One way to resolve this is to implement a very naive undersampling routine.
https://github.com/acmiyaguchi/birdclef-2022/blob/main/notebooks/2022-05-18-acm-training-data-per-species.ipynb https://github.com/acmiyaguchi/birdclef-2022/blob/main/notebooks/2022-05-21-acm-classifier-prediction-dist.ipynb