How to increase accuracy of classification problem in r



Dear All,

my score for Loan Prediction is not going beyond 0.78.

I am using R

I used simple Logistic Regression for this.

May i know how should I improve my score


Hi @tpjoshi ,

Feature Engineering can be one of the ways by which you can improve your score. Figure out which features do not matter, and create more new features which you think might be useful in predicting the loan status for an individual.

You can also try using ensemble techniques. (Which usually work better than logistic regression)


Have a look at this article


Sharing what usually helps me.
Checking for multicolinearity. Use vif() and alias() to help in identifying the variables which do not contribute much to the model.
Next is the cut-off value, usually 0.5 but this many a times may not be the case, varying them helps.
eg. table(pred >0.5,test$Opened)
Yet another approach that was tried in SPSS though, is to change the methods for building the model - (Enter, Forward: Conditional, etc). So at times varying the model also helps.