Altair RISE

A program to recognize and reward our most engaged community members

Nominate Yourself Now!

Extract regression equation

Hi,

I'm building a model to predict a numeric continuous variable by using another numeric variable.

Logarithmic or polynomial equations better fit my data. They have given a lower error rate than linear regression, but I'm unsure how to extract the model equation (formula) from the model I built.

Here is the algorithm I used:

- Generalized Linear Model

- family: gaussian

- link: log

(Green label = prediction)

Image: https://us.v-cdn.net/6030995/uploads/editor/yo/ngpwlnu3kfop.png

Also, I am trying to use polynomial regression for my model since it shows the equation in the model output. However, the prediction is totally off, not sure if I made any mistake in the setting. (Green label = prediction)

Image: https://us.v-cdn.net/6030995/uploads/editor/qo/oe94b05c1vr6.png

Image: https://us.v-cdn.net/6030995/uploads/editor/a1/e6fs3ws8htdq.png

Thanks,

Find more posts tagged with

AI Studio

Accepted answers

Caperez

Hi @Louie,
If you click on the data option into the left menu, you can see a table with the attributes and coefficients. You can construct the model with the more significant variables using the p-value and significance level as criteria.

The attached image is from sample process for GLM regression.

Image: https://us.v-cdn.net/6030995/uploads/editor/o2/bkttzqafi2we.png

Best,

Cesar

All comments

Caperez

Hi @Louie,

There is a good guide about the use and interpretation of GLM. It was very useful for me.

https://statisticsbyjim.com/regression/interpret-coefficients-p-values-regression/

Best,

Cesar

Louie

Hi ceaperez,
Thank you for the reference.
I understand the GLM formula and interpretation. However, I'm not sure where to find the model equation I built which I am planning to apply the model into other application (by using the formula) to predict new testing data (e.g., label = a + b(x) + c (x) or label = a+b log (x)
. Also, I am wondering if the rapidminer GLM could automatically generate interaction terms when I have 2 or more numeric or binomial independent variables?
Thanks,

MartinLiebig

Hi,

maybe you want to check the Function Fitting operator.

cheers,

Martin

Louie

Thanks mschmitz,

The Function Fitting operator could help me for generating interaction terms.
However. I'm still not sure how to directly out up the equation from an GLM .

For example, I can find the equation directly from linear or polynomials model

Image: https://us.v-cdn.net/6030995/uploads/editor/05/dnxjltqr7hbh.png

but not with GLM

Image: https://us.v-cdn.net/6030995/uploads/editor/jg/rrn305pwk0zp.png

Caperez

Best,

Cesar