Since the variable’s units are scaled to percentage units, which makes it easier to understand, the mean absolute percentage error (MAPE) is the most frequently used measure to forecast error [1]. The data should not have any extremes (including zeros) for the best results. In regression analysis and model evaluation, it is frequently used as a loss function.
How is the weighted mean absolute percentage error used?
WMAPE can be used to examine the average error over time between your forecasts and what actually transpires. WMAPE is typically used to compare forecasts over a longer period of time because it exhibits a general trend rather than a very precise day or hour of the forecast’s accuracy. WMAPE can also be used in conjunction with other forecasting metrics to shed light on the precision and adaptability of your data models.
What is weighted mean absolute percentage error?
WMAPE, or weighted mean absolute percentage error, is a way to assess the precision of financial and statistical projections in comparison to actual, or real, outcomes for a sample. For instance, if you forecasted selling five cars and actually sell five cars on that day, your WMAPE would be zero percent as there were no errors in your forecast. Your WMAPE would be 66 if three cars were sold. 6% because the forecast stated one outcome, but the real outcome was different The different parts of WMAPE are:
How is the weighted mean absolute percentage error calculated?
The formula for calculating the WMAPE is:
WMAPE = 1/n x (|Actual – Forecasted|) x 100 / |Actual|
Where the parts of the formula are:
For instance, if the forecasted and actual values are as follows:
The steps listed below may assist you in determining the WMAPE for your data set: Forecasted**Actual**53101555
1. Find all values for |Actual – Forecasted|
You can enter each forecasted and actual value into the equation to determine these values. For example:
|3 – 5| = 2|15 -10| = 5|5 – 5| = 0
This fulfills the requirement of the formula |Actual – Forecasted| and enables you to determine the relative importance of each value. Notably, the first equation changes the value of the negative two into the positive two by using the absolute value symbol |x|. This is due to the fact that the absolute value symbols are more concerned with a number’s distance from zero than its actual value. Here, negative two is two away from zero.
2. For each value, divide by the actual value
You can divide this data by the actual value once you’ve determined the absolute values of the actual and forecasted data. For example:
2 / 3 = 0. 665 / 15 = 0. 330 / 5 = 0.
This satisfies the part of the equation /|Actual|. Due to the use of the absolute value symbols, every result of the calculations is a positive number. With the aid of this, you can determine the weight of each value in your data set.
3. Multiply by 100 and divide by the actual value
Once the results of your earlier calculations have been determined, multiply each by 100. This meets the requirement of the equation x 100, ensures that all of your values are on the same scale as the actual values, and can assist you in determining the weight of each value. For example:
0. 66 x 100 = 660. 33 x 100 = 330 x 100 = 0.
After completing the aforementioned calculations, divide each result by the initial actual value. For example:
66 / 3 = 2233 / 15 = 2. 20 / 5 = 0.
The weights of each calculation are represented by these numbers, which are used in the subsequent steps. You can also assign weights to this step based on the crucial moments in your data set. For instance, you could give Monday a weight of six while giving the other days a weight of one if Monday is the best sales day of the week and all the other days are equal. This means that Monday has 60% of the weight and the other days all have 10% of the weight
4. Calculate the sum of actual values and the sum of weights
You can compute the sum of the actual values now that you have determined the weights for each set of values. To accomplish this, multiply each actual value by the others. For example:
3 + 15 + 0 = 18
Once you’ve determined the total of the actual values, you can combine your weights. For example:
22 + 2.2 = 24.2
5. Calculate the weighted mean absolute percentage error
The weighted mean absolute percentage error can then be determined by dividing the total weights by the total actual values. For example:
24.2 / 18 = 1.34
Converted into a percentage, this value is 1. 34% which is equal to the weighted mean absolute percentage error of the values This is a relatively low percentage error, meaning that the forecasting models were almost 100% accurate
Forecasting: Moving Averages, MAD, MSE, MAPE
FAQ
How do you calculate weighted mean absolute percentage error?
- Find all values for |Actual – Forecasted| …
- For each value, divide by the actual value. …
- Multiply by 100 and divide by the actual value. …
- Add up all of the weights and the actual values.
What is the difference between MAPE and WMAPE?
WMAPE and MAPE are different measures. Mean Absolute Percent Error (MAPE) is the simple average of the percent errors. WMAPE, which stands for Weighted Mean Absolute Percent Error, weighs the errors according to volume, making it more accurate and dependable. The calculation is unaffected by negative errors because this is all absolute error.
Which is better mad MSE or MAPE?
MSE is scale-dependent, MAPE is not. Therefore, MSE cannot be used to compare accuracy between time series with different scales. According to reports, managers understand percentages better than squared errors when used for business purposes, so MAPE is frequently preferred. MAPE can’t be used when percentages make no sense.
What does mean absolute percentage error mean?
The average or mean of the absolute percentage errors of forecasts is known as the mean absolute percentage error (MAPE). Actual or observed value less than forecasted value is how errors are defined. To calculate MAPE, percentage errors are added without regard to sign.
