"Correlation Matrix/Polynomial Regression [SOLVED]"

king79

Hi,

I'd like to analyze the correlation between a product and the parameters which are needed for the production of this.

Example:

I have a set of machine parameters (~20) and another set of optical measured values (~300) from the item which the machine produces. In the tutorials I have seen so far there is only possible to compute the correlation matrix between all parameters/values. Because of the amount of values/parameters (in total over 320 for only one item), I want to minimize the computations. Therefore I want to ask is there a possibility to only compute the correlation between the parameters and the optical measured values and not the correlation between themselves (parameters to parameters and values to values) The same goes for Polynomial Regression.

Thanks in advance

MariusHelf

Hi,

you have to remove all but one of the machine parameters and define that one as label. Then use the operator Weight by Correlation to get correlation values between the optical values and the label. The same way you can calculate a Polynomial Regression to get the relation between the optical measured values and the label.

Repeat this for all machine parameters. If you are more advanced with RapidMiner you can use the Loop operators to automate this process.

Best regards,
Marius