0
$\begingroup$

I came across a multivariate machine learning problem in which I need to detect the biasing column and remove the bias of that particular column in predicting final target variable. train data-> [x1,x2,..xn,BiasColumn,y] (BiasColumn does not contain all the X columns. Its only a fixed subset of X). What should be the standard way to approach this problem?

$\endgroup$
2
  • $\begingroup$ I thought of an approach as: 1)Create classification model for detecting bias column 2)training a separate model for each value in BiasColumns. So in each model we will only take the training data subset having a particular biascolumn. We remove that column from X and train the model on y 3)for predictions we first detect the biascolumn and select the respective regression model for y $\endgroup$ – HarshR Jun 17 at 5:41
  • $\begingroup$ What exactly is meant by bias of a column? Each column is more useful or less useful at predicting a target. Any bias may come exclusively from the method that uses the columns to predict the target. Are you perhaps saying that your specific model performs better when a certain column is excluded from predictors? $\endgroup$ – Aleksejs Fomins Jun 17 at 10:46

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.