Mean Absolute Percent Error
What is mean absolute percent error (MAPE)?
MAPE, or mean absolute percentage error, is a commonly used performance metric for regression defined as the mean of absolute relative errors:
where N is the number of estimates (Et) produced by the regression model and actuals (At) from ground truth data that are being compared when determining the performance of the regression model. MAPE is sometimes expressed as a percentage.
Why is MAPE important?
MAPE’s advantage is that it can be expressed as a percentage, making it understandable to a general audience when applied in any domain. By contrast, other metrics that are not expressed in relative terms or as percentages usually require domain expertise and context to understand the significance of their numerical values.
Because of its formulation relative to actual values, however, MAPE has two main disadvantages:
- It is prone to division-by-zero errors when At = 0 for any t . For this reason, a small positive, constant term (for example or larger/smaller depending on the typical values of At) could be added to the denominator for numerical stability, or zero-valued actuals could be ignored when averaging.
- It is asymmetric. For the same prediction errors, smaller actual values cause the relative error to become larger. Therefore, estimated and actual values are not interchangeable in the formulation. To address that limitation, an alternative formulation, called “symmetric MAPE.” 
How C3 AI helps organizations use MAPE
MAPE is included as a scoring metric in the C3 Agentic AI Platform. C3 AI Energy Management, a product that reduces energy costs and improves building operations, includes several applications that require accurate forecasting of building resources, such as electricity consumption and chilled water production/consumption. Energy management customers often prefer to use MAPE to quantify the performance of AI models that make such forecasts.