MAPE
A commonly used accuracy metric in regression tasks, computed as the average of the absolute difference between predicted and true values divided by the true value, multiplied by 100 percent and expressed as a percentage. Lower values indicate more accurate predictions, but it is not applicable when true values are close to zero.
Timeline (last 90 days)
The experiments were conducted on a synthetic oil and gas dataset, and when filtering using only sensor data, the downstream model's MAPE was 7.1%.
After adding operator text comments on top of the sensor data, the downstream model's MAPE dropped from 7.1% to 5.4%, a decrease of 1.7 percentage points.
MAPE is a commonly used accuracy metric in regression tasks, calculated as the absolute value of the difference between the predicted value and the true value divided by the true value, then averaged over all samples and multiplied by 100%; its limitation is that when the true value is close to zero it gets amplified or even undefined, so it is more applicable in industrial scenarios such as flow rate and production volume where values are always positive and of stable magnitude.
A MAPE decrease of 1.7 percentage points is a significant improvement for a soft-sensor model and typically corresponds to considerable economic benefit.
WAPE is a commonly used evaluation metric in demand forecasting, calculated as the sum of absolute errors across all samples divided by the sum of actual values and then multiplied by 100%; compared with MAPE it is more robust for low-sales or zero-sales items, and a WAPE below 30% is generally considered a good level.
All Facts (5)
The experiments were conducted on a synthetic oil and gas dataset, and when filtering using only sensor data, the downstream model's MAPE was 7.1%.
50%UnverifiedAfter adding operator text comments on top of the sensor data, the downstream model's MAPE dropped from 7.1% to 5.4%, a decrease of 1.7 percentage points.
50%UnverifiedMAPE is a commonly used accuracy metric in regression tasks, calculated as the absolute value of the difference between the predicted value and the true value divided by the true value, then averaged over all samples and multiplied by 100%; its limitation is that when the true value is close to zero it gets amplified or even undefined, so it is more applicable in industrial scenarios such as flow rate and production volume where values are always positive and of stable magnitude.
50%UnverifiedA MAPE decrease of 1.7 percentage points is a significant improvement for a soft-sensor model and typically corresponds to considerable economic benefit.
50%UnverifiedWAPE is a commonly used evaluation metric in demand forecasting, calculated as the sum of absolute errors across all samples divided by the sum of actual values and then multiplied by 100%; compared with MAPE it is more robust for low-sales or zero-sales items, and a WAPE below 30% is generally considered a good level.
50%