Last updated: Jun 15, 2023
Logarithmic loss gives the mean of logarithms that target class probabilities (confidence) in Watson OpenScale. It is also known as Expected log-likelihood and is a measure of model performance.
Logarithmic loss at a glance
- Description: Mean of logarithms target class probabilities (confidence). It is also known as Expected log-likelihood.
- Default thresholds: Lower limit = 80%
- Default recommendation:
- Upward trend: An upward trend indicates that the metric is deteriorating. Feedback data is becoming significantly different than the training data.
- Downward trend: A downward trend indicates that the metric is improving. This means that model retraining is effective.
- Erratic or irregular variation: An erratic or irregular variation indicates that t The feedback data is not consistent between evaluations. Increase the minimum sample size for the Quality monitor.
- Problem type: Binary classification and multiclass classification
- Chart values: Last value in the timeframe
- Metrics details available: None
Do the math
For a binary model, Logarithmic loss is calculated by using the following formula:
-(y log(p) + (1-y)log(1-p))
Where p = true label and y = predicted probability
For a multi-class model, Logarithmic loss is calculated by using the following formula:
M
-SUM Yo,c log(Po,c)
c=1
Where M > 2, p = true label, and y = predicted probability
Learn more
Parent topic: Quality metrics overview