Altair RISE

A program to recognize and reward our most engaged community members

Nominate Yourself Now!

Logistic Regression Losing 1 Polynomial attribute consistently

I have two polynomnials data columns, one being an age group that is has 5 attributes (e.g. 16-20, 21-30....) and another polynomial with 4 attributes (Tariff plan 1, 2, 3, 4). When applying the logistic regression model, one attribute does not show in the model. For example I lose Tariff Plan 2 and Age Group 31-40, every time I run the model.

In the cross validation ExampleSet there is no loss of data, but the logistic regression model does not show it.

Image: https://dl.dropboxusercontent.com/s/4ayya82tmo13jzh/javaw_s1dFXSTKQr.png

Image: https://dl.dropboxusercontent.com/s/samp638mfdt7orl/javaw_a9rr4ZQQYB.png

If I try to change them from polynomial to numerical I see all the attributes in the Logistic Regression, but one shows no analysis. Since it removes collinear columns.

Image: https://dl.dropboxusercontent.com/s/hglgsdt7iq158yf/javaw_HtzoXoBy0V.png

Find more posts tagged with

AI Studio

Regression

Cross Validation

Accepted answers

All comments

MartinLiebig

Hi,

the model uses an one-hot encoding removing one column. Actually you only need classes-1 columns to contain all the information. If you do not want this: Please use nominal to numerical before hand.

Best,

Martin

frederick

I have tried to use nominal to numerical, but as I stated above, the Logistic Regression Model removes the co-linear columns (Tariff Plan 4 and Age Group 16=20).

MartinLiebig

Did you try to remove the remove colinear option?

frederick

Yes, but I wish to use the removed collinear columns for analysis? Sorry, I'm a noob, but It is removing 1 of each of my nominal to numerical columns such as Complaints = True, Tariff Plan = 4, International Plan = No...

I want to use all of these removed attributes for analysis.