How do I build a model to find the best Design of Experiments?

Benvar
Benvar New Altair Community Member
edited November 5 in Community Q&A
Hello, 
My experiment is very expensive and difficult test so many time. I would like to apply Rapieminer's Machine Learning to reduce times of experiments. How do I build a model to find the best Design of Experiments?
=================================================================================
I only have 10 sets of data are like: variable x1, x2, x3, x4, x5, x6, x7 and results like y1, y2, y3, y4, for exapmle,
1st experiment x1, x2, x3, x4, x5, x6, x7  and y1, y2, y3, y4
2nd experiment x1, x2, x3, x4, x5, x6, x7  and y1, y2, y3, y4
.....
10th experiment x1, x2, x3, x4, x5, x6, x7  and y1, y2, y3, y4
to predict 11th~20th and so on. 
I want to build model to predict 11th~20th sets of x1, x2, x3, x4, x5, x6, x7 and result y1, y2, y3, y4 closed to optimization target as possible as Molde can.
The optimization target of y1,y2,y3,y4  are like y1 => 0, y2 => maximum, y3 closed to 3 , y4 has minimum.
=================================================================================
After prediction, I could select 5 sets  x1, x2, x3, x4, x5, x6, x7 to test experiment based on suggesting score.
Is it possible done by Rapidminer?
Appreciate deeply for support. 

Best Answer

  • Mia_Smith
    Mia_Smith New Altair Community Member
    Answer ✓
    Certainly! In RapidMiner, import your 10 sets of data, then use a multi-objective regression algorithm that considers your optimization targets for y1, y2, y3, and y4. Train the model, predict the 11th to 20th experiments, and select the top 5 sets based on the optimization criteria.

Answers

  • Mia_Smith
    Mia_Smith New Altair Community Member
    Answer ✓
    Certainly! In RapidMiner, import your 10 sets of data, then use a multi-objective regression algorithm that considers your optimization targets for y1, y2, y3, and y4. Train the model, predict the 11th to 20th experiments, and select the top 5 sets based on the optimization criteria.