What is the difference between log loss and accuracy in machine learning?

What is the difference between log loss and accuracy in machine learning?

Log loss measures the performance of a classification model where the prediction is a probability value between 0 and 1. Accuracy, on the other hand, is the ratio of correctly predicted instances to the total instances.

Why is log loss preferred over accuracy in some cases?

Log loss is preferred when you want to evaluate the model's probability outputs, especially in imbalanced datasets. It penalizes false classifications more than accuracy does.

Can you give an example of when to use log loss?

Sure! In a medical diagnosis model where predicting a rare disease, log loss helps ensure that the model is not just predicting the majority class.

How can I calculate log loss?

Log loss can be calculated using the formula: -1/N * Σ(y * log(p) + (1-y) * log(1-p)), where y is the true label and p is the predicted probability.

Are there any limitations to using log loss?

Yes, log loss can be sensitive to outliers and may not provide a complete picture of model performance, especially in certain contexts.