📉Root Mean Squared Logarithmic Error

NEW

Also known as: RMSLE

Square root of the mean squared error on log-transformed targets, emphasizing relative differences.

DESCRIPTION

Root Mean Squared Logarithmic Error measures prediction error after applying a logarithmic transform to true and predicted values. It penalizes underestimation and overestimation proportionally, making it suitable for skewed, positive targets and growth processes. RMSLE is robust to large absolute errors when relative accuracy matters, common in demand forecasting and competitions. It assumes nonnegative targets and predictions (often with +1 shifts). Use RMSLE when multiplicative errors are more meaningful than additive ones.

ALTERNATIVE DEFINITIONS

scikit-learn

The mean squared logarithmic error (MSLE) measures the mean squared distance between the log-transformed true and predicted values.

Explains MSLE; RMSLE is the square root of MSLE.

Source

Kaggle Learn

RMSLE is the square root of the mean squared logarithmic error and is useful when you care about relative differences instead of absolute differences.

Emphasizes the relative-error perspective and use cases.

Source

RELATED TERMS

Mean Squared Logarithmic Error Mean Absolute Percentage Error Huber Loss Log Transformation Heteroscedasticity Regression Metrics Forecasting Skewed Targets

COMPETING TERMS

RMSE Mean Absolute Error

KNOWLEDGE AREAS

Data Science AI Mathematics Testing & Validation

TARGET AUDIENCE

Data Scientist AI Engineer MLOps Engineer Data Analyst

EXTERNAL SOURCES

• Wikipedia - Root-mean-square deviation
• scikit-learn - MSLE definition and usage in regression.
• Google ML Crash Course - Loss functions and when to transform targets.

TOP QUESTIONS

What is Root Mean Squared Logarithmic Error?

RMSLE is the root of mean squared error computed on log-transformed values, prioritizing relative over absolute differences.

Category: Definition

When should I use RMSLE instead of RMSE?

Use RMSLE when targets are skewed, positive, and relative error matters (e.g., demand, sales growth).

Category: Relevancy

What are the requirements for using RMSLE?

Targets and predictions must be nonnegative. A +1 shift is often applied before logging to handle zeros.

Category: Requirements

What is a common mistake with RMSLE?

Applying it to data with negatives or interpreting it as an absolute error measure. It reflects multiplicative deviations.

Category: Mistakes

How does RMSLE treat underestimation vs overestimation?

It penalizes relative differences symmetrically after log transform, making proportional errors more comparable.

Category: Opinion

How is RMSLE computed in Python?

Compute MSLE via sklearn.metrics.mean_squared_log_error and take the square root. Ensure inputs are nonnegative.

Category: How-To

What are alternatives to RMSLE?

RMSE and MAE are common. Consider log-transforming targets and using RMSE on log-space as an equivalent approach.

Category: Alternatives

Is RMSLE robust to outliers?

It reduces the influence of large absolute errors due to the log transform but is not outlier-proof.

Category: Risk

What is a simple analogy for RMSLE?

Think in percentages: doubling from 10 to 20 and from 100 to 200 represent similar multiplicative errors.

Category: Example