Micro-média é um método estatístico comumente empregado em tarefas de classificação multiclasse tasks within the field of Inteligência Artificial and Aprendizado de Máquina. It provides a way to assess the performance of a classification model by considering all classes together, rather than calculating metrics for each class separately. This approach is particularly useful when dealing with conjuntos de dados desequilibrados onde algumas classes podem ter significativamente mais amostras do que outras.
In micro-averaging, the true positives, false positives, and false negatives are summed across all classes before calculating the overall precision, recall, or F1 score. This means that each instance contributes equally to the final metric, regardless of which class it belongs to. The formula for micro-average precision (P) and recall (R) can be expressed as:
- Precisão na micro-média: P = (Soma dos Verdadeiros Positivos) / (Soma dos Verdadeiros Positivos + Soma dos Falsos Positivos)
- Recall na micro-média: R = (Soma dos Verdadeiros Positivos) / (Soma dos Verdadeiros Positivos + Soma dos Falsos Negativos)
Micro-average is particularly advantageous in scenarios where the focus is on overall model performance rather than on individual class performance. This can help in providing a more holistic view of a model’s capabilities, especially in applications such as classificação de imagens, text categorization, and speech recognition.