Dear Shokufe,

Generally, the relative importance or so-called contribution factor is obtained toward the following procedure:

First you need to calculate the average values of those responses (target values) obtained from the first level of your desired factor, (e.g., first factor). Redo the first step for the second, third, fourth, and the remaining levels of the concerning factor. Next, repeat the first two steps for the next factors, as well. Now, you can draw a table that includes the average response values for each level. The difference between the maximum and the minimum value of the average response values for each factor is called the “delta value”. The magnitude of the delta value implies the significance/impact of each factors on the target values. Having the delta values, you can simply measure the percentage contribution (e.g., relative importance) of your designated factors.

Regardless of employed modeling or DOE method, this is the general procedure of obtaining relative importance.

Regards,

I am agree with Mohamad Mohamadi-Baghmolaei :

See also ::: https://au.mathworks.com/matlabcentral/answers/466477-how-can-i-select-and-rank-the-input-variables-in-a-neural-net-created-with-the-neural-net-pattern-re

https://machinelearningmastery.com/feature-selection-with-real-and-categorical-data/

https://au.mathworks.com/help/stats/feature-selection.html

Since you are using MATLAB, then you can probably use Random Forests Regression. It allows you to automatically find predictor importance estimates in the property out of bag (OOB) score related to Delta Error. As such, you do not need to calculate anything by yourself.

See more at:

https://se.mathworks.com/help/stats/treebagger.html